AU2021101496A4 - A Process for grape leaf disease identification using machine learning techniques - Google Patents
A Process for grape leaf disease identification using machine learning techniques Download PDFInfo
- Publication number
- AU2021101496A4 AU2021101496A4 AU2021101496A AU2021101496A AU2021101496A4 AU 2021101496 A4 AU2021101496 A4 AU 2021101496A4 AU 2021101496 A AU2021101496 A AU 2021101496A AU 2021101496 A AU2021101496 A AU 2021101496A AU 2021101496 A4 AU2021101496 A4 AU 2021101496A4
- Authority
- AU
- Australia
- Prior art keywords
- image
- disease
- leaf
- thres
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- 201000010099 disease Diseases 0.000 title claims abstract description 71
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 71
- 230000008569 process Effects 0.000 title claims abstract description 34
- 235000014787 Vitis vinifera Nutrition 0.000 title claims abstract description 28
- 235000009754 Vitis X bourquina Nutrition 0.000 title claims abstract description 27
- 235000012333 Vitis X labruscana Nutrition 0.000 title claims abstract description 27
- 238000010801 machine learning Methods 0.000 title claims abstract description 22
- 240000006365 Vitis vinifera Species 0.000 title description 2
- 230000011218 segmentation Effects 0.000 claims abstract description 28
- 241000219095 Vitis Species 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000001914 filtration Methods 0.000 claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 125000000205 L-threonino group Chemical group [H]OC(=O)[C@@]([H])(N([H])[*])[C@](C([H])([H])[H])([H])O[H] 0.000 claims description 26
- 238000012706 support-vector machine Methods 0.000 claims description 19
- 238000007637 random forest analysis Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000003902 lesion Effects 0.000 claims description 7
- 239000000203 mixture Substances 0.000 claims description 5
- 230000010339 dilation Effects 0.000 claims description 4
- 230000003628 erosive effect Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 11
- 238000003709 image segmentation Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 4
- 238000004422 calculation algorithm Methods 0.000 description 19
- 239000013598 vector Substances 0.000 description 11
- 230000008901 benefit Effects 0.000 description 9
- 241000196324 Embryophyta Species 0.000 description 6
- 238000003066 decision tree Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000007636 ensemble learning method Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 241000219991 Lythraceae Species 0.000 description 2
- 235000014360 Punica granatum Nutrition 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 208000024386 fungal infectious disease Diseases 0.000 description 2
- 238000003711 image thresholding Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 241000894006 Bacteria Species 0.000 description 1
- 241000221785 Erysiphales Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 235000007688 Lycopersicon esculentum Nutrition 0.000 description 1
- 241000244206 Nematoda Species 0.000 description 1
- 241000233679 Peronosporaceae Species 0.000 description 1
- 240000003768 Solanum lycopersicum Species 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000003898 horticulture Methods 0.000 description 1
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003612 virological effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The present disclosure relates to a process for grape leaf disease identification using
Machine Learning techniques. This automatic system can be used to detect the type of
disease and to take appropriate action. This uses image processing and machine learning
techniques. The Architecture of the present disclosed invention consists of five different
processes such as image preprocessing, image segmentation, feature extraction, disease
detection, and identification. The system of disease identification comprises a filtering unit to
filter the noise from the grape leaves, a segmentation unit for segmenting the image from the
background image, a feature extraction unit for extracting feature and color from the disease
region, and a training module consisting plurality of classifiers for classifying the extracted
features into healthy, rot, esca, and leaf blight.
