VEL TECH MULTITECH DR. RANGARAJAN DR.SAKUNTHALA
ENGINEERING COLLEGE
Credibility Based Suspicious Behavior Recognition
Using Artificial Intelligence For Theft Prevention
MR.G.VISHNU VARADHAN RAO, M.E.,
Asst. Prof. /ECE
GUIDED BY
Department of Electronics and Communication Engineering
05/08/2021 BATCH NO:8 1
TEAM MEMBERS
AISHWARYA B 113117106004
DHIVYA S 113117106024
SUBALAKSHMI R 113117106087
BATCH NO : 8
05/08/2021 BATCH NO:8 2
05/08/2021 BATCH NO:8 3
ABSTRACT
 The system deals with the development of an application for automation of video surveillance in
critical areas as atm, airports and bank agencies to ensure more security .
 The traditional approaches based on access to restricted places or suspected actions as theft, scam
and loitering are insufficient to identify the suspect behavior.
 The proposed algorithm benefits from the advantages of each technique, but it is challenged by
the Real-Time exigence.
 The system uses face features to recognize the suspect behavior.
 An advanced Human detection system using CCN technique and Artificial Intelligence would be
utilized which would create phenomenal results in the detection of the activities and their
categorization
05/08/2021 BATCH NO:8 4
OBJECTIVE
 To prevent the theft from happening using the concept credibility the system
will give alert before the theft happens based on the suspicious behavior.
 To design an intelligent algorithm for automatic classification of videos of
single person to multi people and hence detect strange and suspicious activities.
 To extend the approach by developing algorithms for detecting real-time
changes in the scene and to trigger an alarm in real time, if the identified
activity is suspicious.
05/08/2021 BATCH NO:8 5
LITERATURE REVIEW
TITLE: Comparative Evaluation of Classifiers for Abnormal Event Detection in ATMs
AUTHOR: Vivek Ashokan, O.V. Ramana Murthy
YEAR : 2017
ABSTRACT:
The proposed model has a two-stream structure that is composed of the appearance and motion
streams. For each stream, a recurrent variation auto encoder can model the probabilistic distribution
of the normal data in a semi-supervised learning scheme. The appearance and motion features from
the two streams can provide complementary information to describe this probabilistic.
TITLE: A Novel Efficient Method for Abnormal Face Detection in ATM
AUTHOR : Xihao Zhang1, Lin Zhou1, Tao Zhang2, Jie Yang2
YEAR : 2014
ABSTRACT:
A novel and time-saving foreground extraction method is proposed to obtain accurate foreground.
After that, we locate the face with two cascaded steps. An empirical rule-based face localization is
utilized to locate the face roughly, then adaptive ellipse fitting helps accurately locate the face. In
order to detect the occluded faces, we use ADABOOST to combine skin color detection and face
templates matching.
05/08/2021 BATCH NO:8 6
TITLE: Design of an ATM Security through Smart-Vision.
AUTHOR: Bharti Thakur, Pof. Bhupinder Verma
YEAR : 2018
ABSTRACT :
The project presents effort reported in this paper tries to incorporate intelligence in the
cameras installed in bank ATMs, to trigger a preventive action to incase of any abnormal
behaviors of person using it. A novel method using HOG based features extraction and
SVM classifier is reported. On the basis of the feature extraction, the Support Vector
Machine (SVM) classifies the activity of the captured images as normal or abnormal.
TITLE : Combining Commercially Available Active and Passive Sensors In to
a Millimeter-Wave Imager for Concealed Weapon Detection.
AUTHOR : Federico García-Rial, Daniel Montesano, Ignacio Gómez, Carlos
Callejero, Francis Bazus, and Jesús Grajal,
YEAR : 2018
ABSTRACT :
In this project, a system combining active and passive mm-wave sensors along with a
depth-sensing camera is presented. These sensors are combined at a hardware stage (being
integrated into a single imager) as well as at software level (via simultaneous processing
and sensor fusion).
