SlideShare a Scribd company logo
1 of 9
Download to read offline
Visual Identification of Apple
Diseases using Image
Processing Techniques on
Android Mobile Devices
Abstract
Apple diseases can cost a gardener their season’s harvest. Using technology that
empowers the average grower to self identify apple diseases will save the grower time,
money, and their season’s crop. An algorithm was presented that automatically detected
and identified three apple diseases: apple bitter rot, blister spot, and scab from an Android
mobile device application. Using OpenCV and Java, a computer vision software library
and programming language, I developed a real-time algorithm that identifies the proper
apple disease using the OpenCV template matching function. The algorithm executed a
pixel-by-pixel transformation using a gray scale filter to improve disease matching
between the templates and the image. A systematic verification process was performed
that took the resulting highest pixel match and performed identification against all disease
templates. The Android application that packages this functionality allowed the user to
take a picture of the diseased area, executed the algorithm, and then displayed the best
match identified. Results suggested that some diseases were identified on either the
wrong disease image or a non-diseased fruit image, while others were accurately
identified. The successful identification of a disease was proven to be 71% accurate,
while the others had a success rate as low as 33%. Therefore alternative or supplemental
methods will need to be explored to improve accuracy across various disease matches.
Anita Garcia
August, 2015
Table of Contents
I. Introduction
II. Methodology
III. Experiments
IV. Experiment Results
V. Apps Created
VI. Discussion
VII. Future Work
References
I. Introduction
Fruit diseases are hazardous in more ways than only their effect on the human
body. The issue of fruit diseases stretches into the grower’s economical well
being and production. According to plant pathologist, Alfredo Martinez, apple
diseases cause a production loss of 0.1 - 5.0 percent and .5 - 153.6 thousands of
dollars are being spent on damage and control [1]. Existing solutions that are
being performed by trained individuals include monitoring and various
management procedures. During the monitoring process, several fruit factors are
observed, like leaf wetness and temperature. These observations are made on
each apple tree and then recorded during the pre bloom, pre harvest and harvest
times. Some management procedures include resistant cultivars, sanitation, and
chemical control. All of this man power and additional disease resistant resources
are costly. In 2008, apple disease cost of control was 267.5 thousands of
dollars[2]. What if the grower could bypass the experts, timely procedures, and
money due to lost of production and cost of control? Automating these
procedures while using template matching, an image processing technique, on
an android phone will eliminate loss in production and economical devastation.
II. Methodology
The primary method used in addressing this agricultural problem is OpenCV’s image
processing method, template matching. In this approach, two images are matched up
against one an other to best identify the most accurate match. A generic, pre-selected
disease template is cross examined against a query image in a pixel by pixel fashion and
results in a float value from -1 to 1, indicating the accuracy of the best match found.
.9993243217468262
[TEMPLATE IMAGE] -> [QUERY IMAGE] -> [DISEASE FOUND]
*In the figure above, the resulting matches found indicate the level of accuracy of the template
being matched against the query image.
Much image variability may be present between the two images being cross examined,
such as lighting, color, and positioning. To best execute template matching, additional
image processing steps are needed to neutralize both images. To best neutralize pixel
variability, I applied an image processing filter that converts both the template and query
image into grayscale.
The second method that I explored was Cascade Classification.There are two main stages
in this classification, one is training and the other is detection[3]. Through an automated
training process, a set of resulting images are generated to be identified with any given
disease. This training process involved collecting positive and negative images in order to
produce a resulting template. A terminal command can be used to begin the training via
openCV’s opencv_traincascade. This work will then be saved in an .xml file,
which is ready to be tested.
III. Experiments
Binary
Much of the image variability questions were taken into account by various experiments.
Any image may give too much pixel information, whether it be color, shine, brightness,
etc. A binary image conversion and comparison was explored, which allowed for each
pixel in the image to have only two possible values, thus eliminating pixel variability. In
my case, I used the black and white colors with the use of a threshold. Both the template
and query image were taken and converted into Binary. These comparisons were done
using template matching and the resulting values closely analyzed. I found that when
using binary images, the resulting accuracy was less than when using the same two
images (template and query) in color. The normalization attempt, in theory, would have
given more accurate results because of the reduction of pixel variety. The opposite was
resulted to be true.