18
100
102
re-procesing a eleavesirmas forfiltenn noiseusin a Gaussian filter
segmentingtheleafpacrtftheimagefronmthebadgroundimagenusingaGrabcutsegmertation 104
techrnue
extractingdiseaseregioncontainingleions,coloredspots,and someyellowishpartoftheleaf 106
fronthesementedleafpart
IA
extractingtextureandcolorfeaturesconsisting ofenergyhonogeneitycontrast dissimilarity, G108
correlation, and angularsecondmorners fromthe disease region
classifying theextracted featiesinto healthy,rot,eca, andleafblightusing classifiers. 110
Figure1
200
lteng ut entationunit204 Featureextradionurt
206
Trainingmodule208
Figure 2
Description
102 re-procesing a eleavesirmas forfiltenn noiseusin a Gaussian filter
segmentingtheleafpacrtftheimagefronmthebadgroundimagenusingaGrabcutsegmertation 104 techrnue
extractingdiseaseregioncontainingleions,coloredspots,and someyellowishpartoftheleaf 106 fronthesementedleafpart
IA extractingtextureandcolorfeaturesconsisting ofenergyhonogeneitycontrast dissimilarity, G108
correlation, and angularsecondmorners fromthe disease region
classifying theextracted featiesinto healthy,rot,eca, andleafblightusing classifiers. 110
Figure1
200
lteng ut entationunit204 Featureextradionurt 206 Trainingmodule208
Figure 2
A Process for grape leaf disease identification using machine learning techniques
The present disclosure relates to a process for grape leaf disease identification using Machine Learning techniques.
Indian Economy is highly dependent on the agricultural productivity of the country. Grape is a very commercial fruit of India. It can easily be grown in all tropical, subtropical, and temperate climatic regions. India has got different types of climate and soil in different parts of the country. This makes grapevines a major vegetative propagated crop with high socio-economic importance. The grape plant will cause poor yield and growth when affected by diseases. The diseases are due to viral, bacteria, and fungi infections which are caused by insects, rust, and nematodes, etc. These diseases are judged by the farmers through their experience or with the help of experts through naked eye observation which is not an accurate and time-consuming process. Early detection of disease is then very much needed in the agriculture and horticulture field to increase the yield of the crops.
Web-enabled disease detection systems have been disclosed in one of the solutions. The system disclosed a segmentation method which has used a mean-based strategy for computing threshold and textual features were extracted and classification was done by SVM.
The survey disclosed in one of the solutions discusses different disease classification techniques used for plant leaf disease and used genetic algorithm for image segmentation. An integrated approach of particle swarm optimization and SVM for plant leaf disease detection and classification was disclosed in one of the solutions. A disease detection system for pomegranate leaves that used color-based segmentation and features like color, morphology, and texture for classifying the leaves was disclosed in one of the solutions. A leaf detection and climatic parametric monitoring of plants using IoT was disclosed in one of the solutions. Neural Network-based classification was disclosed for detecting plant leaf diseases based on the texture features extracted using the GLCM matrix. SVM-based classification by extracting the texture-based features was disclosed in one of the solutions. SVM classifier with different kernel functions including Cauchy kernel, Invmult Kernel, and Laplacian Kernel were employed to evaluate the ability of the approach to detect and identify the infected tomato leaf. Leaf detection system for pomegranate leaves which uses K-means for segmentation and statistical features for classification using SVM was disclosed in one of the solutions. A system for leaf disease classification using a decision tree by extracting different features after segmenting the leaf using ostu thres holding was disclosed in one of the solutions. A system for two types of disease classification such as Downy mildew and Powdery mildew in grape leaves was using Back propagation Neural Network was disclosed in one of the solutions. A fast system for disease detection and classification using Neural Network after extracting the texture features using gray level co-occurrence methodology was disclosed in one of the solutions. A smartphone-based system using machine learning techniques to detect the state of the disease of the plant and also the severity levels of each disease was developed in one of the solutions. Machine leaming-based techniques such as decision tree, Navie Bayes theorem, Neural Network, K-Means, and Random forest algorithms for leaf disease classification using the features such as size, shape, dryness, and wilting were disclosed in one of the solutions. The plurality of disclosure in the literature uses K-means segmentation for segmenting the leaf and extract low-level features of the image to classify the plant leaf diseases.
