Construction target monitoring became one of the key concerns of the construction field. Computer vision
based monitoring methods like 2D image, video and 3D laser scan methods have difficulties in meeting
real-time and data objectivity requirements. This paper points out the applicability of high resolution
remote sensing image (HRRSI) in this task. This paper proposes a way for enhancing Histogram of
Gradients (HOG) feature for HRRSI. HOG is applied as the key feature for machine learning, which is
carried out by the boosting method. To overcome the complexity of HOGs in HRRSIs, this paper builds a
feature framework according to the top, side and height characteristics of engineering ground objects.
Based on the feature framework, this paper presents a HRRSI based framework for construction progress
monitoring. Through an application case, the performance of the framework is presented. Judging from the
result, this method is robust under different seasonal and illuminative conditions.
Human Photogrammetry: Foundational Techniques for Creative Practitionersijcga
Photogrammetry has emerged as a leading approach for photorealistic digital replication and 3D scanning of real-world objects, particularly in areas of cinematic visual effects and interactive entertainment. While the technique generally relies on simple photography methods, the foundational practices for the field of human photogrammetry remain relatively undocumented.Human subjects are significantly more complex than still life, both in terms of photogrammetric capture, and in digital reproduction. Without the documentation of foundational practices for human subjects, there is a significant knowledge barrier for new creative practitioners to operate in the field, stifling innovation and adoption of the technique.Researchers and commercial practitioners currently working in this field continually distribute learnings and research outcomes. These learnings tend to centralise more on advanced practices such as capturing micro-geometry (skin pores), reflectance and skin distortion. However, the standard principles for building capture systems, considerations for human subjects, processing considerations and technology requirements remain elusive. The purpose of this research is to establish foundational practices for human photogrammetry systems. These practices encapsulate the underlying architectures of capture systems, through to necessary data processing for the 3D reconstruction of human subjects.Design-led research was used to construct a scale 21-camera system, designed for high-quality data capture of the human head. Due to its incredible level of surface complexity, the face was used to experiment with a variety of capture techniques and system arrangements, using several human subjects. The methods used were a result of the analysis of existing practitioners and research, refined through numerous iterations of system design.A distinct set of findings were synthesised to form a foundational architecture and blueprint for a scale, human photogrammetry multi-camera system. It covers the necessary knowledge and principles required to construct a production-ready photogrammetry system capable of consistent, high-quality capture that meets the needs of visual effects and interactive entertainment production.
Computer Vision Based 3D Reconstruction : A ReviewIJECEIAES
3D reconstruction are used in many fields starts from the object reconstruction such as site, and cultural artifacts in both ground and under the sea levels. The scientist are beneficial for these task in order to learn and keep the environment into 3D data due to the extinction. In this paper explained vision setup that is commonly used such as single camera, stereo camera, Kinect / Structured Light/ Time of Flight camera and fusion approach. The prior works also explained how the 3D reconstruction perform in many fields and using various algorithms.
Moving Object Detection for Video SurveillanceIJMER
Video surveillance has long been in use to monitor security sensitive areas such as banks, department stores, highways, crowded public places and borders. The advance in computing power, availability of large-capacity storage devices and high speed network infrastructure paved the way for cheaper, multi sensor video surveillance systems. Traditionally, the video outputs are processed online by human operators and are usually saved to tapes for later use only after a forensic event. The increase in the number of cameras in ordinary surveillance systems overloaded both the human operators and the storage devices with high volumes of data and made it infeasible to ensure proper monitoring of sensitive areas for long times. In order to filter out redundant information generated by an array of cameras, and increase the response time to forensic events, assisting the human operators with identification of important events in video by the use of “smart” video surveillance systems has become a critical requirement. The making of video surveillance systems “smart” requires fast, reliable and robust algorithms for moving object detection, classification, tracking and activity analysis.
Human Photogrammetry: Foundational Techniques for Creative Practitionersijcga
Photogrammetry has emerged as a leading approach for photorealistic digital replication and 3D scanning of real-world objects, particularly in areas of cinematic visual effects and interactive entertainment. While the technique generally relies on simple photography methods, the foundational practices for the field of human photogrammetry remain relatively undocumented.Human subjects are significantly more complex than still life, both in terms of photogrammetric capture, and in digital reproduction. Without the documentation of foundational practices for human subjects, there is a significant knowledge barrier for new creative practitioners to operate in the field, stifling innovation and adoption of the technique.Researchers and commercial practitioners currently working in this field continually distribute learnings and research outcomes. These learnings tend to centralise more on advanced practices such as capturing micro-geometry (skin pores), reflectance and skin distortion. However, the standard principles for building capture systems, considerations for human subjects, processing considerations and technology requirements remain elusive. The purpose of this research is to establish foundational practices for human photogrammetry systems. These practices encapsulate the underlying architectures of capture systems, through to necessary data processing for the 3D reconstruction of human subjects.Design-led research was used to construct a scale 21-camera system, designed for high-quality data capture of the human head. Due to its incredible level of surface complexity, the face was used to experiment with a variety of capture techniques and system arrangements, using several human subjects. The methods used were a result of the analysis of existing practitioners and research, refined through numerous iterations of system design.A distinct set of findings were synthesised to form a foundational architecture and blueprint for a scale, human photogrammetry multi-camera system. It covers the necessary knowledge and principles required to construct a production-ready photogrammetry system capable of consistent, high-quality capture that meets the needs of visual effects and interactive entertainment production.
Computer Vision Based 3D Reconstruction : A ReviewIJECEIAES
3D reconstruction are used in many fields starts from the object reconstruction such as site, and cultural artifacts in both ground and under the sea levels. The scientist are beneficial for these task in order to learn and keep the environment into 3D data due to the extinction. In this paper explained vision setup that is commonly used such as single camera, stereo camera, Kinect / Structured Light/ Time of Flight camera and fusion approach. The prior works also explained how the 3D reconstruction perform in many fields and using various algorithms.
Moving Object Detection for Video SurveillanceIJMER
Video surveillance has long been in use to monitor security sensitive areas such as banks, department stores, highways, crowded public places and borders. The advance in computing power, availability of large-capacity storage devices and high speed network infrastructure paved the way for cheaper, multi sensor video surveillance systems. Traditionally, the video outputs are processed online by human operators and are usually saved to tapes for later use only after a forensic event. The increase in the number of cameras in ordinary surveillance systems overloaded both the human operators and the storage devices with high volumes of data and made it infeasible to ensure proper monitoring of sensitive areas for long times. In order to filter out redundant information generated by an array of cameras, and increase the response time to forensic events, assisting the human operators with identification of important events in video by the use of “smart” video surveillance systems has become a critical requirement. The making of video surveillance systems “smart” requires fast, reliable and robust algorithms for moving object detection, classification, tracking and activity analysis.
Augmented Reality to Access Dam Cracks InformationIJERA Editor
Augmented reality technology is a tool with great potential for application in several areas, including
engineering. During the early phases of the enterprise life cycle, as design and construction, augmented reality
has been widely used. Concrete dams, during operation and maintenance phase, require regular inspections as
part of the necessary activities to keep their structural safety. Cracks are common pathology in these structures
that must be detected, mapped and measured as their evolution may evidence structure compromising, need for
maintenance or intervention. In this context this paper aims to present a prototype to access in situ the dam
cracks information through the use of augmented reality, thus giving a contribution to the effectiveness and
efficiency of inspections.
A Review on Overview of Image Processing Techniquesijtsrd
Image processing is actually among the fast growing innovations across various areas of a business with applications. Image processing frequently forms key scientific areas within the areas of electronics and computer science. Image processing is a tool for refining raw photographs obtained in our everyday lives from rockets, ships, space samples or military identification flights. Thanks to technologically powerful personal computers, broad databases of current devices and the Graphic Technology and the accessible resources for such software and apps, this area is strong and common. The provided input is an image and its output an enhanced high quality image according to the techniques used in the image processing procedure. Image processing is typically called digital image processing, although it is often possible to optically process and analogy photograph. An overview of image processing methods is given in this article. This article focuses mainly on identifying specific methods utilized in various image processing phases. Hirdesh Chack | Vijay Kumar Kalakar | Syed Tariq Ali "A Review on Overview of Image Processing Techniques" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-4 | Issue-5 , August 2020, URL: https://www.ijtsrd.com/papers/ijtsrd31819.pdf Paper Url :https://www.ijtsrd.com/engineering/electronics-and-communication-engineering/31819/a-review-on-overview-of-image-processing-techniques/hirdesh-chack
Novel hybrid framework for image compression for supportive hardware design o...IJECEIAES
Performing the image compression over the resource constrained hardware is quite a challenging task. Although, there has been various approaches being carried out towards image compression considering the hardware aspect of it, but still there are problems associated with the memory acceleration associated with the entire operation that downgrade the performance of the hardware device. Therefore, the proposed approach presents a cost effective image compression mechanism which offers lossless compression using a unique combination of the non-linear filtering, segmentation, contour detection, followed by the optimization. The compression mechanism adapts analytical approach for significant image compression. The execution of the compression mechanism yields faster response time, reduced mean square error, improved signal quality and significant compression ratio performance.