05/08/2021 BATCH NO:8 7
EXISTING SYSTEM
 In the existing system, light weight encryption approach
is used. The video signal is converted into each frame.
 Each frame as a whole is loaded into a cloud .
 Cryptography techniques such as, encryption and
decryption methods are used to encode and decode the
frame. Then layer cellular algorithm is used for analyzing
each frame.
This leads to drastic increase in overall cost of the
system which is directly proportional to scalability of the
system.
05/08/2021 BATCH NO:8 8
DISADVANTAGES OF EXIXTING SYSTEM
HIGH COST
05/08/2021 BATCH NO:8 9
PROPOSED SYSTEM
The proposed system makes efficient utilization of digital image
processing through algorithms such as Canny edge algorithm and Viola-Jones
algorithm.
The above mentioned algorithm can be used to detect and recognize faces,
identify objects, classify human actions in videos, track camera movements,
track moving objects finally ending up with the detection and identification of
the necessary action for the prevention of such type of activities.
The proposed system includes the specialized mechanisms for
Camera
Tampering
Collision of
Humans
Long Time
Tracking
05/08/2021 BATCH NO:8 10
BLOCK DIAGRAM
05/08/2021 BATCH NO:8 11
CANNY EDGE DETECTION ALGORITHM
The Canny edge detector is an edge
detection operator that uses a multi-
stage algorithm to detect a wide range
of edges in images.
Canny edge detection is a image processing
method used to detect edges in an image while
suppressing noise.
Advantages -
05/08/2021 BATCH NO:8 12
STEPS INVOLVED
 Apply Gaussian filter to smooth the image in
order to remove the noise. Find the intensity
gradients of the image.
 Apply non-maximum suppression to get rid of
spurious response to edge detection.
 Apply double threshold to determine potential
Edges.
 Finalize the detection of edges by suppressing
all the other edges that are weak and not
connected to strong edges.
05/08/2021 BATCH NO:8 13
 Presence of Gaussian filter allows
removing of any noise in an image.
 The signal can be enhanced with respect to
the noise ratio by non-maxima suppression
method which results in one pixel wide
ranges as the output.
 Detects the edges in a noisy state by
applying the threshold method.
 This algorithm gives a good localization,
response and is immune to a noisy
environment.
Original image
Image reduced to
grayscale Gaussian
filter with σ=1.4
The intensity
gradient of the
previous image.
Non-maximum
suppression
applied
05/08/2021 BATCH NO:8 14
VIOLA-JONES ALGORITHM
 The Viola-Jones algorithm is widely used mechanism
for object detection.
 This algorithm uses Haar basis feature filters, so it
does not use multiplications.
 The efficiency of the Viola-Jones algorithm can be
significantly increased by first generating the integral
image.
 Viola Jones algorithm is used to detect many faces
and area of interested features like eyes, nose, mouth etc.
Accurate detection of features increases.
BATCH NO:8 15
The characteristics of Viola–Jones algorithm which make it a good
detection algorithm are:
ROBUST
REAL TIME
PROCESSED
FACE
DETECTION
05/08/2021
05/08/2021 BATCH NO:8 16
HAAR FEATURES
All human faces share some similar
properties.
The following regularities may be matched
using Haar Features.
The eye region is darker than the upper-
cheeks.
The nose bridge region is brighter than the
eyes
Location and size: eyes, mouth, bridge of nose
05/08/2021 BATCH NO:8 17
FACE DETECTION
Viola-Jones algorithm is used for face detection the image is
converted into grayscale, since it is easier to work with and acquires
lesser data to process.
The Viola-Jones algorithm first detects the face on the grayscale
image and then finds the location on the colored image.
Viola-Jones outlines a box and searches for a face within the box. It
is essentially searching for the haar-like features.
 The box moves a step to the right after going through every tile in
the picture and the number of boxes used to detect haar features and
the data of all of those boxes put together, helps the algorithm
determine where the face is.