However, to ensure good quality and high production, it is necessary to have accurate disease diagnosis and control actions to prevent them in time. Grape is a widely grown crop in India and it May be affected by different types of diseases on leaf, stem, and Fruit. In order to overcome the aforementioned drawbacks, there exists a need to develop a process for grape leaf disease identification using Machine Learning techniques.
The present disclosure seeks to provide a process for grape leaf disease identification using image processing and machine learning technique that can be used to detect the type of diseases and to take appropriate actions.This uses image processing and machine learning techniques. The Architecture of the present disclosed invention consists of five different processes such as image preprocessing, image segmentation, feature extraction, disease detection, and identification. The system segments the leaf (Region of Interest) from the background image using grab cut segmentation method. From the segmented leaf part the diseased region is fruther segmented based on two different methods such as global thres holding and using semi-supervised technique. The features are extracted from the segmented diseased part and it has been classified as healthy, rot, esca, and leaf blight using different machine learning techniques such as Support Vector Machine (SVM), adaboost and Random Forest tree.
The present disclosure seeks to provide a process for grape leaf disease identification using Machine learning technique. The process comprises:pre-processing grape leaves images for filtering noise using a Gaussian filter;segmenting the leaf part of the image from the background image using a Grab cut segmentation technique; extraction disease region containing lesions, colored spots, and some yellowish part of the leaf from the segmented leaf part;extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation, and angular second moment from the disease region; andclassifying the extracted features into healthy, rot, esc a, and leaf blight using classifiers.
The present disclosure also seeks to provide a system for grape leaf disease identification using machine Learning techniques, the system comprises:a filtering unit equipped with a Gaussian filter for filtering noise from grape leaves images;a segmentation unit employing a Grabcut segmentation technique for segmenting the leaf part of the image from the background image;wherein disease region containing lesions, colored spots, and some yellowish part of the leaf is further extracted from the segmented leaf part;a feature extraction unit for extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation, and angular second moment from the disease region; anda training module consists of a plurality of classifiers for classifying the extracted features into healthy, rot, esca, and leaf blight.
An objective of the present disclosure is to, segments the leaf (Region of Interest) from the background image using the grab cut segmentation method.
Another object of this disclosure is to, classify the segmented leaves as healthy, black-rot, esca, and leaf blight.
Another object of the present disclosure is to, scale down the images to a standard width and height,to reduce the computational complexity.
Another object of the present disclosure is to, further segment the disease based on two different methods such as global thres holding and using semi-supervised technique.
Yet another object of the present disclosure is to provide an automated disease detection and classification system for grape leaves using traditional image processing and machine learning techniques.
To further clarify advantages and features of the present disclosure, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
Figure 1 illustrates a flow chart of a disease identification process for grape leaf using machine learning techniques in accordance with an embodiment of the present disclosure; Figure 2 illustrates a block diagram of a grape leaf disease identification system using machine learning techniques in accordance with an embodiment of the present disclosure; Figure 3 illustrates different types of disease in grape leaves in accordance with an embodiment of the present disclosure; Figure 4 illustrates the Architecture of the present disclosed system in accordance with an embodiment of the present disclosure; Figure 5 illustrates the Result of applying Grabcut segmentation algorithm for a sample image in accordance with an embodiment of the present disclosure; Figure 6 illustrates the Identification of disease part of the leaves using Global Thres holding in accordance with an embodiment of the present disclosure; Figure 7 illustrates the Identification of disease part of the leaves using Semi Supervised learning in accordance with an embodiment of the present disclosure;
Further, skilled artisans will appreciate that elements in the drawings are illustrated for simplicity and may not have been necessarily been drawn to scale. For example, the flow charts illustrate the method in terms of the most prominent steps involved to help to improve understanding of aspects of the present disclosure. Furthermore, in terms of the construction of the device, one or more components of the device may have been represented in the drawings by conventional symbols, and the drawings may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the drawings with details that will be readily apparent to those of ordinary skill in the art having benefit of the description herein.