Framework on Retrieval of Hypermedia Data using Data mining Techniquerahulmonikasharma
Image Annotation is a method to reveal the meaning for a specific image .The embedded meaning in the image is identified and mined. The Scenario is identified through the image annotation scheme with in a provided training. The focus is on the blur images, noisy images and images with pixels lost. The image annotation can be done on the good resolution image. The analysis carried outon the image data to derive the information and image restoration takes place. Image mining deals with extracting embedded details, patterns and their relationship in images. Embedded details in the image could be extracted using high-level features that are robust. Inpainting techniques can be utilized for cleaning the image .The analytics is applied on enormous amount of data, techniques performed on the test images sets for better accuracy.
Efficient Image Compression Technique using Clustering and Random PermutationIJERA Editor
Multimedia data compression is a challenging situation for compression technique, due to the possibility of loss
of data as well as it require large amount of storage place. The minimization of storage place and proper
transmission of these data need compression. In this dissertation we proposed a block based DWT image
compression technique using genetic algorithm and HCC code matrix. The HCC code matrix compressed into
two different set redundant and non-redundant which generate similar pattern of block coefficient. The similar
block coefficient generated by particle of swarm optimization. The process of particle of swarm optimization is
select for the optimal block of DWT transform function. For the experimental purpose we used some standard
image such as Lena, Barbara and cameraman image. The size of resolution of this image is 256*256. The source
of image is Google.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
Satellite and Land Cover Image Classification using Deep Learningijtsrd
Satellite imagery is very significant for many applications including disaster response, law enforcement and environmental monitoring. These applications require the manual identification of objects and facilities in the imagery. Because the geographic area to be covered are great and the analysts available to conduct the searches are few, automation is required. The traditional object detection and classification algorithms are too inaccurate, takes a lot of time and unreliable to solve the problem. Deep learning is a family of machine learning algorithms that can be used for the automation of such tasks. It has achieved success in image classification by using convolutional neural networks. The problem of object and facility classification in satellite imagery is considered. The system is developed by using various facilities like Tensor Flow, XAMPP, FLASK and other various deep learning libraries. Roshni Rajendran | Liji Samuel "Satellite and Land Cover Image Classification using Deep Learning" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-4 | Issue-5 , August 2020, URL: https://www.ijtsrd.com/papers/ijtsrd32912.pdf Paper Url :https://www.ijtsrd.com/computer-science/other/32912/satellite-and-land-cover-image-classification-using-deep-learning/roshni-rajendran
COMPLETE END-TO-END LOW COST SOLUTION TO A 3D SCANNING SYSTEM WITH INTEGRATED...ijcsit
3D reconstruction is a technique used in computer vision which has a wide range of applications in
areas like object recognition, city modelling, virtual reality, physical simulations, video games and
special effects. Previously, to perform a 3D reconstruction, specialized hardwares were required.
Such systems were often very expensive and was only available for industrial or research purpose.
With the rise of the availability of high-quality low cost 3D sensors, it is now possible to design
inexpensive complete 3D scanning systems. The objective of this work was to design an acquisition and
processing system that can perform 3D scanning and reconstruction of objects seamlessly. In addition,
the goal of this work also included making the 3D scanning process fully automated by building and
integrating a turntable alongside the software. This means the user can perform a full 3D scan only by
a press of a few buttons from our dedicated graphical user interface. Three main steps were followed
to go from acquisition of point clouds to the finished reconstructed 3D model. First, our system
acquires point cloud data of a person/object using inexpensive camera sensor. Second, align and
convert the acquired point cloud data into a watertight mesh of good quality. Third, export the
reconstructed model to a 3D printer to obtain a proper 3D print of the model.
Transforming Construction Projects: The Crucial Role of Drone Mapping and Mod...Eye-bot
Are you wondering and asking, "Can we do the site inspection with drone technology? If so, let us tell you the pivotal role of drone mapping and modeling in construction sites for ensuring a multitude of critical measures.
Design and implement a reality-based 3D digitisation and modelling project3D ICONS Project
Design and implement a reality-based 3D digitisation and modelling project
Fabio Remondino, Fabio Menna, 3D Optical Metrology (3DOM) unit Bruno Kessler Foundation (FBK)
Anestis Koutsoudis, Christos Chamzas, Athena Research and Innovation Centre
Sabry El-Hakim, 4DHistory
Paper presented at Digital Heritage 2013
Augmented Reality to Access Dam Cracks InformationIJERA Editor
Augmented reality technology is a tool with great potential for application in several areas, including
engineering. During the early phases of the enterprise life cycle, as design and construction, augmented reality
has been widely used. Concrete dams, during operation and maintenance phase, require regular inspections as
part of the necessary activities to keep their structural safety. Cracks are common pathology in these structures
that must be detected, mapped and measured as their evolution may evidence structure compromising, need for
maintenance or intervention. In this context this paper aims to present a prototype to access in situ the dam
cracks information through the use of augmented reality, thus giving a contribution to the effectiveness and
efficiency of inspections.
A Review on Overview of Image Processing Techniquesijtsrd
Image processing is actually among the fast growing innovations across various areas of a business with applications. Image processing frequently forms key scientific areas within the areas of electronics and computer science. Image processing is a tool for refining raw photographs obtained in our everyday lives from rockets, ships, space samples or military identification flights. Thanks to technologically powerful personal computers, broad databases of current devices and the Graphic Technology and the accessible resources for such software and apps, this area is strong and common. The provided input is an image and its output an enhanced high quality image according to the techniques used in the image processing procedure. Image processing is typically called digital image processing, although it is often possible to optically process and analogy photograph. An overview of image processing methods is given in this article. This article focuses mainly on identifying specific methods utilized in various image processing phases. Hirdesh Chack | Vijay Kumar Kalakar | Syed Tariq Ali "A Review on Overview of Image Processing Techniques" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-4 | Issue-5 , August 2020, URL: https://www.ijtsrd.com/papers/ijtsrd31819.pdf Paper Url :https://www.ijtsrd.com/engineering/electronics-and-communication-engineering/31819/a-review-on-overview-of-image-processing-techniques/hirdesh-chack
Novel hybrid framework for image compression for supportive hardware design o...IJECEIAES
Performing the image compression over the resource constrained hardware is quite a challenging task. Although, there has been various approaches being carried out towards image compression considering the hardware aspect of it, but still there are problems associated with the memory acceleration associated with the entire operation that downgrade the performance of the hardware device. Therefore, the proposed approach presents a cost effective image compression mechanism which offers lossless compression using a unique combination of the non-linear filtering, segmentation, contour detection, followed by the optimization. The compression mechanism adapts analytical approach for significant image compression. The execution of the compression mechanism yields faster response time, reduced mean square error, improved signal quality and significant compression ratio performance.
Framework on Retrieval of Hypermedia Data using Data mining Techniquerahulmonikasharma
Image Annotation is a method to reveal the meaning for a specific image .The embedded meaning in the image is identified and mined. The Scenario is identified through the image annotation scheme with in a provided training. The focus is on the blur images, noisy images and images with pixels lost. The image annotation can be done on the good resolution image. The analysis carried outon the image data to derive the information and image restoration takes place. Image mining deals with extracting embedded details, patterns and their relationship in images. Embedded details in the image could be extracted using high-level features that are robust. Inpainting techniques can be utilized for cleaning the image .The analytics is applied on enormous amount of data, techniques performed on the test images sets for better accuracy.
Efficient Image Compression Technique using Clustering and Random PermutationIJERA Editor
Multimedia data compression is a challenging situation for compression technique, due to the possibility of loss
of data as well as it require large amount of storage place. The minimization of storage place and proper
transmission of these data need compression. In this dissertation we proposed a block based DWT image
compression technique using genetic algorithm and HCC code matrix. The HCC code matrix compressed into
two different set redundant and non-redundant which generate similar pattern of block coefficient. The similar
block coefficient generated by particle of swarm optimization. The process of particle of swarm optimization is
select for the optimal block of DWT transform function. For the experimental purpose we used some standard
image such as Lena, Barbara and cameraman image. The size of resolution of this image is 256*256. The source
of image is Google.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
Satellite and Land Cover Image Classification using Deep Learningijtsrd
Satellite imagery is very significant for many applications including disaster response, law enforcement and environmental monitoring. These applications require the manual identification of objects and facilities in the imagery. Because the geographic area to be covered are great and the analysts available to conduct the searches are few, automation is required. The traditional object detection and classification algorithms are too inaccurate, takes a lot of time and unreliable to solve the problem. Deep learning is a family of machine learning algorithms that can be used for the automation of such tasks. It has achieved success in image classification by using convolutional neural networks. The problem of object and facility classification in satellite imagery is considered. The system is developed by using various facilities like Tensor Flow, XAMPP, FLASK and other various deep learning libraries. Roshni Rajendran | Liji Samuel "Satellite and Land Cover Image Classification using Deep Learning" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-4 | Issue-5 , August 2020, URL: https://www.ijtsrd.com/papers/ijtsrd32912.pdf Paper Url :https://www.ijtsrd.com/computer-science/other/32912/satellite-and-land-cover-image-classification-using-deep-learning/roshni-rajendran
COMPLETE END-TO-END LOW COST SOLUTION TO A 3D SCANNING SYSTEM WITH INTEGRATED...ijcsit
3D reconstruction is a technique used in computer vision which has a wide range of applications in
areas like object recognition, city modelling, virtual reality, physical simulations, video games and
special effects. Previously, to perform a 3D reconstruction, specialized hardwares were required.
Such systems were often very expensive and was only available for industrial or research purpose.