05/08/2021 BATCH NO:8 18
CAMERA TAMPERING
Camera tampering is defined as any sustained
event which dramatically alters the image seen
by the camera
If camera lens is sprayed by glue and the
image is darkened.
The proposed system will generate an alert
message.
 Alert message will be generated for all cases
where the image turns dark or the lens is
sprayed.
05/08/2021 BATCH NO:8 19
HUMAN COLLISION
In the Human detection module every human
part is bounded by the respective bounding
boxes and tracking is continued.
 If more persons are in the same environment,
the multiple detection and tracking is also
possible. So from that the system has multiple
boxes in the environment.
If the collision is occurs between the
boundaries then the violations may happened,
so the system sends the alert message to the
admin.
05/08/2021 BATCH NO:8 20
LONG TIME TRACKING
Haar cascade is the library which works with
image processing to detect the human.
Long time tracking is that, when the human
enters into the ATM center /Bank the
Surveillance camera start tracking with the
bounding box by the identification of human
parts.
The proposed system calls the timer to
measure the time.
05/08/2021 BATCH NO:8 21
SOFTWARE DESCRIPTION
MATLAB is a high-performance language for technical computing. It integrates computation,
visualization, and programming in an easy-to-use environment where problems and solutions are
expressed in familiar mathematical notation.
Typical uses include
 Math and computation
 Algorithm development
 Modeling, simulation, and prototyping
 Data analysis, exploration, and visualization
 Scientific and engineering graphics
 Application development
05/08/2021 BATCH NO:8 22
 M-file – An M-file is a MATLAB document
the user creates to store the code they write for
their specific application.
The advantage of using an M-file is that the
user, after modifying their code, must only tell
MATLAB to run the M-file, rather than re-
enter each line of code individually.
 An M-file is useful because it saves the code
the user has written for their application.
 It can be manipulated and tested until it
meets the user’s specifications.
M-FILE FOR LOADING
IMAGE
M-FILE FOR SAVING
IMAGE
05/08/2021 BATCH NO:8 23
RESULT CLIPS
CAMERA TAMPERING
05/08/2021 BATCH NO:8 24
FACE NOT RECOGNISED
05/08/2021 BATCH NO:8 25
HUMAN COLLISION
05/08/2021 BATCH NO:8 26
CONCLUSION
Suspicious activity recognition remains to be an important problem in
computer vision.
Methodologies and technologies have made tremendous development in the
past decades and have kept developing up to date. However, challenges still exist
when facing realistic sceneries, in addition to the inherent intraclass variation and
interclass similarity problem.
This proposed system consists of abnormal activities such as( camera
covering, Human collision, long time tracking).
The proposed system alarm automatically, it will greatly improve financial
security and facilitate the detective work of the police. Therefore, the different
suspicious activity is recognized before the theft happens.
05/08/2021 BATCH NO:8 27
REFERENCES
[1] H. He and E. A. Garcia, “Learning from imbalanced data,” IEEE Transactions on Knowledge
and Data Engineering, vol. 21, no. 9, pp. 1263–1284, Sept 2009.
[2] S. Zhou, W. Shen, D. Zeng, M. Fang, Y. Wei, and Z. Zhang, “Spatial–temporal convolutional
neural networks for anomaly detection and localization in crowded scenes,” Signal Processing:
Image Communication, vol. 47, pp. 358–368, 2016.
[3] C. Lu, J. Shi, and J. Jia, “Abnormal event detection at 150 fps in matlab,” in Proceedings of the
IEEE International Conference on Computer Vision, 2013, pp. 2720–2727.
[4] S.-H. Yen and C.-H. Wang, “Abnormal event detection using hosf,” in IT Convergence and
Security (ICITCS), 2013 International Conference on. IEEE, 2013, pp. 1–4.
[5] X. Zhu, J. Liu, J. Wang, C. Li, and H. Lu, “Sparse representation for robust abnormality
detection in crowded scenes,” Pattern Recognition, vol. 47, no. 5, pp. 1791–1799, 2014.