.8070833683013916
[TEMPLATE IMAGE] -> [QUERY IMAGE] -> [DISEASE FOUND]
*In the figure above, the resulting matches found indicate the level of accuracy of the template
being matched against the query image.
Gray Scale
For the same reasons of reducing image variability, the experimentation of the grey scale
filter was used. In this experiment, both the query image and the template were
converted into a grey scaled image with a built in OpenCV method before the OpenCV
template matching function was performed. The normalization that grey scale conversion
brought to both, the template and query image, was beneficial to all matching
experiments. Matching results from the template using this filter yielded a higher
matching accuracy rate. With a higher accuracy rate, this experiment was selected to be
used in the final algorithm as a pre-processing step before template matching was to be
executed.
IV. Experiment Results
The template matching results vary on the amount of templates examined against each
image subject; whether each image subject is tested against a single disease template, or
all eight templates. Single template matches: 33 - 100 % Matching template matches: 5 -
71 % Each of the three has two corresponding templates to improve the accuracy rate.
The smaller the template feature is, the more easily identifiable the disease is across
multiple image subjects, making the accuracy rate drop.
V.Apps Created
Several Applications were developed prior to the completion of the final application.
Four applications were created in order to test key features, image pre-processing
experiments and OpenCV function. The applications created included: Opening Camera,
Image Pre-processing, Template Matching, Test Algorithm, Final Application: Phyto
Sense It.
Camera Access
The purpose of this application was solely to gain Android camera access. The specificity
of this application allows for a camera view to open up.
Image Pre-processing
The purpose of the application is to convert an image to binary, and gray scale. This
application was capable of using the a Camera View instead of the native Android
Camera. This app allowed for a photo to be taken, image filtration to be applied, and the
converted image was displayed.
Template Matching
The purpose of this application is to test the OpenCv template matching function. This
application had multiple stages: At it’s beginning stage, it allowed the user to press a
button that would match the template to an query image taken from a drawable. Later,
the application progressed to match the template against a query image that was captured
from the camera. The resulting template match shows a rectangle around the best
detected match. The last purpose of this application was to test which pre-processing
filter had the best accuracy rates.
Test Algorithm
The purpose of this application is collect the name of the disease, state of fruit, disease
detected, and maximum accuracy rate of each sample photo. In order to attain this
comprehensive data set, each sample picture is matched against each template.
1) Store all images in drawables as diseasename_#
2) Run all templates against each photo individually, by each batch
3) Store maximum values in array
4) After Image batch is iterated through, look for highest max
5) The index of where the maxvalue is found will correspond to the disease
* 0-1: healthy, 2-3: bitterrot, 3-4: blisterspot, 5-6: scab
6) If index of max value is 0 or 1, then fruit is healthy, otherwise fruit is diseased
Name State Disease Detected Max. Accuracy Rate
Final Application
This is the final product, which is a combination of all previous applications put together.
This serves the purpose of disease inspection using the camera and allowing user to snap
a picture after a home screen of photography tips for best chances of disease detection,
run inspection algorithm on the image taken, then display resulting disease detected, if
any, along with a short description of disease, picture of disease, link for more
information and option to try again.
VI. Discussion
Using template matching as my primary method for disease inspection proved to be less
accurate than hypothesized. The similarity of each disease template used made the results
appear on false image subjects. For example, apple blister spot was often detected in
apple bitter rot and apple scab because of it’s subset similarity.
VII. Future Work
Future work will include further exploring cascade classifier or any other methods to
improve accuracy. The initial verification of whether the object being captured is an apple
or not before proceeding into image detection. Conducting a thorough accuracy
experimentation and analysis to pin point the best method compatible with android
platforms. Once this is attained, a larger disease database will need to be implemented.
Any future implications of a functioning, accurate and robust algorithm that detects and
identifies diseases can be utilized as a cost effective tool for large crop growers.
References
[1] Jayamala K. Patil, Raj Kumar, “Advances in Image Processing For Detection of Plant
Diseases,” Journal of Advanced Bioinformatics Applications and Research, vol. 2, Issue,
2, pp. 135-141, June, 2011.
[2] David B. Langston, Jr. (2008). 2008 Georgia Plant Disease Loss Estimates (AP
102-1) [online]. Available: http://extension.uga.edu/publications/detail.cfm?
number=AP102-1#Apple
[3] OpenCV 2.4.11.0 documentation. Cascade Classifier Training [online]. Available:
http://docs.opencv.org/doc/user_guide/ug_traincascade.html