With the rise of the availability of high-quality low cost 3D sensors, it is now possible to design
inexpensive complete 3D scanning systems. The objective of this work was to design an acquisition and
processing system that can perform 3D scanning and reconstruction of objects seamlessly. In addition,
the goal of this work also included making the 3D scanning process fully automated by building and
integrating a turntable alongside the software. This means the user can perform a full 3D scan only by
a press of a few buttons from our dedicated graphical user interface. Three main steps were followed
to go from acquisition of point clouds to the finished reconstructed 3D model. First, our system
acquires point cloud data of a person/object using inexpensive camera sensor. Second, align and
convert the acquired point cloud data into a watertight mesh of good quality. Third, export the
reconstructed model to a 3D printer to obtain a proper 3D print of the model.
Transforming Construction Projects: The Crucial Role of Drone Mapping and Mod...Eye-bot
Are you wondering and asking, "Can we do the site inspection with drone technology? If so, let us tell you the pivotal role of drone mapping and modeling in construction sites for ensuring a multitude of critical measures.
Design and implement a reality-based 3D digitisation and modelling project3D ICONS Project
Design and implement a reality-based 3D digitisation and modelling project
Fabio Remondino, Fabio Menna, 3D Optical Metrology (3DOM) unit Bruno Kessler Foundation (FBK)
Anestis Koutsoudis, Christos Chamzas, Athena Research and Innovation Centre
Sabry El-Hakim, 4DHistory
Paper presented at Digital Heritage 2013
Utilizing computer vision to change the “vision” of modern and complex constr...iFieldsmart Technologies
Construction vision is changing the landscape of onsite construction. This technology allows onsite construction personnel to measure and
analyzes objects in real-time and real life. Computer vision tools and technology use digital media like images and video to calculate the
incalculable. Construction sites can be complex and high-risk at the same time.
Onsite challenges.
Onsite personnel faces a myriad of challenges that include accidents, unsafe conditions, quality, snag checks, and onsite monitoring. In the
present, computer vision has become a crucial application that draws the true capabilities of Artificial Intelligence (AI) and Machine Learning.
This technology has immense potential in the construction industry based on object identification and recognition. It can also be integrated
with BIM processes and construction management software to assess work comparisons, evaluate video data in real-time, identifies onsite
work quality, and deviation from set work plans.
Using top construction management software from iFieldSmart Technologies can support onsite teams to streamline construction workflows
through various technology integrations.
A snapshot of computer vision benefits.
As mentioned earlier, computer vision is extremely beneficial to monitoring site safety through real-time camera footage, hardhat detection,
vest visibility, onsite goggles, etc. required for personnel working on high-altitude projects. The AI in computer vision can assess the absence of
onsite gear on workers, evaluate threats, etc. and caution site managers to take the necessary steps and avoid onsite accidents.
Live streaming data or footage from a camera, drone, or laser scanner can be evaluated to check for defects and quality problems. This
identification can help the design team make changes in the BIM model, or resolve As-Built challenges to save project time, money, and
resources. Computer vision can also help build 3D models to monitor progress, manage autonomous technology or equipment, etc. to monitor
site activities.
Optimized and efficient job sites.
Comprehensive inventory monitoring can be achieved by using HD cameras or aerial laser scanners. By analyzing data received from these
devices, project managers and contractors can track and modify analytics into actionable insights for ongoing job site processes. Computer
vision has great potential to optimize processes, mitigate bottlenecks, check for worker travel routes, and improve construction material
inventory.
Computer vision integrated with machine learning can reduce onsite idle hours leading to reduced project delays. Under-utilization or Overutilization of resources can be reduced, actionable insights can be leveraged, and real-time intelligence can be deployed.
Computer vision can drive data analysis through the cloud to cut down on costs. Progress tracking with 3D BIM models can help estimate
project completion timelines, planned vs actual progress, and more.
A Hardware Model to Measure Motion Estimation with Bit Plane Matching AlgorithmTELKOMNIKA JOURNAL
The multistep approach involving combination of techniques is referred as motion estimation.
The proposed approach is an adaptive control system to measure the motion from starting point to limit of
search. The motion patterns are used to analyze and avoid stationary regions of image. The algorithm
proposed is robust efficient and the calculations justify its advantages. The motivation of the work is to
maximize the encoding speed and visual quality with the help of motion vector algorithm. In this work a
hardware model is developed in which a frame of pictures are captured and sent via serial port to the system.
MATLAB simulation tool is used to detect the motion among the picture frame. Once any motion is detected
that signal is sent to the hardware which will give the appropriate sign accordingly. This system is developed
on two platforms (hardware as well software) to estimate and measure the motion vectors
Traffic Congestion Detection Using Deep Learningijtsrd
Despite the huge amount of traffic surveillance videos and images have been accumulated in the daily monitoring, deep learning approaches have been underutilized in the application of traffic intelligent management and control. Traffic images, including various illumination, weather conditions, and vast scenarios are considered and preprocessed to set up a proper training dataset. In order to detect traffic congestion, a network structure is proposed based on residual learning to be pre trained and fine tuned. The network is then transferred to the traffic application and retrained with self established training dataset to generate the Traffic Net. The accuracy of Traffic Net to classify congested and uncongested road states reaches 99 for the validation dataset and 95 for the testing dataset. The trained model is stored in cloud storage for easy access for application from anywhere. The proposed Traffic Net can be used by a regional detection of traffic congestion on a large scale surveillance system. The effectiveness and efficiencies are magnificently demonstrated with quick detection in the high accuracy in the case study. The experimental trial could extend its successful application to traffic surveillance system and has potential enhancement for intelligent transport system in future. Anusha C | Dr. J. Bhuvana "Traffic Congestion Detection Using Deep Learning" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-6 | Issue-2 , February 2022, URL: https://www.ijtsrd.com/papers/ijtsrd49401.pdf Paper URL: https://www.ijtsrd.com/computer-science/other/49401/traffic-congestion-detection-using-deep-learning/anusha-c
Embedded Implementations of Real Time Video Stabilization Mechanisms A Compre...YogeshIJTSRD
Video Stabilization has been widely researched and still is under active research considering the advancements that are being done in the field of digital imagery. Despite all the works, Hardware based Real time Video Stabilization systems especially works dealing with implementation of Technology on prototype Implementation boards have been very few. This review works focuses on the specific aspect of the modern day Stabilization systems implemented in prototyping boards and the algorithms that have been found suitable for such implementation taking considerations of cost, size and speed as the principal criterions. Mohammed Ahmed | Dr. Laxmi Singh "Embedded Implementations of Real Time Video Stabilization Mechanisms: A Comprehensive Review" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-5 | Issue-3 , April 2021, URL: https://www.ijtsrd.com/papers/ijtsrd39963.pdf Paper URL: https://www.ijtsrd.com/engineering/other/39963/embedded-implementations-of-real-time-video-stabilization-mechanisms-a-comprehensive-review/mohammed-ahmed
Similar to A FRAMEWORK FOR AUTOMATED PROGRESS MONITORING BASED ON HOG FEATURE RECOGNITION AND HIGH RESOLUTION REMOTE SENSING IMAGE (20)
A Strategic Approach: GenAI in EducationPeter Windle
Artificial Intelligence (AI) technologies such as Generative AI, Image Generators and Large Language Models have had a dramatic impact on teaching, learning and assessment over the past 18 months. The most immediate threat AI posed was to Academic Integrity with Higher Education Institutes (HEIs) focusing their efforts on combating the use of GenAI in assessment. Guidelines were developed for staff and students, policies put in place too. Innovative educators have forged paths in the use of Generative AI for teaching, learning and assessments leading to pockets of transformation springing up across HEIs, often with little or no top-down guidance, support or direction.
This Gasta posits a strategic approach to integrating AI into HEIs to prepare staff, students and the curriculum for an evolving world and workplace. We will highlight the advantages of working with these technologies beyond the realm of teaching, learning and assessment by considering prompt engineering skills, industry impact, curriculum changes, and the need for staff upskilling. In contrast, not engaging strategically with Generative AI poses risks, including falling behind peers, missed opportunities and failing to ensure our graduates remain employable. The rapid evolution of AI technologies necessitates a proactive and strategic approach if we are to remain relevant.
Francesca Gottschalk - How can education support child empowerment.pptxEduSkills OECD
Francesca Gottschalk from the OECD’s Centre for Educational Research and Innovation presents at the Ask an Expert Webinar: How can education support child empowerment?
Safalta Digital marketing institute in Noida, provide complete applications that encompass a huge range of virtual advertising and marketing additives, which includes search engine optimization, virtual communication advertising, pay-per-click on marketing, content material advertising, internet analytics, and greater. These university courses are designed for students who possess a comprehensive understanding of virtual marketing strategies and attributes.Safalta Digital Marketing Institute in Noida is a first choice for young individuals or students who are looking to start their careers in the field of digital advertising. The institute gives specialized courses designed and certification.
for beginners, providing thorough training in areas such as SEO, digital communication marketing, and PPC training in Noida. After finishing the program, students receive the certifications recognised by top different universitie, setting a strong foundation for a successful career in digital marketing.
Model Attribute Check Company Auto PropertyCeline George
In Odoo, the multi-company feature allows you to manage multiple companies within a single Odoo database instance. Each company can have its own configurations while still sharing common resources such as products, customers, and suppliers.
Introduction to AI for Nonprofits with Tapp NetworkTechSoup
Dive into the world of AI! Experts Jon Hill and Tareq Monaur will guide you through AI's role in enhancing nonprofit websites and basic marketing strategies, making it easy to understand and apply.