05/08/2021 BATCH NO:8 28

new ppt.pptx

  • 1.
    VEL TECH MULTITECHDR. RANGARAJAN DR.SAKUNTHALA ENGINEERING COLLEGE Credibility Based Suspicious Behavior Recognition Using Artificial Intelligence For Theft Prevention MR.G.VISHNU VARADHAN RAO, M.E., Asst. Prof. /ECE GUIDED BY Department of Electronics and Communication Engineering 05/08/2021 BATCH NO:8 1 TEAM MEMBERS AISHWARYA B 113117106004 DHIVYA S 113117106024 SUBALAKSHMI R 113117106087 BATCH NO : 8
  • 2.
  • 3.
    05/08/2021 BATCH NO:83 ABSTRACT  The system deals with the development of an application for automation of video surveillance in critical areas as atm, airports and bank agencies to ensure more security .  The traditional approaches based on access to restricted places or suspected actions as theft, scam and loitering are insufficient to identify the suspect behavior.  The proposed algorithm benefits from the advantages of each technique, but it is challenged by the Real-Time exigence.  The system uses face features to recognize the suspect behavior.  An advanced Human detection system using CCN technique and Artificial Intelligence would be utilized which would create phenomenal results in the detection of the activities and their categorization
  • 4.
    05/08/2021 BATCH NO:84 OBJECTIVE  To prevent the theft from happening using the concept credibility the system will give alert before the theft happens based on the suspicious behavior.  To design an intelligent algorithm for automatic classification of videos of single person to multi people and hence detect strange and suspicious activities.  To extend the approach by developing algorithms for detecting real-time changes in the scene and to trigger an alarm in real time, if the identified activity is suspicious.
  • 5.
    05/08/2021 BATCH NO:85 LITERATURE REVIEW TITLE: Comparative Evaluation of Classifiers for Abnormal Event Detection in ATMs AUTHOR: Vivek Ashokan, O.V. Ramana Murthy YEAR : 2017 ABSTRACT: The proposed model has a two-stream structure that is composed of the appearance and motion streams. For each stream, a recurrent variation auto encoder can model the probabilistic distribution of the normal data in a semi-supervised learning scheme. The appearance and motion features from the two streams can provide complementary information to describe this probabilistic. TITLE: A Novel Efficient Method for Abnormal Face Detection in ATM AUTHOR : Xihao Zhang1, Lin Zhou1, Tao Zhang2, Jie Yang2 YEAR : 2014 ABSTRACT: A novel and time-saving foreground extraction method is proposed to obtain accurate foreground. After that, we locate the face with two cascaded steps. An empirical rule-based face localization is utilized to locate the face roughly, then adaptive ellipse fitting helps accurately locate the face. In order to detect the occluded faces, we use ADABOOST to combine skin color detection and face templates matching.
  • 6.
    05/08/2021 BATCH NO:86 TITLE: Design of an ATM Security through Smart-Vision. AUTHOR: Bharti Thakur, Pof. Bhupinder Verma YEAR : 2018 ABSTRACT : The project presents effort reported in this paper tries to incorporate intelligence in the cameras installed in bank ATMs, to trigger a preventive action to incase of any abnormal behaviors of person using it. A novel method using HOG based features extraction and SVM classifier is reported. On the basis of the feature extraction, the Support Vector Machine (SVM) classifies the activity of the captured images as normal or abnormal. TITLE : Combining Commercially Available Active and Passive Sensors In to a Millimeter-Wave Imager for Concealed Weapon Detection. AUTHOR : Federico García-Rial, Daniel Montesano, Ignacio Gómez, Carlos Callejero, Francis Bazus, and Jesús Grajal, YEAR : 2018 ABSTRACT : In this project, a system combining active and passive mm-wave sensors along with a depth-sensing camera is presented. These sensors are combined at a hardware stage (being integrated into a single imager) as well as at software level (via simultaneous processing and sensor fusion).