More Related Content

What's hot

Tomato leaves diseases detection approach based on support vector machines
Tomato leaves diseases detection approach based on support vector machinesTomato leaves diseases detection approach based on support vector machines
Tomato leaves diseases detection approach based on support vector machinesAboul Ella Hassanien
 
Identification of Disease in Leaves using Genetic Algorithm
Identification of Disease in Leaves using Genetic AlgorithmIdentification of Disease in Leaves using Genetic Algorithm
Identification of Disease in Leaves using Genetic Algorithmijtsrd
 
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET Journal
 
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUES
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUESAUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUES
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUESJournal For Research
 
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...IRJET- Result on the Application for Multiple Disease Prediction from Symptom...
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...IRJET Journal
 
Segmentation of unhealthy region of plant leaf using image processing techniques
Segmentation of unhealthy region of plant leaf using image processing techniquesSegmentation of unhealthy region of plant leaf using image processing techniques
Segmentation of unhealthy region of plant leaf using image processing techniqueseSAT Journals
 
GreenR: Automatic Plant Disease Diagnosis
GreenR: Automatic Plant Disease DiagnosisGreenR: Automatic Plant Disease Diagnosis
GreenR: Automatic Plant Disease DiagnosisAshish Gupta
 
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...IRJET Journal
 
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...Chakkrit (Kla) Tantithamthavorn
 
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...International Center for Biometric Research
 
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...The Impact of Class Rebalancing Techniques on the Performance and Interpretat...
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...Chakkrit (Kla) Tantithamthavorn
 
Automated parameter optimization should be included in future 
defect predict...
Automated parameter optimization should be included in future 
defect predict...Automated parameter optimization should be included in future 
defect predict...
Automated parameter optimization should be included in future 
defect predict...Chakkrit (Kla) Tantithamthavorn
 
IRJET- An Android based Image Processing Application to Detect Plant Disease
IRJET-  	  An Android based Image Processing Application to Detect Plant DiseaseIRJET-  	  An Android based Image Processing Application to Detect Plant Disease
IRJET- An Android based Image Processing Application to Detect Plant DiseaseIRJET Journal
 
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...Fault Detection in Mobile Communication Networks Using Data Mining Techniques...
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...ijcisjournal
 
Classification of Apple diseases through machine learning
Classification of Apple diseases through machine learningClassification of Apple diseases through machine learning
Classification of Apple diseases through machine learningMuqaddas Bin Tahir
 
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...Chakkrit (Kla) Tantithamthavorn
 
Improving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approachImproving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approachIJERA Editor
 
Genetic algorithm based approach for
Genetic algorithm based approach forGenetic algorithm based approach for
Genetic algorithm based approach forIJCSES Journal
 

What's hot (20)

Tomato leaves diseases detection approach based on support vector machines
Tomato leaves diseases detection approach based on support vector machinesTomato leaves diseases detection approach based on support vector machines
Tomato leaves diseases detection approach based on support vector machines
 
Identification of Disease in Leaves using Genetic Algorithm
Identification of Disease in Leaves using Genetic AlgorithmIdentification of Disease in Leaves using Genetic Algorithm
Identification of Disease in Leaves using Genetic Algorithm
 
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
 
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUES
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUESAUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUES
AUTOMATED BUG TRIAGE USING ADVANCED DATA REDUCTION TECHNIQUES
 
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...IRJET- Result on the Application for Multiple Disease Prediction from Symptom...
IRJET- Result on the Application for Multiple Disease Prediction from Symptom...
 
Segmentation of unhealthy region of plant leaf using image processing techniques
Segmentation of unhealthy region of plant leaf using image processing techniquesSegmentation of unhealthy region of plant leaf using image processing techniques
Segmentation of unhealthy region of plant leaf using image processing techniques
 
GreenR: Automatic Plant Disease Diagnosis
GreenR: Automatic Plant Disease DiagnosisGreenR: Automatic Plant Disease Diagnosis
GreenR: Automatic Plant Disease Diagnosis
 
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...
IRJET - Breast Cancer Prediction using Supervised Machine Learning Algorithms...
 
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...
Explainable Artificial Intelligence (XAI) 
to Predict and Explain Future Soft...
 