For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates.
It will be understood by those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not intended to be restrictive thereof.
Reference throughout this specification to "an aspect", "another aspect" or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, appearances of the phrase "in an embodiment", "in another embodiment" and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment.
The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process or method that comprises a list of steps does not include only those steps but may include other steps not expressly listed or inherent to such process or method. Similarly, one or more devices or sub-systems or elements or structures or components proceeded by "comprises...a" does not, without more constraints, preclude the existence of other devices or other sub-systems or other elements or other structures or other components or additional devices or additional sub-systems or additional elements or additional structures or additional components.
Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The system, methods, and examples provided herein are illustrative only and not intended to be limiting.
Embodiments of the present disclosure will be described below in detail with reference to the accompanying drawings.
Figure 1 illustrates a flow chart of a disease identification process for grape leaf using machine learning techniques in accordance with an embodiment of the present disclosure. At the step 102 the process 100 includes, pre-processing grape leaves images for filtering noise using a Gaussian filter. The images are acquired from the web and are from different sources and sizes. The images also contain noise due to bad lighting conditions, weather occlusion, etc. To reduce the computational complexity the images are scaled down to a standard width and height. This scaled image is then processed to filter the noise using a Gaussian filter. The Gaussian blur is a low pass filter that reduces the high-frequency components, 5*5 kernel size is used to filter the noise.
At step 104, process 100 includes, segmenting the leaf part of the image from the background image using a Grabcut segmentation technique. From the preprocessed image, the leaf part of the image is segmented from the background image Grabcut segmentation algorithm. This algorithm label a pixel as foreground or background using Gaussian Mixture Model (GMM) and also takes the initial rectangle which is a rough segmentation between background and foreground. A rectangle of dimensions (10, 10, w-30, and h-20) is used as the bounding box where w and h are the width and height of the image.
At step 106, process 100 includes, extracting the disease region containing lesions, colored spots, and some yellowish parts of the leaf from the segmented leaf part. From extracted foreground i.e. the leaf part, the diseased parts are extracted. The disease part contains lesions, colored spots, and some yellowish parts of the leaf. For extracting diseased regions from the leaves there are two different methods. Global thres holding and semi supervised learning.
At step 108, process 100 includes, extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation, and angular second moment from the disease region. The texture of an image is usually expressed, by contrast, uniformity, entropy, etc. A statistical method of examining the texture of an image is the Gray-level Co occurrence matrix (GLCM). GLCM extracts second-order statistical texture features from the training images. GLCM an (NXN) matrix where N is the number of grey levels in the image. Before extracting GLCM features, wavelet decomposition of an image is done. The Discrete Wavelet Transform decomposes an image into different sub-band images such as LL (low low), LH (low- high), HL (high-low), and HH (high-high). LL image is used for texture analysis as it contains the highest quantity of information. In a GLCM matrix M, if i represents a row and j represents a column, the element M (i, j) represents the count of a pair of two pixels having i and j as their intensities that are separated by a distance d and lies along a direction specified by an angle theta. The algorithm takes d and theta values as input by the user. In this system, six texture features were extracted namely, energy (randomness of intensity distribution), homogeneity (closeness of distribution of GLCM element to the diagonal of GLCM), contrast (amount of local variations), dissimilarity, correlation (image linearity), and angular second moment in combination with different values of d and theta. The values of d and theta are chosen as (1, 2, 3) and (0, 45, 90, 135) respectively. So the total number of features are (6 x 3 x 4) = 72 texture features for an image. After segmentation of the disease-affected portion of the leaf, color features were examined to find whether the given input leaf is healthy or not. If the input leaf image is a healthy leaf then the segmented image will not have any region and it contains only black pixels. For healthy leaves image color feature vector contains only Os whereas for non-healthy leaves feature vector contains some color information.