  • 7.
    05/08/2021 BATCH NO:87 EXISTING SYSTEM  In the existing system, light weight encryption approach is used. The video signal is converted into each frame.  Each frame as a whole is loaded into a cloud .  Cryptography techniques such as, encryption and decryption methods are used to encode and decode the frame. Then layer cellular algorithm is used for analyzing each frame. This leads to drastic increase in overall cost of the system which is directly proportional to scalability of the system.
  • 8.
    05/08/2021 BATCH NO:88 DISADVANTAGES OF EXIXTING SYSTEM HIGH COST
  • 9.
    05/08/2021 BATCH NO:89 PROPOSED SYSTEM The proposed system makes efficient utilization of digital image processing through algorithms such as Canny edge algorithm and Viola-Jones algorithm. The above mentioned algorithm can be used to detect and recognize faces, identify objects, classify human actions in videos, track camera movements, track moving objects finally ending up with the detection and identification of the necessary action for the prevention of such type of activities. The proposed system includes the specialized mechanisms for Camera Tampering Collision of Humans Long Time Tracking
  • 10.
    05/08/2021 BATCH NO:810 BLOCK DIAGRAM
  • 11.
    05/08/2021 BATCH NO:811 CANNY EDGE DETECTION ALGORITHM The Canny edge detector is an edge detection operator that uses a multi- stage algorithm to detect a wide range of edges in images. Canny edge detection is a image processing method used to detect edges in an image while suppressing noise. Advantages -
  • 12.
    05/08/2021 BATCH NO:812 STEPS INVOLVED  Apply Gaussian filter to smooth the image in order to remove the noise. Find the intensity gradients of the image.  Apply non-maximum suppression to get rid of spurious response to edge detection.  Apply double threshold to determine potential Edges.  Finalize the detection of edges by suppressing all the other edges that are weak and not connected to strong edges.
  • 13.
    05/08/2021 BATCH NO:813  Presence of Gaussian filter allows removing of any noise in an image.  The signal can be enhanced with respect to the noise ratio by non-maxima suppression method which results in one pixel wide ranges as the output.  Detects the edges in a noisy state by applying the threshold method.  This algorithm gives a good localization, response and is immune to a noisy environment. Original image Image reduced to grayscale Gaussian filter with σ=1.4 The intensity gradient of the previous image. Non-maximum suppression applied
  • 14.
    05/08/2021 BATCH NO:814 VIOLA-JONES ALGORITHM  The Viola-Jones algorithm is widely used mechanism for object detection.  This algorithm uses Haar basis feature filters, so it does not use multiplications.  The efficiency of the Viola-Jones algorithm can be significantly increased by first generating the integral image.  Viola Jones algorithm is used to detect many faces and area of interested features like eyes, nose, mouth etc. Accurate detection of features increases.
  • 15.
    BATCH NO:8 15 Thecharacteristics of Viola–Jones algorithm which make it a good detection algorithm are: ROBUST REAL TIME PROCESSED FACE DETECTION 05/08/2021
  • 16.
    05/08/2021 BATCH NO:816 HAAR FEATURES All human faces share some similar properties. The following regularities may be matched using Haar Features. The eye region is darker than the upper- cheeks. The nose bridge region is brighter than the eyes Location and size: eyes, mouth, bridge of nose
  • 17.
    05/08/2021 BATCH NO:817 FACE DETECTION Viola-Jones algorithm is used for face detection the image is converted into grayscale, since it is easier to work with and acquires lesser data to process. The Viola-Jones algorithm first detects the face on the grayscale image and then finds the location on the colored image. Viola-Jones outlines a box and searches for a face within the box. It is essentially searching for the haar-like features.  The box moves a step to the right after going through every tile in the picture and the number of boxes used to detect haar features and the data of all of those boxes put together, helps the algorithm determine where the face is.
  • 18.