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...
(2011) The Impact of Force on Fingerprint Image Quality, Minutiae Count and P...
 
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...The Impact of Class Rebalancing Techniques on the Performance and Interpretat...
The Impact of Class Rebalancing Techniques on the Performance and Interpretat...
 
Automated parameter optimization should be included in future 
defect predict...
Automated parameter optimization should be included in future 
defect predict...Automated parameter optimization should be included in future 
defect predict...
Automated parameter optimization should be included in future 
defect predict...
 
TBerger_FinalReport
TBerger_FinalReportTBerger_FinalReport
TBerger_FinalReport
 
IRJET- An Android based Image Processing Application to Detect Plant Disease
IRJET-  	  An Android based Image Processing Application to Detect Plant DiseaseIRJET-  	  An Android based Image Processing Application to Detect Plant Disease
IRJET- An Android based Image Processing Application to Detect Plant Disease
 
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...Fault Detection in Mobile Communication Networks Using Data Mining Techniques...
Fault Detection in Mobile Communication Networks Using Data Mining Techniques...
 
Classification of Apple diseases through machine learning
Classification of Apple diseases through machine learningClassification of Apple diseases through machine learning
Classification of Apple diseases through machine learning
 
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...
Software Analytics In Action: A Hands-on Tutorial on Mining, Analyzing, Model...
 
IJET-V2I6P28
IJET-V2I6P28IJET-V2I6P28
IJET-V2I6P28
 
Improving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approachImproving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approach
 
Genetic algorithm based approach for
Genetic algorithm based approach forGenetic algorithm based approach for
Genetic algorithm based approach for
 

Similar to RSI_ReportPDF

Multi-class Bio-images Classification
Multi-class Bio-images ClassificationMulti-class Bio-images Classification
Multi-class Bio-images ClassificationZhuo Li
 
Mini Pro PPT.pptx
Mini Pro PPT.pptxMini Pro PPT.pptx
Mini Pro PPT.pptxToMuCh
 
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET-  	  An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET-  	  An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET Journal
 
Fruit Disease Detection And Fertilizer Recommendation
Fruit Disease Detection And Fertilizer RecommendationFruit Disease Detection And Fertilizer Recommendation
Fruit Disease Detection And Fertilizer RecommendationIRJET Journal
 
LEARNING BASES OF ACTICITY
LEARNING BASES OF ACTICITYLEARNING BASES OF ACTICITY
LEARNING BASES OF ACTICITYPadma Kannan
 
Android application for detection of leaf disease (Using Image processing and...
Android application for detection of leaf disease (Using Image processing and...Android application for detection of leaf disease (Using Image processing and...
Android application for detection of leaf disease (Using Image processing and...IRJET Journal
 
IRJET- Survey on Face Recognition using Biometrics
IRJET-  	  Survey on Face Recognition using BiometricsIRJET-  	  Survey on Face Recognition using Biometrics
IRJET- Survey on Face Recognition using BiometricsIRJET Journal
 
A Review Paper on Automated Plant Leaf Disease Detection Techniques
A Review Paper on Automated Plant Leaf Disease Detection TechniquesA Review Paper on Automated Plant Leaf Disease Detection Techniques
A Review Paper on Automated Plant Leaf Disease Detection TechniquesIRJET Journal
 
Classify Rice Disease Using Self-Optimizing Models and Edge Computing with A...
Classify Rice Disease Using Self-Optimizing Models and  Edge Computing with A...Classify Rice Disease Using Self-Optimizing Models and  Edge Computing with A...
Classify Rice Disease Using Self-Optimizing Models and Edge Computing with A...Damian R. Mingle, MBA
 
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCE
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCEFRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCE
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCEIRJET Journal
 
Plant Disease Detection Using ML.pptx
Plant Disease Detection Using ML.pptxPlant Disease Detection Using ML.pptx
Plant Disease Detection Using ML.pptxjmjiniyamandal
 
How to Implement the Digital Medicine in the Future
How to Implement the Digital Medicine in the FutureHow to Implement the Digital Medicine in the Future
How to Implement the Digital Medicine in the FutureYoon Sup Choi
 