At step 110, process 100 includes classifying the extracted features into healthy, rot, esca, and leaf blight using classifiers. The extracted feature vectors are then used to train different classifiers and the results were analyzed. Support Vector Machine: Given a labeled training data, SVM outputs an optimal separating hyperplane. This hyper- plane categorizes new data point into classes. In order to improve the accuracy of SVM, some parameters of the SVM classifier need to be tuned. One of the parameters is kernel which defines whether separation should be linear or non- linear. Another parameter is regularization which defines the extent to which misclassification of a training sample needs to be avoided. Linear kernel and regularization parameter with value 1000 is used in this system. A larger value of regularization chooses a small margin of hyper plane if it ensures minimum misclassification of training examples.Random Forest: Random Forest is an ensemble learning method and also a supervised learning algorithm. It builds a forest of decision trees. Many trees fit into a random forest classifier. The extracted feature vector is passed as an input vector to each tree of the forest and a decision rule is obtained. In other words, the trees vote for a class. The class having the majority votes by the trees is chosen by the forest. AdaBoost: AdaBoost is used to boost the performance of a machine learning algorithm. It is used with weak learners. Weak models are trained using weighted training data where each instance is weighted. After training the model, the misclassification rate is calculated. This error rate is modified further and used to update the weights of training instances. The purpose of weight updating is to give more weight to misclassified instances. This process continues until there is no scope for improvement. In this way, the AdaBoost algorithm improves the learning of weak learners.
Figure 2 illustrates a block diagram of a grape leaf disease identification system using machine learning techniques in accordance with an embodiment of the present disclosure. The system 200 includes a filtering unit 202. A filtering unit is equipped with a Gaussian filter for filtering noise due to bad lighting conditions, weather occlusion, etc. from grape leaves images. In an embodiment, the segmentation unit 204 employing a Grabcut segmentation technique for segmenting the leaf part of the image from the background image using a gaussian mixture model (GMM) and also takes an initial rectangle which is a rough segmentation between background and foreground.
In an embodiment, the feature extraction unit 206 is employed in the system for extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation, and angular second moment from the disease region. These features represent certain distinctive characteristics that can be used for differentiating among the categories of input patterns. In the present disclosure, texture and color features of the images are used for classification. In an embodiment, the training module 208 consists of plurality of classifiers for classifying the extracted features into healthy, rot, esca, and leaf blight. The extracted feature vectors are then used to train different classifiers and the results were analyzed.Given a labeled training data, SVM (Support Vector Machine) outputs an optimal separating hyper plane. This hyper-plane categorizes new data point into classes.Random Forest is an ensemble learning method and also a supervised learning algorithm. It builds a forest of decision trees. AdaBoost is used to boost the performance of a machine learning algorithm. It is used with weak learners. Weak models are trained using weighted training data where each instance is weighted.
Figure 3 illustrates different types of disease in grape leaves in accordance with an embodiment of the present disclosure. These diseases are caused due to fungi infection on the leaves. Each disease has different characteristics where black rot appears to be circular in shape and has dark margins, esca appears as dark red stripes and leaf blight appears to be solid reddish-purple spots.