2024.06.01 Introducing a competency framework for languag learning materials ...Sandy Millin
http://sandymillin.wordpress.com/iateflwebinar2024
Published classroom materials form the basis of syllabuses, drive teacher professional development, and have a potentially huge influence on learners, teachers and education systems. All teachers also create their own materials, whether a few sentences on a blackboard, a highly-structured fully-realised online course, or anything in between. Despite this, the knowledge and skills needed to create effective language learning materials are rarely part of teacher training, and are mostly learnt by trial and error.
Knowledge and skills frameworks, generally called competency frameworks, for ELT teachers, trainers and managers have existed for a few years now. However, until I created one for my MA dissertation, there wasn’t one drawing together what we need to know and do to be able to effectively produce language learning materials.
This webinar will introduce you to my framework, highlighting the key competencies I identified from my research. It will also show how anybody involved in language teaching (any language, not just English!), teacher training, managing schools or developing language learning materials can benefit from using the framework.
Embracing GenAI - A Strategic ImperativePeter Windle
Artificial Intelligence (AI) technologies such as Generative AI, Image Generators and Large Language Models have had a dramatic impact on teaching, learning and assessment over the past 18 months. The most immediate threat AI posed was to Academic Integrity with Higher Education Institutes (HEIs) focusing their efforts on combating the use of GenAI in assessment. Guidelines were developed for staff and students, policies put in place too. Innovative educators have forged paths in the use of Generative AI for teaching, learning and assessments leading to pockets of transformation springing up across HEIs, often with little or no top-down guidance, support or direction.
This Gasta posits a strategic approach to integrating AI into HEIs to prepare staff, students and the curriculum for an evolving world and workplace. We will highlight the advantages of working with these technologies beyond the realm of teaching, learning and assessment by considering prompt engineering skills, industry impact, curriculum changes, and the need for staff upskilling. In contrast, not engaging strategically with Generative AI poses risks, including falling behind peers, missed opportunities and failing to ensure our graduates remain employable. The rapid evolution of AI technologies necessitates a proactive and strategic approach if we are to remain relevant.
A FRAMEWORK FOR AUTOMATED PROGRESS MONITORING BASED ON HOG FEATURE RECOGNITION AND HIGH RESOLUTION REMOTE SENSING IMAGE
1. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
DOI:10.5121/ijcsit.2017.9601 1
A FRAMEWORK FOR AUTOMATED PROGRESS
MONITORING BASED ON HOG FEATURE
RECOGNITION AND HIGH RESOLUTION REMOTE
SENSING IMAGE
Xu Ruhang 1,2
1
School of Economics and Management, North China Electric Power University,
Beijing, PCR
2
Beijing Key Laboratory of New Energy and Low-Carbon Development (North China
Electric Power University), Changping Beijing, 102206;
ABSTRACT
Construction target monitoring became one of the key concerns of the construction field. Computer vision
based monitoring methods like 2D image, video and 3D laser scan methods have difficulties in meeting
real-time and data objectivity requirements. This paper points out the applicability of high resolution
remote sensing image (HRRSI) in this task. This paper proposes a way for enhancing Histogram of
Gradients (HOG) feature for HRRSI. HOG is applied as the key feature for machine learning, which is
carried out by the boosting method. To overcome the complexity of HOGs in HRRSIs, this paper builds a
feature framework according to the top, side and height characteristics of engineering ground objects.
Based on the feature framework, this paper presents a HRRSI based framework for construction progress
monitoring. Through an application case, the performance of the framework is presented. Judging from the
result, this method is robust under different seasonal and illuminative conditions.
KEYWORDS
Network Object recognition, HOG feature, High resolution remote sensing image, Progress monitoring,
Construction project
1. INTRODUCTION
Construction target monitoring became one of the key concerns of the construction field.
Generally, monitoring targets include construction targets which may have critical influences
inside and outside the construction projects, like progress, quality, safety and environment. Some
researchers acquire real-world construction site data through several kinds of sensors, like RFIDs
(radio frequency identification devices), CSSs (chirp spread spectrum) and GPSs (global
positioning system) [1-4]. The computer vision technology also shows great potential in
construction target monitoring applications. Firstly, most of the computer vision data acquiring
ways are not expensive. Secondly, vision data can provide information not only about building
objects, but also about moving workers, moving equipment, occupied space and so on [5-7].
The on-site construction progress monitoring can benefit project control by detecting and
correcting progress deviations. There are 3 major technical approaches for computer-vision-based
construction target monitoring.
2. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
2
1.1.2D IMAGE BASED
Xiaonan et al. emphasized that the computer-vision-based construction progress automation
system can greatly help not only progress but also cost control [8]. Photographs are combined
with BIM to generate as-built 3D model of interior construction works. Seungjun et al. proposed a
way to build this kind of model by applying the photograph-based object recognition algorithm
[9]. Quality monitoring, the quality of the tile construction work is assessed by analyzing
computer-vision geometric characteristics [10]. Kevin et al. emphasized the importance of the on-
site operation-level progress monitoring. It is claimed that the operation-level real-time progress
monitoring can reveal engineering details and provide sufficient deviation information for project
control. They combined BIM with site photo logs to create a detailed real-time on-site progress
model [11]. Firstly, the BIM plan should be built. To monitor the interior progress of the
construction works, the inside BIM model should be fully built. Secondly, objects in the real
world should be detected and attached with corresponding plan objects. Normally, this progress is
done by applying computer vision methods based on photographs or videos. Thirdly, the progress
is analyzed by comparing the BIM-plan with photograph-real-world information [12]. Real-time
3D modeling of built objects is another way to achieve real-time progress monitoring. Pablo et al.
proposed an as-built 3D reconstruction method by processing 2D images acquired from an
unmanned aerial system [13].
1.2. VIDEO BASED
To perform progress tracking, some researchers paid attention to moving object tracking other
than building part tracking like worker and equipment movement tracking. Gong et al. proposed a
worker movement tracking method through the vision analysis based on video and photograph
[14]. Man-Woo et al. carried out a comparative study on the vision tracking methods for
construction site object tracking. They categorized 2D vision tracking methods into contour-
based, kernel-based and point-based methods. They concluded that for current technical situation,
the kernel-based 2D tracker may be the most suitable one for construction site object tracking
[15]. Man-Woo combined the background subtraction method, histogram of oriented gradients
and the HSV color histogram together and proposed a construction site worker tracking method
[16]. Video-based analysis was also applied in quality control. Mahmoud et al. applied a video
processing method to detecting defects in the city sewer system.
1.3. 3D LASER SCAN BASED
Some researchers applied laser scanners to acquiring the as-built raw data. Frédéric presented a
3D computer-aided-design object recognition method in laser scanning results and emphasized
the advantage of this method in construction dimensional compliance control [17]. Chao et al.
proposed a way to automatically extract building geometries from unorganized laser-scanning
point clouds, and attach them to as-planned BIM objects, so as to detect progress deviations [18].
As for 3D as-built object reconstruction, there are two main approaches: the image-based and the
laser-scan-based. Mani et al. pointed out that the former may not be as accurate as the latter in
defect detection or alignment tasks. They also concluded that the former has an advantage in
operation cost [19]. Some researchers considered the possibility of combining 2D and 3D sensing
technologies to build a multifunctional progress monitoring system, which is still in its initial
stage yet [20-21].
It can be summarized from previous studies in the field that each kind of technology has its
advantages and disadvantages. 2D image-based construction as-built modeling is less expensive.
Compared with 3D laser scanners, 2D image sensors with satisfying technical parameters are far
less expensive. At the same time, the usage of 2D image sensors, such as a digital camera, is
3. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
3
simpler than a 3D laser scanner, which requires operation training. As mentioned in literatures,
2D real-world images can provide more information than a simple 3D point cloud. The
information of elements like shadows, surface textures, movements and specific patterns can be
provided in a visible light exposure 2D image, which can be significant for modeling. The
deployment of a 2D image as-built sensing system may be simpler than a 3D-laser-scanning one
because 2D-based image monitoring system application is more mature in many fields like
security monitoring.
At the same time, the disadvantages of a 2D image-based approach are obvious. The information,
especially distance information provided in a 2D photograph is less accurate. Due to the light
sensing mechanism of a 2D image sensor, the results can be interfered by many factors like
illumination, view point and temperature. The advantages of a laser-scanning based monitoring
method are obvious. Firstly, 3D laser scanning can acquire precise coordinates of each point,
which is significant in quality deviation monitoring. Secondly, 3D-scanning results can be
directly transformed into a 3D model. With the precise coordinates, this 3D modeling way can be
faster and more accurate than a 2D-way. Besides, an as-built 3D model can provide visualized
construction process information to project managers and customers. However, disadvantages
came along. 3D-scanning equipment is much more expensive than a 2D camera under current
technical conditions. What’s more, 3D scanning maybe more difficult to deploy and provide
mostly position information at the same time.
In general, there are obstacles for both the 2D and the 3D ways in practical construction
monitoring application. A deployment solution should be provided for the whole construction
period. Actually, monitoring equipment is hard to be fixed in one location during the construction
process. Because the interior environment changes a lot while the engineering proceeds. If the
data cannot be automatically acquired by previously deployed equipment, a manned acquiring
way is the only option, which can still be error-prone and costly.
Practically, there is virtually no proven deployment solution for either a 2D or 3D monitoring
way for a practical entire construction process. It can be concluded that there are obstacles for
data acquisition when the size of the demand data granularity is too small. In fact, in practical
applications, some monitoring tasks only require a bigger size of data granularity.