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD Editor
 
IRJET- Greensworth: A Step Towards Smart Cultivation
IRJET- Greensworth: A Step Towards Smart CultivationIRJET- Greensworth: A Step Towards Smart Cultivation
IRJET- Greensworth: A Step Towards Smart CultivationIRJET Journal
 
Automatic detection of rust disease of Lentil by machine learning system usin...
Automatic detection of rust disease of Lentil by machine learning system usin...Automatic detection of rust disease of Lentil by machine learning system usin...
Automatic detection of rust disease of Lentil by machine learning system usin...IJECEIAES
 
Skin Cancer Detection Application
Skin Cancer Detection ApplicationSkin Cancer Detection Application
Skin Cancer Detection ApplicationIRJET Journal
 
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNING
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNINGA SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNING
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNINGIRJET Journal
 
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNN
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNNLEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNN
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNNIRJET Journal
 

Similar to RSI_ReportPDF (20)

COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
 
Multi-class Bio-images Classification
Multi-class Bio-images ClassificationMulti-class Bio-images Classification
Multi-class Bio-images Classification
 
Mini Pro PPT.pptx
Mini Pro PPT.pptxMini Pro PPT.pptx
Mini Pro PPT.pptx
 
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET-  	  An Expert System for Plant Disease Diagnosis by using Neural NetworkIRJET-  	  An Expert System for Plant Disease Diagnosis by using Neural Network
IRJET- An Expert System for Plant Disease Diagnosis by using Neural Network
 
Fruit Disease Detection And Fertilizer Recommendation
Fruit Disease Detection And Fertilizer RecommendationFruit Disease Detection And Fertilizer Recommendation
Fruit Disease Detection And Fertilizer Recommendation
 
MAINS PPT1.pptx
MAINS PPT1.pptxMAINS PPT1.pptx
MAINS PPT1.pptx
 
LEARNING BASES OF ACTICITY
LEARNING BASES OF ACTICITYLEARNING BASES OF ACTICITY
LEARNING BASES OF ACTICITY
 
Android application for detection of leaf disease (Using Image processing and...
Android application for detection of leaf disease (Using Image processing and...Android application for detection of leaf disease (Using Image processing and...
Android application for detection of leaf disease (Using Image processing and...
 
IRJET- Survey on Face Recognition using Biometrics
IRJET-  	  Survey on Face Recognition using BiometricsIRJET-  	  Survey on Face Recognition using Biometrics
IRJET- Survey on Face Recognition using Biometrics
 
A Review Paper on Automated Plant Leaf Disease Detection Techniques
A Review Paper on Automated Plant Leaf Disease Detection TechniquesA Review Paper on Automated Plant Leaf Disease Detection Techniques
A Review Paper on Automated Plant Leaf Disease Detection Techniques
 
Classify Rice Disease Using Self-Optimizing Models and Edge Computing with A...
Classify Rice Disease Using Self-Optimizing Models and  Edge Computing with A...Classify Rice Disease Using Self-Optimizing Models and  Edge Computing with A...
Classify Rice Disease Using Self-Optimizing Models and Edge Computing with A...
 
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCE
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCEFRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCE
FRUIT DISEASE DETECTION AND CLASSIFICATION USING ARTIFICIAL INTELLIGENCE
 
Plant Disease Detection Using ML.pptx
Plant Disease Detection Using ML.pptxPlant Disease Detection Using ML.pptx
Plant Disease Detection Using ML.pptx
 
How to Implement the Digital Medicine in the Future
How to Implement the Digital Medicine in the FutureHow to Implement the Digital Medicine in the Future
How to Implement the Digital Medicine in the Future
 
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
 
IRJET- Greensworth: A Step Towards Smart Cultivation
IRJET- Greensworth: A Step Towards Smart CultivationIRJET- Greensworth: A Step Towards Smart Cultivation
IRJET- Greensworth: A Step Towards Smart Cultivation
 
Automatic detection of rust disease of Lentil by machine learning system usin...
Automatic detection of rust disease of Lentil by machine learning system usin...Automatic detection of rust disease of Lentil by machine learning system usin...
Automatic detection of rust disease of Lentil by machine learning system usin...
 