Figure 4 illustrates the Architecture of the present disclosed system in accordance with an embodiment of the present disclosure. The Architecture consists of five different processes such as image preprocessing, image segmentation, feature extraction, disease detection, and identification. Image Preprocessing:The images are acquired from the web and are from different sources and sizes. The images also contain noise due to bad lighting conditions, weather occlusion, etc. To reduce the computational complexity the images are scaled down to a standard width and height. This scaled image is then processed to filter the noise using a Gaussian filter. The Gaussian blur is a low pass filter that reduces the high-frequency components, 5*5 kernel size is used to filter the noise B. Image Segmentation From the preprocessed image, the leaf part of the image is segmented from the background image Grabcut segmentation algorithm. This algorithm label a pixel as foreground or background using Gaussian Mixture Model (GMM) and also takes the initial rectangle which is a rough segmentation between background and foreground. A rectangle of dimensions (10, 10, w-30, and h-20) is used as the bounding box where w and h are width and height of the image. From extracted foreground i.e. the leaf part, the diseased parts are extracted. The disease part contains lesions, colored spots, and some yellowish parts of the leaf. For extracting diseased regions from the leaves two different methods have been used. 1) Diseased Part Identification- Global Thres holding: In this method, the RGB image is converted into grey scale image and then global thres holding is applied to convert the image into a binary image. On the thres holded image, connected component labeling is applied to find the contours. The contour with the largest area is then identified and morphological operations such as dilation and erosion are applied. The original image is converted to an HSV image and in the h channel, thres holding is applied. Binary AND operator is then applied to the contour detected image and the HSV image. The resultant image was again thres holded using binary invert thres holding. 2) Diseased Part Identification - Semi supervised Learning: The diseased part of the leaves generally appears in blue color in the BGR image. To segment, the diseased part blue color pixels are filtered out by converting the RGB image into a BGR image. To filter blue color pixels training image is used to find the lower and upper boundary of blue color pixels. The pixels which lie within the lower and upper boundary are then filtered as blue pixels from the input image. In the filtered image thres holding is applied and finally, the diseased areas are identified. Feature Extraction: Image features provide rich information about the content of the image. These features represent certain distinctive characteristics that can be used for differentiating among the categories of input patterns. In the present discloser texture and color features of the images are used for classification. The texture of an image is usually expressed by contrast, uniformity, entropy etc. A statistical method of examining the texture of image is Gray-level Co-occurrence matrix (GLCM). GLCM extracts second-order statistical texture features from the training images. GLCM an (NXN) matrix where N is the number of grey levels in the image. Before extracting GLCM features, wavelet decomposition of an image is done. The Discrete Wavelet Transform decomposes an image into different sub-band images such as LL (low-low), LH (low-high), HL (high-low), and HH (high-high). LL image is usedfor texture analysis as it contains the highest quantity of information. In a GLCM matrix M, if i represents a row and j represents a column, the element M (i, j) represents the count of a pair of two pixels having i and j as their intensities that are separated by a distance d and lies along a direction specified by an asix texture features namely, energy (randomness of intensity distribution), homogeneity (closeness of distribution of GLCM element to the diagonal of GLCM), contrast (amount of local variations), dissimilarity, correlation (image linearity) and angular second moment in combination with different values of d and theta. The values of d and theta are chosen as (1, 2, 3) and (0, 45, 90, 135) respectively. So the total number of features are (6 x 3 x 4) = 72 texture features for an image. After segmentation of disease affected portion of the leaf, colour features were examined to find whether the given input leaf is healthy or not. If the input leaf image is healthy leaf then the segmented image will not have any region and it contains only black pixels. For healthy leaves image colour feature vector contains only Os whereas for non-healthy leaves feature vector contains some color information. Classification using Different Classifiers:The extracted feature vectors are then used to train different classifiers and the results were analysed. Given a labelled training data, SVM (Support Vector Machine) outputs an optimal separating hyper plane. This hyper plane categorizes new data point into classes. In order to improve the accuracy of SVM, some parameters of the SVM classifier needs to be tuned. One of the parameters is kernel which defines whether separation should be linear or non- linear. Another parameter is regularization which defines the extent to which misclassification of a training sample needs to be avoided. Linear kernel and regularization parameter with value 1000 is used in this system. A larger value of regularization chooses small margin of hyper plane if it ensures minimum misclassification of training examples. Random Forest is an ensemble learning method and also a supervised learning algorithm. It builds a forest of decision trees. Many trees fit into a random forest classifier. The extracted feature vector is passed as input vector to each tree of the forest and a decision rule is obtained. In other words, the trees vote for a class. The class having the majority votes by the trees is chosen by the forest. AdaBoost is used to boost the performance of a machine learning algorithm. It is used with weak learners. Weak models are trained using weighted training data where each instance is weighted. After training the model, the misclassification rate is calculated. This error rate is modified further and used to update the weights of training instances. The purpose of weight updating is to give more weight to misclassified instances. This process continues until there is no scope for improvement. In this way, the AdaBoost algorithm improves the learning of weak learners.