The project monitoring in the Chinese government-leading investment is a good example. In last
decades, the government’s investment in fixed assets played an important role in economy
simulation in China. Great amount of GDP growth has been stimulated by the government fixed
assets investment since 2000. To handle the international economy crisis, the Chinese
government decided to carry out a 4,000 billion RMB investment in 2008, mostly in public
infrastructures. Although Public Private Partnership became population in recent years, there is
no doubt that government-leading fixed assets investment plays a significant role in the economy
development. There is a unique character of government-leading fixed asset investment. That is
the public resource utilization. This kind of utilization comes in many forms, like public finance,
land use and franchise. So, the utilization process must be monitored so as to guarantee that
public resources are put into proper use.
In fact, problems came along with the government investment in China in last decades’ practices.
Some local projects start construction before legal procedures, which means unauthorized public
resource occupation. Even in some authorized projects, misuse of the public resources happens.
For example, land-use ways can be changed. In the practice, some developers change some of the
development targets and made them more profitable for themselves. For example, they may add a
commercial building in the development plan on the land which they were authorized to construct
for public benefit. In some cases, the developer changed the construction target, to make it better-
built meanwhile more expensive for their own benefit. Some targets were changed during the
4. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
4
construction procedure, so that the government had to increase the budget. For the government,
there can be great amount of projects in construction, and these projects vary in stages, types and
locations, which makes the monitoring very difficult [22].
The structure of this paper is as follows. Section 1 is the background introduction. Section 2
discusses the applicability of HRRSI in progress monitoring of construction projects. Section 3
gives the feature framework for the engineering ground object recognition task in a HRRSI based
on HOG feature. Section 4 gives the progress monitoring framework of a construction project.
Section 5 presents an example of the methods mentioned in this paper. Section 6 gives the
conclusion.
2. THE APPLICABILITY OF HIGH RESOLUTION REMOTE SENSING IMAGE IN
PROGRESS MONITORING OF CONSTRUCTION PROJECTS
At present, some problems still exist in nowadays technical framework for effective monitoring
of construction projects, as shown below:
Real-time data acquisition. Traditional monitoring methods have a common problem, that is,
how to effectively obtain related data in real time. In most of these methods, data need to be
subjectively measured, which creates two derivative problems. Firstly, if the data are
provided by the under-monitoring entity, the authenticity of the data is difficult to be
guaranteed. Secondly, if the supervisor uses the method of manual data collection, it will
face a lot of data acquisition costs and it is too difficult to achieve the real-time performance.
Data objectivity. There is a major problem with current monitoring that the use of
information asymmetry by the supervised may lead to the damage to project owner's interests
and further cause a series of conspiracy problems. Besides, many of the indicators in
traditional monitoring approaches are subjective. Thus, there is a common problem in the
monitoring approach outlined in this section. As long as the subjective data acquisition is
involved, the original problem will perhaps become a more complex game problem.
Therefore, using subjective data for monitoring has inherent defects.
The combination of the Computer Vision (CV) technology and High Resolution Remote Sensing
Image (HRRSI) is suitable for the above problem scenarios. The phrase "high resolution" here
refers specifically to "high spatial resolution" and "high spectral resolution". For example, the
spatial resolutions of American IKONOS, QuickBird and WorldView-02B satellite panchromatic
image have respectively reached 1m, 0.61m, 0.5m. The adaptability is manifested in the
following three aspects:
(1) Satellite data has a variety of application scenarios. The monitoring application is only a small
part of its application value. Using high-resolution satellite remote sensing data for construction
project monitoring has the advantage of low cost from the cost-sharing perspective.
(2) Satellite remote sensing data has a natural advantage in wide geographical distributed project
monitoring. Supervisors do not need to install and maintain any sensing equipment at the project
site, nor do they need to send personnel to the site to collect data. They only need to process the
data sent by the satellite, which will greatly improve the monitoring efficiency.
(3) Satellite remote sensing data meets the monitoring requirements of the data size. Supervisors
do not need to deal with too specific project data.
The technical principle of the monitoring framework based on the satellite remote sensing image
and the CV processing method will be discussed in Section 3.
5. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
5
3. THE MONITORING OBJECT RECOGNITION METHOD BASED ON THE
ENGINEERING APPARENT VISION PROCESS FEATURE
3.1. ENHANCED HOG FEATURE ALGORITHM FOR GROUND OBJECT
HOG refers to Histogram of Oriented Gradient. Navneet Dalal and Bill Triggs firstly mentioned
this method for pedestrian detection in [25]. This method presents the feature of a certain object
in an image through the histogram of different gradient orientations [26].
The basic steps of the method are described as follows. First, an image is grayscale and
normalization processed and the gradients of the image is calculated. Then, the image is divided
into cells by certain window, for example, 16 16× or 32 32× . The histograms of gradients in cells
can be calculated. The feature vector of a cell is the weighted average of gradients in different
directions. Next, certain amount of cells is combined into blocks. The block connect feature
vector of cells in the block is normalized. Feature vector of blocks in the image is the HOG
feature of the image.
The algorithm of feature vector F for a certain image ( , )I x y ( x [1,m]∈ , y [1,n]∈ ) is described in
Algorithm 1.
Algorithm 1: HOG feature for ground object
Input: ( , )I x y
Output: F
I(x,y) [ 1,0,1]← ⊗ −Iy
I(x,y) [ 1,0,1]← ⊗ − T
Ix
←w window width
←o 360/9
for 1: /←i m w do
for 1: /←j n w do
←xw x direction gradient in the window
←pw phase in the window
for 1:←p w do
for 1:←q w do
←ang residual value of
arctan( ( , )) 180
360π
×
×
pw p q
if ( , ) 0<xw p q
if 90<ang
180← +ang ang
if 270>ang
180← −ang ang
,(N,Ang) ←i jhist gradient direction histogram; N is number; {1,2,...9}=Ang
//block width is 2
, , , 1 1, 1, 1{ (N,Ang) , (N,Ang) , (N,Ang) , (N,Ang) }+ + + +←k s i j i j i j i jF hist hist hist hist
This paper hopes to recognize engineering ground object by top and side outline features. The
most outstanding characteristics of these features are the closed geometric figures in a HRRSI.
However, the inside and outside textures of a target can also have closed geometric figures. To
suppress the disturbance brought by this phenomenon, this paper proposes Algorithm 2. Inspired
6. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
6
by literature [27], every pixel is smoothed by its peer group so as to make engineering ground
object more outstanding considering 3 tunnel information.
Algorithm 2: Enhancement of outline feature by peer group filter in 3 tunnels
Input: ( , )I x y
Output: ( , )I x y
←w window width
for 1: /←i m w do
for 1: /←j n w do
( )
, ←r
i jR red tunnel in the window
(g)
, ←i jR green tunnel in the window
( )
, ←b
i jR blue tunnel in the window
(r) (r) 2 ( ) ( ) 2 ( ) ( ) 2
, w/2,w/2 , w/2,w/2 , w/2,w/2( ( () ) )− + − + −← g g b b
i j i j i jR R R R R Rd
←sd sort d by ascending sequence
←k length of sd
for 1:←i k do
0←temp
for 1: 1← −i i do
(j,1) temp← +stemp d
1 /←ia temp i
for 1:←i k do
0←temp
for 1: 1← −i i do
(j,1) temp← +stemp d
2 / (k 1 )← + −ia temp i
for 1:←i k do
0←temp
for 1: 1← −i i do
2
1( (j,1) a ) temp← − +s itemp d
1 ←is temp
for 1:←i k do
0←temp
for 1: 1← −i i do
2
2( (j,1) a ) temp← − +s itemp d
2 ←is temp
2
1 2 1 2( ) / (s )← − +J a a s
mJ ← max of J
( , ) ←PeerGroup x y whichever ( , ) mJ<I x y in the window
( , ) ←I x y ( , )PeerGroup x y convolution with Gaussian filter
3.2 TRAINING OF THE FEATURE-BASED CLASSIFIER
This section intends to adopt HOG feature to recognize monitoring objects.
7. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
7
After determining what features are used to recognize objects, a classifier shall be set up to
identify the target image. The classifier training method used in this paper is based on 25 (PAC),
called boosting. The specific method of setting up such kind of classifier is consistent with [23,
24]. The classifier is a multi-stage one. Each stage can be regarded as a sub-classifier and each
sub-classifier is composed of multiple "weak classifiers" based on trained weights. According to
the PAC theory, a "weak classifier" can be trained into a "strong classifier". The staged classifier
can speed up processing. Firstly, the classifier chooses different window sizes to slide traverse
through the object. Then, the classifier sequentially uses each stage to perform the recognition
judgment on the images within the window. Once this stage has rejected the object with regarding
it as a non-recognition object in the current window, the classifier will move to the next window.
This process is shown in Fig.1.
Figure 1. The structure of the classifier
Each of the classifiers has a certain false positive rate, which should be controlled primarily.
Firstly, the false positive rate of the classifier in different stages is set as [ ]1, 1,2, , 0,1= ∈K if i N f .
Then through N stages, this rate can be controlled at ∏ i
i
f . But it should be noted that when
setting the true positive rate of each sub-classifier in different stages as [ ], 1,2, , 0,1= ∈Ki ik i N k ,
the total true positive rate will be lower than the true positive rate of each sub-classifier.