Skin Cancer Detection Application
Skin Cancer Detection ApplicationSkin Cancer Detection Application
Skin Cancer Detection Application
 
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNING
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNINGA SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNING
A SURVEY ON BLOOD DISEASE DETECTION USING MACHINE LEARNING
 
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNN
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNNLEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNN
LEAF DISEASE IDENTIFICATION AND REMEDY RECOMMENDATION SYSTEM USINGCNN
 

RSI_ReportPDF

  • 1. Visual Identification of Apple Diseases using Image Processing Techniques on Android Mobile Devices Abstract Apple diseases can cost a gardener their season’s harvest. Using technology that empowers the average grower to self identify apple diseases will save the grower time, money, and their season’s crop. An algorithm was presented that automatically detected and identified three apple diseases: apple bitter rot, blister spot, and scab from an Android mobile device application. Using OpenCV and Java, a computer vision software library and programming language, I developed a real-time algorithm that identifies the proper apple disease using the OpenCV template matching function. The algorithm executed a pixel-by-pixel transformation using a gray scale filter to improve disease matching between the templates and the image. A systematic verification process was performed that took the resulting highest pixel match and performed identification against all disease templates. The Android application that packages this functionality allowed the user to take a picture of the diseased area, executed the algorithm, and then displayed the best match identified. Results suggested that some diseases were identified on either the wrong disease image or a non-diseased fruit image, while others were accurately identified. The successful identification of a disease was proven to be 71% accurate, while the others had a success rate as low as 33%. Therefore alternative or supplemental methods will need to be explored to improve accuracy across various disease matches. Anita Garcia August, 2015
  • 2. Table of Contents I. Introduction II. Methodology III. Experiments IV. Experiment Results V. Apps Created VI. Discussion VII. Future Work References
  • 3. I. Introduction Fruit diseases are hazardous in more ways than only their effect on the human body. The issue of fruit diseases stretches into the grower’s economical well being and production. According to plant pathologist, Alfredo Martinez, apple diseases cause a production loss of 0.1 - 5.0 percent and .5 - 153.6 thousands of dollars are being spent on damage and control [1]. Existing solutions that are being performed by trained individuals include monitoring and various management procedures. During the monitoring process, several fruit factors are observed, like leaf wetness and temperature. These observations are made on each apple tree and then recorded during the pre bloom, pre harvest and harvest times. Some management procedures include resistant cultivars, sanitation, and chemical control. All of this man power and additional disease resistant resources are costly. In 2008, apple disease cost of control was 267.5 thousands of dollars[2]. What if the grower could bypass the experts, timely procedures, and money due to lost of production and cost of control? Automating these procedures while using template matching, an image processing technique, on an android phone will eliminate loss in production and economical devastation. II. Methodology The primary method used in addressing this agricultural problem is OpenCV’s image processing method, template matching. In this approach, two images are matched up against one an other to best identify the most accurate match. A generic, pre-selected disease template is cross examined against a query image in a pixel by pixel fashion and results in a float value from -1 to 1, indicating the accuracy of the best match found. .9993243217468262 [TEMPLATE IMAGE] -> [QUERY IMAGE] -> [DISEASE FOUND] *In the figure above, the resulting matches found indicate the level of accuracy of the template being matched against the query image. Much image variability may be present between the two images being cross examined, such as lighting, color, and positioning. To best execute template matching, additional
  • 4. image processing steps are needed to neutralize both images. To best neutralize pixel variability, I applied an image processing filter that converts both the template and query image into grayscale. The second method that I explored was Cascade Classification.There are two main stages in this classification, one is training and the other is detection[3]. Through an automated training process, a set of resulting images are generated to be identified with any given disease. This training process involved collecting positive and negative images in order to produce a resulting template. A terminal command can be used to begin the training via openCV’s opencv_traincascade. This work will then be saved in an .xml file, which is ready to be tested. III. Experiments Binary Much of the image variability questions were taken into account by various experiments. Any image may give too much pixel information, whether it be color, shine, brightness, etc. A binary image conversion and comparison was explored, which allowed for each pixel in the image to have only two possible values, thus eliminating pixel variability. In my case, I used the black and white colors with the use of a threshold. Both the template and query image were taken and converted into Binary. These comparisons were done using template matching and the resulting values closely analyzed. I found that when using binary images, the resulting accuracy was less than when using the same two images (template and query) in color. The normalization attempt, in theory, would have given more accurate results because of the reduction of pixel variety. The opposite was resulted to be true. .8070833683013916 [TEMPLATE IMAGE] -> [QUERY IMAGE] -> [DISEASE FOUND] *In the figure above, the resulting matches found indicate the level of accuracy of the template being matched against the query image. Gray Scale For the same reasons of reducing image variability, the experimentation of the grey scale filter was used. In this experiment, both the query image and the template were converted into a grey scaled image with a built in OpenCV method before the OpenCV