Figure 5 illustrates the Result of applying Grabcut segmentation algorithm for a sample image in accordance with an embodiment of the present disclosure. The images are acquired from the web and are from different sources and sizes. The images also contains noise due to bad lightening condition, weather occlusion etc. To reduce the computational complexity the images are scaled down to a standard width and height. This scaled image are then processed to filter the noise using Gaussian filter. The Gaussian blur is a low pass filter that reduces the high frequency components, 5*5 kernel size is used to filter the noise From the preprocessed image, the leaf part of the image is segmented from the background image Grabcut segmentation algorithm. This algorithm label a pixel as foreground or background using Gaussian Mixture Model (GMM) and also takes initial rectangle which is a rough segmentation between background and foreground. A rectangle of dimension (10, 10, w-30 and h-20) is used as the bounding box where w and h are width and height of the image.
Figure 6 illustrates the identification of disease Part using Global Thres holding. Wherein A is Input Image B is Thres holded Image C is Dilation D is Erosion E is HSV Image F is h channel Image G is Thres holded Image F is Binary AND Between D and G in accordance with an embodiment of the present disclosure. In this method, the RGB image is converted into grey scale image and then global thres holding is applied to convert the image into a binary image. On the thres holded image, connected component labelling is applied to find the contours. The contour with the largest area is then identified and morphological operations such as dilation and erosion is applied. The original image is converted to an HSV image and in the h channel, thres holding is applied. Binary AND operator is then applied to contour detected image and the HSV image. The resultant image was again thres holded using binary invert thres holding.
Figure 7 illustrates the identification of the disease Part using semi supervised learning. Wherein A is Input Image B is BGR Image C is HSV Image D is Filtered Blue pixels in accordance with an embodiment of the present disclosure. The diseased part of the leaves generally appears in blue color in BGR image. To segment the diseased part blue color pixels are filtered out by converting the RGB image into BGR image. To filter blue color pixels training image is used to find the lower and upper boundary of blue color pixels. The pixels which lie within lower and upper boundary is then filtered as blue pixels from the input image. In the filtered image thres holding is applied and finally the diseased areas are identified.
The drawings and the forgoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, orders of processes described herein may be changed and are not limited to the manner described herein. Moreover, the actions of any flow diagram need not be implemented in the order shown; nor do all of the acts necessarily need to be performed. Also, those acts that are not dependent on other acts may be performed in parallel with the other acts. The scope of embodiments is by no means limited by these specific examples. Numerous variations, whether explicitly given in the specification or not, such as differences in structure, dimension, and use of material, are possible. The scope of embodiments is at least as broad as given by the following claims.
Benefits, other advantages, and solutions to problems have been described above with regard to specific embodiments. However, the benefits, advantages, solutions to problems, and any component(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential feature or component of any or all the claims.
Claims (10)
1. A process for grape leaf disease identification using Machine Learning techniques, the process comprises: pre-processing grape leaves images for filtering noise using a gaussian filter; segmenting the leaf part of the image from the background image using a Grabcut segmentation technique; extracting disease region containing lesions, colored spots and some yellowish part of the leaf from the segmented leaf part; extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation and angular second moment from the disease region; and classifying the extracted features into healthy, rot, esca, and leaf blight using classifiers.
2. The process as claimed in claim 1, wherein the images are scaled down to a standard width and height prior to pre-processing to reduce the computational complexity.
3. The process as claimed in claim 1, wherein the Grabcut segmentation technique labels a pixel as foreground or background using a Gaussian Mixture Model (GMM) for extracting the disease region from the foreground pixel.