The training of classifiers requires a large number of samples and there are two types of such
samples. One is the positive sample and the other is the negative sample. The number of samples
is mainly determined by the number of stages of the selected classifier and the total true positive
rate. Both positive samples and negative samples should be collected under a variety of different
conditions. Thousands of positive samples and negative samples are needed for creating a
classifier with a higher true positive rate. It need to be pointed out that this recognition technology
has a basic uniformity requirement of the object's appearance characteristics. For the case where
there are variations in observation angles, the recognition cannot be achieved by a single
classifier. Therefore, it is necessary to establish a logical framework for ground engineering
object features so as to establish a classifier system.
3.3 THE GROUND ENGINEERING OBJECT FEATURE FRAMEWORK
3.3.1 TOP FEATURES
To reduce the priori determination of the concept of ground objects, the feature description
method is adopted in this section. Computer learning of the features of a ground object makes it
8. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
8
possible for the computer to automatically determine the presence of the ground object in a
remote sensing image. This paper employs HOG feature as the description feature.
From the human eye perspective, differences may occur in images collected at different time for
the same ground engineering object because of the following four reasons: ①The ground object
presents different colour features due to different illumination conditions; ② The ground object
presents different shadow features due to different illumination conditions; ③The ground object
with a certain height may have different side features due to different satellite imaging angles; ④
Imaging quality may be different, that is, the images may have different degrees of noise due to
the impact of uncertainty factors.
The differences under the above conditions can be seen from the building image taken at different
times, shown in Fig.2. It can be seen that there are significant differences in building angles of
each image. Fig.2a differs from the other figures without apparent shadows. The building image
in Fig.2e is noticeably darker and there are apparent differences in hue because of the difference
in imaging conditions. A 32 32× size cell HOG feature is calculated for these images. Seen from
the feature map, the HOG feature is not sensitive to illumination conditions and shadow
differences. When the imaging noise is small, the HOG feature has robustness to some extent.
Figure 2. The HOG feature of a building under different conditions of illumination, colour temperature and
noise.
Then, another group of building images are compared as shown in Fig.3. It can be seen that when
the height of the building is high, its side features show a great difference. There are mainly three
cases (excluding buildings with peculiar forms): ①The east, south, west, north side, and the top
of the building are individually obviously imaged; ② Only two sides, that is the southeast,
southwest, northeast, northwest, and the top of the building are obviously imaged; ③ Only the
top of the building is obviously imaged. The HOG features of these images are calculated
separately. Seen from the result in Fig.3, the effects of different imaging angles are relatively
large, mainly in the presence of different side features. Hence, the feature collection should be
carried out from different imaging angles for a certain ground object when a ground engineering
object feature library is established.
9. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
9
For most buildings, the most obvious facet in a satellite image is their top outline. The HOG
feature used in this paper is sensitive to orientation, so the influence of building orientation on
feature recognition under the same external condition is studied firstly.
First of all, the standard observation window used in this paper is north points up and south points
down. This is essentially different from the classic recognition of pedestrians. As pedestrians are
standing on the ground, the up direction of the image is vertically upward and the down direction
of the image is vertically downward in images taken for pedestrians. It is necessary to regulate the
orientation of the observation window because if not, the HOG feature of the same object in the
same image may be different.
Figure 3. The HOG feature of a building under different side features.
At present, most of ground buildings can be divided into two kinds: ① the north-south
orientation; ②the east-west orientation. In feature training, some ground objects can be attributed
to the north-south orientation, and the others can be attributed to the east-west orientation, as
shown in Fig.4. In order to improve the feature recognition rate, it is necessary to set up sample
libraries against the two orientations to carry out feature training. In reality, for example in China,
most ground buildings are in the north-south orientation. Thus, the north-south orientation can be
regarded as the main recognition orientation for the feature library and classifier when the object
recognition is performed.
Figure 4. The orientation feature of buildings.
10. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
10
Secondly, the top outline of building ground objects can be divided into several types with
differences in the HOG feature. The types are: ①the rectangular feature; ②the round feature; ③
the triangle feature; ④ the combination feature; ⑤ the special type. The first three are basic
shapes.
(1) The rectangular feature On the basis of the intuitive appearance, the buildings with the
rectangular feature can be divided into three types. ①Standard rectangle. The top outline of this
type of building attribute to the standard rectangle according to human eye recognition. ②Micro-
deformation rectangular. The top outline of some buildings changes slightly on the basis of the
rectangle. ③Square. The top outline of many towers is intuitively similar to the square. Since the
HOG feature is concerned with the direction distribution of the gradient in the image, the building
attributed to the micro-deformation rectangular type and the square type can still be trained in
accordance with the HOG feature library of the standard rectangle type, and identified by the
corresponding classifier. Some images of the buildings with the rectangle outline feature and their
HOG features are shown in Fig.5.
Figure 5. Rectangle outline features and their HOG feature.
(2) The roundness feature The basic intuitive feature of this type of building is that there are no
obvious straight edges, as well as obvious corners. The HOG features of circles and rectangles are
very different, making it difficult to establish a unified classifier. Therefore, this type of building
requires an independent classifier. There is a shape transformation from a standard circle to an
oval, and the main change is the length ratio variation of the long axis and short axis. The top
outlines of such buildings have a quite uniform feature. Some images of the buildings with the
roundness outline feature and their HOG features are shown in Fig.6.
11. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
11
Figure 6. Roundness outline features and their HOG feature.
(3) The triangle feature The basic intuitive feature of this type of building is that there are three
distinct edges that form a triangle. It’s necessary to establish a classifier for this type of building
as the features of basic triangles differ from rectangles and circles. Similarly, the top outlines of
this type of building also contain a series of transformations. For example, some of the three
angles may appear to be round. But in general, the top outlines of such buildings have a quite
uniform feature. Some images of the buildings with the triangle outline feature and their HOG
features are shown in Fig.7.
Figure 7. Triangle outline features and their HOG feature.
(4) The combination feature and peculiar types
12. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
12
Figure 8. Combination outline features and their HOG feature.
Another type of the top outline is the combination of one or multiples of rectangles, rounds, and
triangles, which can be clearly and directly seen, as shown in Fig.8. On the whole, the basic
outline of the combination type building changes a lot because a variety of combinations can be
made according to the design scheme and the land use requirements. From this perspective, it is
difficult to establish a unified comprehensive characterization of combination-type buildings.
However, the top outline of this type consists of some basic outlines and their features are not
affected and remain unchanged. Seen from the HOG feature images in Fig.8, the outlines contain
some different orientation rectangular features. Hence, the building with the combination outline
feature can be identified divisionally by the classifier based on basic top outline types.
Moreover, there is a type of outline shape that can hardly be classified as either basic top outline
types or the combination type, as shown in Fig.9. Thus, this type of building is classified as the
peculiar type in this paper. Such complex buildings usually have a special function like landmark
and their overall construction volume is usually large. However, the objects in the construction
process still have some basic outlines. For example, they always have rectangular foundation pits.
Therefore, the method mentioned in this paper can still be adopted for the recognition of such
objects.
Figure 9. Some peculiar building outlines.
13. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
13
3.2.2 THE SIDE FEATURE
Due to satellite orbits and satellite imaging angles, the sides of a ground object can be noticeable
in the satellite image sometimes. According to this nature, ground building objects can be divided
into two categories: ① the ground building object with obvious sides; ② the ground building
object without an obvious side.
In the process of collecting feature samples through artificial visual observation, ground building
objects with apparent sides shall be distinguished from ground building objects without an
apparent side so as to establish different classifier sample libraries. As the current collection of
samples can only be conducted with artificial visual observation, there is no quantitative standard
for the distinguishing process. Here this paper proposes an empirical standard. When a side area
of a building is larger than 1/5 of its top area, it is considered to have a non-negligible side.
According to empirical observations, in northern China, the buildings with apparent sides in
satellite imaging are usually high-rises. In the Chinese standard, the definition of a "high-rise
building" is: the residential building with a height greater than 27 meters and the non-monolayer
factory building, warehouse and other civil buildings with a height greater than 24 meters. Some
ground building object images with apparent sides and their HOG feature are shown in Fig.10.
Figure 10. Building images with side features and their HOG feature.
Buildings with no apparent side are usually low in height. Many buildings have this feature, such
as low-rise housing areas, Beijing's traditional residential areas (such as Hutong district),
temporary living and working areas in the construction site, large farmer product market trading
areas and so on. Some ground building object images with no apparent side are shown in Fig.11.
It should be noted that when the satellite optical axis is parallel to the vertical direction, high-rise
buildings may not have apparent sides. However, it still has some other features, such as a clear
shadow. In this case, the building may be recognized on the basis of its top outline. Thus, the
classifier should be selected in accordance with reality in a specific object recognition task.
14. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
14
Figure 11. Some samples of ground engineering objects without apparent side surfaces.
3.3 THE FRAMEWORK OF A CLASSIFIER
A ground building object will undergo two kinds of transformations in a construction process.
One kind is top texture transformation, because top textures are related to specific work contents
and construction organizations. Only when a building's main structure is completed and the
exterior decoration is finished will the transformation stop. The other kind is the side
transformation, which cannot be observed if the ground object always has no apparent side from
the start to the completion of the construction. If a ground object is a high-rise building, it will
experience the process from none-side existence to side existence.
