Image processing and Machine learning for automated fruit grading system: A Technical Review
-
Upload
cgpitbardoli -
Category
Technology
-
view
1.431 -
download
1
description
Transcript of Image processing and Machine learning for automated fruit grading system: A Technical Review
Image processing and Machine learning for Automated fruit grading system –A Review
Guided by : Prof.Nikunj Gamit Prepared By: Rashmi Pandey
• Image Processing in Agriculture
• Image processing and analysis
• Fruit Grading Process
• Literature review based on color feature extraction techniques
• Workflow
• Future Work
Outline of Presentation
Image Processing in Agriculture[1]
• Detection of disease in leaf, stem fruits and vegetables.
• Determination of shape and color characterization of fruits.
• In remote sensing and irrigation.
• In fact, quantification of the visual properties of
horticultural products and plants can play an important role
to improve and automate agricultural management tasks.
Image Processing and analysis[2]
Fruit Grading Process[3]
Extracted features from processed image
Parameters Processing
Color Color value and degree of color distribution are
Measured based on R, G, and B color
component ratio.
Shape Shape is measured as boundary-based
features, region-based features, mathematical
morphology, and so on.
Size Size is measured from the maximum length or area
on upper image or calculated volume from several
images.
B Back
Paper Title Year Author Published In Method
Automated strawberry grading system based on image processing
2010 Xu Liming, Zhao Yanchao
Computers and Electronics in Agriculture, Elsevier
Dominant Colour Method
Computer vision based date fruit grading system: Design and implementation
2011 Yousef Al Ohali Journal of King Saud University – Computer and Information Sciences, Elsevier
Intensity Distribution Method
Mango Grading By Using Fuzzy Image Analysis
2012 Tajul Rosli B. Razak, Mahmod B. Othman, Mohd Nazari bin Abu Bakar, Khairul Adilah bt Ahmad, Ab Razak Mansor
International Conference on Agricultural, Environment and Biological Sciences
Mean of colour in images
Literature Survey
Paper Title Year Author Published In Method
Application of neural networks to the color grading of apples
1997 Kazuhiro Nalcano Computers and Electronics in Agriculture, Elsevier
Nine Color Characteristic Data
Development of a lemon sorting system based on color and size
2010 M. Khojastehnazhand, M. Omid and A. Tabatabaeefar
African Journal of Plant Science HSI Colour Model Technique
Adaptive texture and color segmentation for tracking moving objects
2002 Ercan Ozyildiz, Nils Krahnst-over, Rajeev Sharma
Pattern Recognition, Elsevier YES Colour Model Technique
Literature Survey
Paper Title Year Author Published In Method
Automated Grading of Palm Oil Fresh Fruit Bunches (FFB) using Neuro-Fuzzy Technique
2009 Nursuriati Jamil, Azlinah Mohamed and Syazwani Abdullah
International Conference of Soft Computing and Pattern Recognition
Simple RGB model
A practical solution for ripe tomato recognition and localization.
2013 Xuming Chen and Simon X. Yang
Journal of Real-Time Image Processing
The Segmentation Method
Rapid Color Grading for Fruit Quality Evaluation Using Direct Color Mapping
2011 Dah-Jye Lee, James K. Archibald and Guangming Xiong
IEEE transactions on automation science and engineering
Direct Color Mapping Technique
Post-harvest profile of mango
2013 A Report
Literature Survey
Input image
Segmentation
Classification (FL, NN or
SVM)
Grade I Grade II Grade III
Implementation Work Flow
Colour Feature Extraction
Disease
Pre-processing
Segmentation Techniques[4] Segmentation Method Description Advantages Disadvantages
Histogram thresholding Histogram is constructed having peaks which correspond to a region.
Low computational complexity. No prior information needed.
Spatial details not considered, cannot guarantee the segmented regions to be contiguous.
Region based approaches Pixels are grouped in the homogeneous regions, and region merging, splitting or their combination is used.
Noise immune in edge detection approach.
High computational complexity. In region splitting segments appear square due to splitting scheme.
Edge detection approaches
Tries to locate the points having changes in gray level.
Works well for high contrast images.
Less immune to noise and doesn’t work well if the image have too many edges.
Segmentation Techniques[4] Segmentation Method Description Advantages Disadvantages
Fuzzy approaches Fuzzy operators, inference rules and properties are applied.
Approximate inference can be performed by fuzzy IF-THEN rules.
Computation can be intensive and determination of membership function is not an easy job.
Neural network approaches
Classification and clustering can be performed.
Less complicated and parallel nature of neural network can be used.
Training time is long and results can be affected by initialization.
Back
Different Colour Feature Extraction Techniques[1] • Dominant Colour Method
• Intensity Distribution
• Mean of Colour
• Nine Colour Characteristic Data
• HSI Colour Model
• Simple RGB
• Segmentation
• Direct Colour Mapping
Dominant Colour Method
• One of the techniques for the colour feature extraction used in the automated strawberry grading is the Dominant Colour Method.
• It represented the image of strawberry in L*a*b* colour model.
• Generally, the human sight is more interested in main colour of the image means that colour which appears frequently in the image.
• So this Dominant Colour Method was used on a* channel to extract the colour feature from the image.
Intensity Distribution Method
• Dates were graded according to their flabbiness.
• The best quality was given to the flabbiest date.
• In order to estimate the flabbiness they used the colour intensity distribution in the image.
• The image is then converted in to gray level and then colour intensity was found from that.
• Flabbiest date is brighter and less flabby date darker.
Mean of colour in images
• In order to determine the colour of the mango the mean of the colour array for red, green and blue was calculated as follows:
• Mean image = (Red value (Find size image) + Green value (Find size image) + Blue value (Find size image))/3
• Simple and easy to implement.
• Doesn't give the accurate colour.
Back
Classification[1]
• Neural Networks • A neural network consists of neurons, arranged in layers, which convert an
input vector into some output. • Each neuron takes an input, applies a function to it and then passes the
output on to the next layer.
• Fuzzy Logic • The simplest fuzzy rule-based classifier . • It is a fuzzy if-then system.
• Support Vector Machine
• SVM are supervised learning models . • Used for classification and regression analysis Back
Grading Quality Standards[6]
Grade designation
Grade Requirements Color Grade tolerances
Grade I Mangoes must be of superior quality They must be free of defects
5% Tolerance
Grade II
Mangoes must be of good quality. Mangoes may have slight defects in shape; slight skin defects due to rubbing or sunburn, and healed bruises not exceeding 2,3,4,5 sq. cm.
10% Tolerance
Grade III Mangoes may have slight defects in shape; slight skin defects due to rubbing or sunburn, and healed bruises not exceeding 4,5,6,7 sq. cm.
10% Tolerance
Back
Future Work
References 1) Image processing and machine learning for automated fruit grading system: A
Technical Review
2) Gonzalez, Rafael C., Richard E. Woods, and S. L. Eddins. "Image segmentation." Digital Image Processing (2002): 577-581
3) Website. [Online] http://webee.technion.ac.il/ido.pdf
4) Cheng, Heng-Da, X. H. Jiang, Ying Sun, and Jingli Wang. "Color image segmentation: advances and prospects." Pattern recognition 34, no. 12 (2001): 2259-2281.
5) A report: Post-harvest profile of mango ,year 2013
Any Q’s???