    05/08/2021 BATCH NO:818 CAMERA TAMPERING Camera tampering is defined as any sustained event which dramatically alters the image seen by the camera If camera lens is sprayed by glue and the image is darkened. The proposed system will generate an alert message.  Alert message will be generated for all cases where the image turns dark or the lens is sprayed.
  • 19.
    05/08/2021 BATCH NO:819 HUMAN COLLISION In the Human detection module every human part is bounded by the respective bounding boxes and tracking is continued.  If more persons are in the same environment, the multiple detection and tracking is also possible. So from that the system has multiple boxes in the environment. If the collision is occurs between the boundaries then the violations may happened, so the system sends the alert message to the admin.
  • 20.
    05/08/2021 BATCH NO:820 LONG TIME TRACKING Haar cascade is the library which works with image processing to detect the human. Long time tracking is that, when the human enters into the ATM center /Bank the Surveillance camera start tracking with the bounding box by the identification of human parts. The proposed system calls the timer to measure the time.
  • 21.
    05/08/2021 BATCH NO:821 SOFTWARE DESCRIPTION MATLAB is a high-performance language for technical computing. It integrates computation, visualization, and programming in an easy-to-use environment where problems and solutions are expressed in familiar mathematical notation. Typical uses include  Math and computation  Algorithm development  Modeling, simulation, and prototyping  Data analysis, exploration, and visualization  Scientific and engineering graphics  Application development
  • 22.
    05/08/2021 BATCH NO:822  M-file – An M-file is a MATLAB document the user creates to store the code they write for their specific application. The advantage of using an M-file is that the user, after modifying their code, must only tell MATLAB to run the M-file, rather than re- enter each line of code individually.  An M-file is useful because it saves the code the user has written for their application.  It can be manipulated and tested until it meets the user’s specifications. M-FILE FOR LOADING IMAGE M-FILE FOR SAVING IMAGE
  • 23.
    05/08/2021 BATCH NO:823 RESULT CLIPS CAMERA TAMPERING
  • 24.
    05/08/2021 BATCH NO:824 FACE NOT RECOGNISED
  • 25.
    05/08/2021 BATCH NO:825 HUMAN COLLISION
  • 26.
    05/08/2021 BATCH NO:826 CONCLUSION Suspicious activity recognition remains to be an important problem in computer vision. Methodologies and technologies have made tremendous development in the past decades and have kept developing up to date. However, challenges still exist when facing realistic sceneries, in addition to the inherent intraclass variation and interclass similarity problem. This proposed system consists of abnormal activities such as( camera covering, Human collision, long time tracking). The proposed system alarm automatically, it will greatly improve financial security and facilitate the detective work of the police. Therefore, the different suspicious activity is recognized before the theft happens.
  • 27.
    05/08/2021 BATCH NO:827 REFERENCES [1] H. He and E. A. Garcia, “Learning from imbalanced data,” IEEE Transactions on Knowledge and Data Engineering, vol. 21, no. 9, pp. 1263–1284, Sept 2009. [2] S. Zhou, W. Shen, D. Zeng, M. Fang, Y. Wei, and Z. Zhang, “Spatial–temporal convolutional neural networks for anomaly detection and localization in crowded scenes,” Signal Processing: Image Communication, vol. 47, pp. 358–368, 2016. [3] C. Lu, J. Shi, and J. Jia, “Abnormal event detection at 150 fps in matlab,” in Proceedings of the IEEE International Conference on Computer Vision, 2013, pp. 2720–2727. [4] S.-H. Yen and C.-H. Wang, “Abnormal event detection using hosf,” in IT Convergence and Security (ICITCS), 2013 International Conference on. IEEE, 2013, pp. 1–4. [5] X. Zhu, J. Liu, J. Wang, C. Li, and H. Lu, “Sparse representation for robust abnormality detection in crowded scenes,” Pattern Recognition, vol. 47, no. 5, pp. 1791–1799, 2014.
  • 28.