In summary, it is difficult to establish a unified feature classifier for ground engineering objects
with a variety of internal and external differences. Because the recognition method based on the
HOG feature can only establish a classifier for objects that have similar features. Therefore, from
the broadest definition, classifiers should be established separately for the various types of
features so as to achieve random remote sensing image recognition. According to this, the total
number of classifiers that should be established is 2 9 3 2-2 3 8-2 3=54× × × × × × , as shown in
Fig.12. Among them, 'the middle-rise or low-rise building with no apparent side' has no side
feature, and the “no-side” feature is not necessary to be established for 'the high-rise building with
obvious sides'.
15. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
15
Figure 12. The structure of the classifier sequence.
If the false positive rate of each classifier is 2% and the number of objects identified by each
classifier is , 1,2, 54,= =∑Ki i
i
n i n M , then the average number of false-positively identified
objects is 2%M . When the number M is large, the number of false-positively identified objects
will be large. If some external information is added in the recognition process, the required
number of classifiers can be greatly reduced, thereby reducing the value of M . For instance, if
the attitude angle of the satellite with respect to the ground during the imaging can be determined
before analysing the remote sensing image, possible side features can be determined. In this case,
only a total of 12 classifiers are needed. If the orientation of the target in the recognition window
can also be determined, only six classifiers are required then. The apparent progress development
will be studied in Section 4. Combined with the apparent progress and some external information,
the accuracy of regulation and recognition will be greatly enhanced. For example, the most
important thing in the apparent progress is to determine the existence of the object, the planned
coordinates and the planned orientation. Hence, the existence of the target object can be easily
determined using the classifiers if the satellite imaging posture and the nature of the object to be
identified is determined before recognition.
The classifier has a true negative rate and for this situation, it is necessary to manually compare
with the apparent progress to find out that whether the recognition is missing or the corresponding
project object is not completed.
4. A FRAMEWORK FOR CONSTRUCTION PROJECT PROGRESS MONITORING
BASED ON THE HIGH RESOLUTION REMOTE SENSING SATELLITE
4.1 THE MAPPING OF THE PROJECT PROGRESS TO THE APPARENT PROGRESS
For the project progress, what needs to be defined firstly is the project period. As this paper
focuses primarily on the apparent nature changes of the construction project, the project progress
is only discussed from the start to the end of the construction. The basic objective of monitoring
is to compare the actual progress of the project with the planned progress. For the purpose of
achieving the automated progress monitoring, the progress of a project needs to be mapped to its
apparent progress. As the monitoring based on the apparent nature requires medium level
resolution information of the project, it mainly focuses on the following nature of the project: ①
Whether a particular object appears in the entire observation scene; ② Whether the location of
the object is consistent with the plan; ③ Whether the time when the object appears and
disappears is consistent with the plan.
16. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
16
In this paper, the remote sensing image with 0.6m resolution is adopted. In order to facilitate the
comparison, an apparent planning system for the project should be established. Firstly, a ground
coordinate system is established. According to the general layout plan of different construction
phases and the resolution of the specific observation window, the apparent coordinate system can
be set up with a point O of the site as the reference point and two fixed directions respectively as
the X-axis and Y-axis, as shown in Fig.13.
Figure 13. The coordinate system of a construction project.
Based on this system, the coordinates of the engineering objects in the observation window can
be obtained. In order to reduce the omission, engineering objects should be identified from the
project work breaking-down system (WBS). According to the project progress plan, the time
point when each engineering object appears, transforms and disappears can be acquired. By
sorting out this information, a project apparent plan can be obtained, as shown in Fig.14. The
apparent plan should indicate that: in what coordinates, what object should appear at what time
point; what object should be present at what time point; what object should disappear at what
time point. In Fig.14, blue windows indicate the positions of the planned objects.
Figure 14. The apparent progress plan of a construction project at a certain time point.
17. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
17
By comparing the actually acquired images that are also identified using the classifier with the
apparent progress, the automated construction project progress comparison can be achieved, as
shown in Fig. 15.
Figure 15. The comparison between apparent progress plan and actual HRRSI.
4.2. OBJECT ALIGNMENT AND OBJECT SEARCH OF THE REMOTE SENSING IMAGE AND
THE APPARENT PLAN
In order to achieve the comparison between the actual apparent progress and the apparent
progress plan of a construction project, it is necessary to carry out the 'object alignment' process
of the collected HRRSIs and the apparent plans.
Figure 16. The basic satellite imaging model of engineering ground project.
Assume that the imaging scenes of the target engineering ground object comply with the
following assumptions: ① The scene of the ground object is relatively flat so that the terrain
factor does not have a big impact on the satellite imaging accuracy; ② The engineering ground
objects are perpendicular to the earth surface. The basic imaging model of engineering ground
objects is shown in Fig.16, where γ is the angle between the optical axis of the satellite and the
18. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
18
vertical direction. Because of this angle, there will exist not only the top image but also the image
of its partial bottom and side in the engineering ground object imaging when the value of γ is
significant, as shown in Fig.16. θ is the azimuth angle of the satellite imaging. It is the clockwise
angle between the north direction and the imaging direction line of the satellite.
The satellite images of the same feature at different time will be slightly different due to the
difference in satellite gestures. Assume that the basic imaging plane remains unchanged and set
the coordinates of a point in the three-dimensional space as ( ), ,x y z when it has a certain height.
When γ of the satellite imaging is significant, the relationship between the imaging coordinates
( )', 'x y and the original coordinates is shown in Fig.17.
Figure 17. Coordinates relation in satellite imaging.
It can be seen from Fig.17 that the relationship between imaging coordinates and original
coordinates is:
' tan sinγ θ= + ⋅ ⋅x x z (1)
y' tan cosγ θ= + ⋅ ⋅y z (2)
Assume that the observed ground is flat. When the satellite is performing remote sensing
imaging, there will be two different imaging points of the same target location. Their imaging
distance will be different supposing that the satellite azimuth keeps unchanged while optical axis
inclinations are different. As shown in Fig.18, the imaging distance of point A and point B on the
ground is AB when the satellite imaging plane is parallel to the ground. The imaging distance of
point A and point B is ' 'A B when the satellite imaging plane is not parallel to the ground. There
is an inclination angleκ . As can be seen from Fig.18, ' ' cosκ=A B AB .
Hence, when there is no significant height of the ground object, its imaging patterns will also
show a certain difference because of the existence of the optical axis inclination. The endpoint B
of line AB , is used as the reference point. Let '=B B . Thus, although line AB is a line on the
ground without a height, its endpoint A cannot completely coincide. If the satellite images twice,
the angles between the imaging plane and the ground plane are respectively κ and 'κ , and the
imaging lines are respectively AB and ' 'A B . If ' ''=B B , the distance difference between 'A and
''A is:
( )' ' '' '' cos cos 'κ κ∆ = − = −d A B A B AB (3)
19. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
19
Figure 18. The influence of illumination inclination on imaging.
In the three-dimensional case, when the intersection of the satellite imaging plane and the ground
plane is known, the projection position of the new point should be found in line with the
intersection’ normal directions in each plane, as shown in Fig.19.
Figure 19. The alignment error direction in 3D condition.
To sum up, in the satellite imaging of the ground three-dimensional object, a difference between
the object’s top imaging coordinates and the object’s bottom imaging coordinates is caused
because of the object’s own height, the optical axis inclination and the azimuth of the satellite.
Besides, a basic stretch or compression along the normal direction with the ratio ( )cos cos 'κ κ−
appears because of the inclination difference of the imaging plane at each imaging time causing.
For instance, assuming the satellite respectively takes image one time on the monitored area in
two monitoring time points and the imaging inclinations are respectively 25o
and 30o
, then the
basic error of the two imaging is ( )cos25 cos30 =0.86%−o o
.
The so-called object alignment is to align the images collected by remote sensing satellites with
the corresponding objects of the project apparent progress. For simplicity, a fixed search radius r
can be determined since the basic error is often relatively small where max( )> ∆r d and ∆d is
the possible error distance under the current satellite parameter and window size condition. For
collected remote sensing images and the apparent progress, the coordinate system needs to be
aligned firstly, and then the searching for the actual apparent object’s existence within the search
radius r should be conducted with a base point A of a certain apparent progress object as the
standard. The point A chosen should be the ground point of the apparent object. Fig.20 shows the
search direction alignment.
20. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
20
Figure 20. The search direction of appearance progress plan and real image alignment.
In summary, the alignment and recognition of a ground object in the apparent progress
coordinates is shown in Fig.21. The yellow window represents the automatic recognition result of
certain ground objects, and the blue window identifies the apparent plan location of the objects at
the corresponding time point.
Figure 21. Monitoring compare in the appearance progress plan of a construction project at a certain time
point.
5. CASE STUDY
Taking a residential construction project in Beijing as an example, this paper established
corresponding classifiers for ground building objects. Now, the side feature is taken as the
example to describe the sensitivity of classifiers to different feature types. For the project
construction site, it can be seen from naked eyes that the buildings in the image are mainly high-
level rectangular outline buildings in the north-south direction with no obvious side. Classifier A
for the objects has feature sequence: 'high-level', 'north-south orientation', 'no obvious side' and
'rectangular outline'. Classifier B has feature sequence: 'high-level', 'north-south orientation', 'a
southwest side' and 'rectangular outline'. The comparative result is shown in Fig.22. As can be
seen, Classifier A correctly identified a large number of ground building objects while the correct
recognition rate of Classifier B was significantly lower. The results are compared as shown in
Table.1.
21. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
21
Figure 22. The recognition result of different classifier.
Table 1. The recognition result of different classifier.