  • 5. template matching function was performed. The normalization that grey scale conversion brought to both, the template and query image, was beneficial to all matching experiments. Matching results from the template using this filter yielded a higher matching accuracy rate. With a higher accuracy rate, this experiment was selected to be used in the final algorithm as a pre-processing step before template matching was to be executed. IV. Experiment Results The template matching results vary on the amount of templates examined against each image subject; whether each image subject is tested against a single disease template, or all eight templates. Single template matches: 33 - 100 % Matching template matches: 5 - 71 % Each of the three has two corresponding templates to improve the accuracy rate. The smaller the template feature is, the more easily identifiable the disease is across multiple image subjects, making the accuracy rate drop. V.Apps Created Several Applications were developed prior to the completion of the final application. Four applications were created in order to test key features, image pre-processing experiments and OpenCV function. The applications created included: Opening Camera, Image Pre-processing, Template Matching, Test Algorithm, Final Application: Phyto Sense It. Camera Access The purpose of this application was solely to gain Android camera access. The specificity of this application allows for a camera view to open up.
  • 6. Image Pre-processing The purpose of the application is to convert an image to binary, and gray scale. This application was capable of using the a Camera View instead of the native Android Camera. This app allowed for a photo to be taken, image filtration to be applied, and the converted image was displayed. Template Matching The purpose of this application is to test the OpenCv template matching function. This application had multiple stages: At it’s beginning stage, it allowed the user to press a button that would match the template to an query image taken from a drawable. Later, the application progressed to match the template against a query image that was captured from the camera. The resulting template match shows a rectangle around the best detected match. The last purpose of this application was to test which pre-processing filter had the best accuracy rates.
  • 7. Test Algorithm The purpose of this application is collect the name of the disease, state of fruit, disease detected, and maximum accuracy rate of each sample photo. In order to attain this comprehensive data set, each sample picture is matched against each template. 1) Store all images in drawables as diseasename_# 2) Run all templates against each photo individually, by each batch 3) Store maximum values in array 4) After Image batch is iterated through, look for highest max 5) The index of where the maxvalue is found will correspond to the disease * 0-1: healthy, 2-3: bitterrot, 3-4: blisterspot, 5-6: scab 6) If index of max value is 0 or 1, then fruit is healthy, otherwise fruit is diseased Name State Disease Detected Max. Accuracy Rate
  • 8. Final Application This is the final product, which is a combination of all previous applications put together. This serves the purpose of disease inspection using the camera and allowing user to snap a picture after a home screen of photography tips for best chances of disease detection, run inspection algorithm on the image taken, then display resulting disease detected, if any, along with a short description of disease, picture of disease, link for more information and option to try again. VI. Discussion Using template matching as my primary method for disease inspection proved to be less accurate than hypothesized. The similarity of each disease template used made the results appear on false image subjects. For example, apple blister spot was often detected in apple bitter rot and apple scab because of it’s subset similarity. VII. Future Work Future work will include further exploring cascade classifier or any other methods to improve accuracy. The initial verification of whether the object being captured is an apple or not before proceeding into image detection. Conducting a thorough accuracy experimentation and analysis to pin point the best method compatible with android platforms. Once this is attained, a larger disease database will need to be implemented. Any future implications of a functioning, accurate and robust algorithm that detects and identifies diseases can be utilized as a cost effective tool for large crop growers.
  • 9. References [1] Jayamala K. Patil, Raj Kumar, “Advances in Image Processing For Detection of Plant Diseases,” Journal of Advanced Bioinformatics Applications and Research, vol. 2, Issue, 2, pp. 135-141, June, 2011. [2] David B. Langston, Jr. (2008). 2008 Georgia Plant Disease Loss Estimates (AP 102-1) [online]. Available: http://extension.uga.edu/publications/detail.cfm? number=AP102-1#Apple [3] OpenCV 2.4.11.0 documentation. Cascade Classifier Training [online]. Available: http://docs.opencv.org/doc/user_guide/ug_traincascade.html