Seen from the comparison results, only when the appropriate classifier is applied can the objects
be identified.
The robustness of this recognition method to the conditions such as the season and illumination is
observed. Firstly, HRRSIs of different seasons for the same construction site are obtained. It can
be seen that the two images are collected in summer and winter respectively. The summer image
is characterized by the presence of large green-vegetation-covered portions while the winter
image is characterized by the presence of large snow-covered portions. Then, the classifier: 'high-
level', 'north-south orientation', 'a northwest side' and 'rectangular outline', and the classifier:
'high-level', 'north-south orientation', 'a southwest side' and 'rectangular outline' are respectively
used to identify these two images. The comparison results are shown in Fig.23 and Table.2.
Figure 23. The influence of seasons on the recognition result
22. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
22
Table 2. The influence of seasons on the recognition result.
It can clearly be seen that the method has strong robustness to seasons. The main reason for the
error is the similarity between the target building and the surrounding ground objects, which can
be called "camouflages".
Next, the robustness of the recognition method to illumination conditions is observed. First of all,
HRRSIs for the same construction site under different illumination conditions are obtained. These
two images have significant shadows and colour differences due to differences in illumination
conditions. Then, the classifier: 'high-level', 'north-south orientation', 'a northwest side' and
'rectangular outline', and the classifier: 'high-level', in 'north-south orientation', with 'a northeast
side' and 'rectangular outline' are applied to identify the images. The comparison results are
shown in Fig. 24 and Table.3.
Figure 24. The influence of illumination on the recognition result.
Table 3. The influence of illumination on the recognition result.
6. CONCLUSION
This paper presents the advantages and disadvantages of computer vision based construction
monitoring by 2D image, video and 3D lase scan. For a monitoring task, traditional methods bear
the problems of real-time requirements and date objectivity requirements. This paper points out
the applicability of HRRSI in this task. HOG feature is applied as the key feature for machine
learning, which is carried out by the boosting method. To overcome the complexity of HOGs in
HRRSIs, this paper builds a feature framework according to the top, side and height
characteristics of the engineering ground object. To realize progress monitoring, this paper firstly
23. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
23
defines the basic tasks of automated progress monitoring: ①Whether a particular object appears
in the entire observation scene; ② Whether the location of the object is consistent with the plan;
③ Whether the time point when the object appears and disappears is consistent with the plan.
Secondly, this paper gives a framework for real-time HRRSI and apparent progress plan
comparison to realize monitoring. Finally, an object alignment and search method are derived.
Through an application case, the performance of this framework is presented. Judging from the
result, we can see that this method is robust under different seasonal and illuminative conditions.
Recently, deep learning method gives good performance in object recognition. For further studies,
the combination of the framework given in this work and deep learning is encouraged.
ACKNOWLEDGEMENTS
This paper is supported by the Fundamental Research Funds for the Central Universities
(2016MS70) and the Research Funds of Beijing Social Science (16GLC069).
REFERENCE
[1] Alsemmeari R A, Bakhsh S T, Alsemmeari H, (2016). Free Space Optics Vs Radio Frequency
Wireless Communication, International Journal of Information Technology and Computer Science
8(9), pp:1-8.
[2] Faridaddin Vahdatikhakia, Amin Hammadb, Hassaan Siddiquic, (2015). Optimization-based
excavator pose estimation using real-timelocation systems. Automation in Construction 56, pp: 76–
92.
[3] YManoj P K, Manoj K S, (2013). GPS Tracking System Coupled with Image Processing in Traffic
Signals to Enhance Life Security. International Journal of Computer Science & Information
Technology 5(4), pp:131-142.
[4] Yuxi Liu, Denghua Zhong, Bo Cui, Guiliang Zhong, Yongxin Wei, (2015). Study on real-time
construction quality monitoring of storehousesurfaces for RCC dams. Automation in Construction 49,
pp: 100–112.
[5] Fei Daia, Man-Woo Parkb, Matthew Sandidge, Ioannis Brilakis, (2015). A vision-based method for
on-road truck height measurement inproactive prevention of collision with overpasses and tunnels.
Automation in Construction 50, pp: 29–39.
[6] Stefania C. Radopouloua, Ioannis Brilakis, (2015). Patch detection for pavement assessment.
Automation in Construction 53, pp: 95–104.
[7] Jane Matthews, Peter E.D. Love, Sam Heinemann, Robert Chandler, Chris Rumsey, Oluwole Ol,
(2015). Real time progress management: Re-engineering processes forcloud-based BIM in
construction. Automation in Construction 58, pp: 38–47.
[8] Xiaonan Zhanga, Nick Bakisa, Timothy C. Lukins, Yahaya M. Ibrahim, Song Wu, Mike Kagioglou,
Ghassan Aouad, Ammar P. Kaka, Emanuele Trucco, (2009). Automating progress measurement of
construction projects. Automation in Construction 18 pp: 294–301.
[9] Seungjun Roha, Zeeshan Aziza, Feniosky Pena-Mora, (2011). An object-based 3D walk-through
model for interior construction progress monitoring. Automation in Construction 20 pp: 66–75.
[10] Kuo-Liang Lina, Jhih-Long Fang, (2013). Applications of computer vision on tile alignment
inspection. Automation in Construction 35 pp: 562–567.
[11] Kevin K. Hana, Mani Golparvar-Fardb, (2015). Appearance-based material classification for
monitoring of operation-level construction progress using 4D BIM and site photologs. Automation in
Construction, 53 pp: 44-57.
[12] Kuo-Liang Lina, Jhih-Long Fang, (2013). Applications of computer vision on tile alignment
inspection. Automation in Construction 35 pp: 562–567.
[13] Pablo Rodriguez-Gonzalvez Diego Gonzalez-Aguilera, Gemma Lopez-Jimenez, Inmaculada Picon-
Cabrera, (2014). Image-based modeling of built environment from an unmanned aerial system.
Automation in Construction 48 pp: 44–52.
[14] Jie Gong, Carlos H. Caldas, (2011). An object recognition, tracking, and contextual reasoning-based
video interpretationmethod for rapid productivity analysis of construction operations. Automation in
Construction 20 pp:1211–1226.
24. International Journal of Computer Science & Information Technology (IJCSIT) Vol 9, No 6, December 2017
24
[15] Man-Woo Parka, Atefe Makhmalbaf, Ioannis Brilakis, (2011). Comparative study of vision tracking
methods for tracking of constructionsite resources. Automation in Construction 20 pp: 905–915.
[16] Man-Woo Parka, Ioannis Brilakis, (2012). Construction worker detection in video frames for
initializing vision trackers. Automation in Construction 28 pp: 15–25.
[17] Frédéric Bosché, (2010). Automated recognition of 3D CAD model objects in laser scans and
calculationof as-built dimensions for dimensional compliance control in construction. Advanced
Engineering Informatics 24 pp: 107–118.
[18] Chao Wang, Yong K. Cho, Changwan Kim, (2015). Automatic BIM component extraction from point
clouds of existingbuildings for sustainability applications. Automation in Construction 56 1–13.
[19] Mani Golparvar-Farda, Jeffrey Bohnb, Jochen Teizerb, Silvio Savaresec, Feniosky Pena-Mora,
(2011). Evaluation of image-based modeling and laser scanning accuracy for emergingautomated
performance monitoring techniques. Automation in Construction 20 pp: 1143–1155.
[20] Samir El-Omari, Osama Moselhi, (2008). Integrating 3D laser scanning and photogrammetry for
progress measurement ofconstruction work. Automation in Construction 18 pp: 1–9
[21] Ioannis Brilakis, Manolis Lourakis, Rafael Sacks, Silvio Savarese, Symeon Christodoulou, Jochen
Teizer, Atefe Makhmalbaf, (2010). Toward automated generation of parametric BIMs based on
hybrid videoand laser scanning data. Advanced Engineering Informatics 24 pp: 456–465.
[22] Chang Chen-Yu, Chen Shi, (2016). Transitional Public-Private Partnership Model in China:
Contracting with Little Recourse to Contracts. Journal of construction engineering and management
142(10) pp: 1-34.
[23] Sascha E.A. Muenzinga, Bram van Ginnekenb, Max A. Viergevera, Josien P.W. Pluim, (2014).
DIRBoost–An algorithm for boosting deformable image registration: Application to lung CT intra-
subject registration. Medical Image Analysis 18(3) pp: 449-459.
[24] Jingjing Cao, Sam Kwong, Ran Wang, (2012). A noise-detection based AdaBoost algorithm for
mislabeled data. Pattern Recognition 45(12) pp: 4451-4465.
[25] Dalal N, Triggs B, Schmid C (2006). Human Detection Using Oriented Histograms of Flow and
Appearance. European Conference on Computer Vision. Springer-Verlag pp: 428-441.
[26] Prates R F, Cámara-Chávez G, Schwartz W R, et al, (2014). Brazilian License Plate Detection Using
Histogram of Oriented Gradients and Sliding Windows. International Journal of Computer Science &
Information Technology 5(6):39-52.
[27] Smolka B (2011). Adaptive Edge Enhancing Technique of Impulsive Noise Removal in Color Digital
Images. International Conference on Computational Color Imaging. Springe, pp:60-74.
AUTHORS
Xu Ruhang received his PhD in management science in School of Economics and
Management in North China Electric Power University (NCEPU), China 2015. From 2015
until now he is a researcher and lecturer in North China Electric Power University. His
research interest includes computing in construction and image processing.