4. The process as claimed in claim 3, wherein the disease region is extracted from the foreground pixel using global thres holding and semi-supervised learning.
5. The process as claimed in claim 4, wherein identification of diseased region using global thres holding comprises:
converting RGB image into grey scale image and applying global thres holding to convert the image into a binary image; finding contours with the largest area from the binary image by applying connected component labelling applying morphological operations consists dilation and erosion; converting the original image into an HSV image and thereafter applying thres holding in the h channel; applying binary AND operator to contour detected image and the HSV image; and thres holding resultant image using binary invert thres holding for identifying disease region.
6. The process as claimed in claim 4, wherein identification of diseased region using semi supervised learning comprises:
filtering blue color pixels upon converting the RGB image into BGR image to segment the diseased region; finding lower and upper boundary of blue color pixels using training image; filtering pixels lies within lower and upper boundary as blue pixels from input image; and thres holding filtered image and thereby identifying the disease region.
7. The process as claimed in claim 1 and 6, comprises a statistical method for examining the texture of image using Gray-level Co-occurrence matrix (GLCM) upon extracting second order statistical texture features from the training images.
8. The process as claimed in claim 1, wherein the classifiers selected from the group of classifiers including support vector machine (SVM), random forest, and AdaBoost is continuously trained using extracted features.
9. The process as claimed in claim 7, wherein wavelet decomposition of an image performed before extracting GLCM features, wherein the discrete wavelet transform decomposes the image into different sub-band images such as LL (low-low), LH (low high), HL (high-low), and HH (high-high) in which LL image is used for texture analysis.
10. A system for grape leaf disease identification using machine Learning techniques, the system comprises: a filtering unit equipped with a gaussian filter for filtering noise from grape leaves images; a segmentation unit employing a Grabcut segmentation technique for segmenting the leaf part of the image from the background image; wherein disease region containing lesions, colored spots and some yellowish part of the leaf is further extracted from the segmented leaf part; a feature extraction unit for extracting texture and color features consisting of energy, homogeneity, contrast, dissimilarity, correlation and angular second moment from the disease region; and a training module consists a plurality of classifiers for classifying the extracted features into healthy, rot, esca, and leaf blight.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2021101496A AU2021101496A4 (en) | 2021-03-23 | 2021-03-23 | A Process for grape leaf disease identification using machine learning techniques |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2021101496A AU2021101496A4 (en) | 2021-03-23 | 2021-03-23 | A Process for grape leaf disease identification using machine learning techniques |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2021101496A4 true AU2021101496A4 (en) | 2021-05-13 |
Family
ID=75829034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2021101496A Ceased AU2021101496A4 (en) | 2021-03-23 | 2021-03-23 | A Process for grape leaf disease identification using machine learning techniques |
Country Status (1)
| Country | Link |
|---|---|
| AU (1) | AU2021101496A4 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113295690A (en) * | 2021-05-17 | 2021-08-24 | 福州大学 | Strawberry maturity rapid discrimination method based on machine learning |
-
2021
- 2021-03-23 AU AU2021101496A patent/AU2021101496A4/en not_active Ceased
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113295690A (en) * | 2021-05-17 | 2021-08-24 | 福州大学 | Strawberry maturity rapid discrimination method based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jaisakthi et al. | Grape leaf disease identification using machine learning techniques | |
| Iqbal et al. | Detection of potato disease using image segmentation and machine learning | |
| Tan et al. | Recognising blueberry fruit of different maturity using histogram oriented gradients and colour features in outdoor scenes | |
| Palacios et al. | Automated grapevine flower detection and quantification method based on computer vision and deep learning from on-the-go imaging using a mobile sensing platform under field conditions | |
| Dias et al. | Apple flower detection using deep convolutional networks | |
| Dubey et al. | Apple disease classification using color, texture and shape features from images | |
| Shah et al. | A survey on detection and classification of rice plant diseases | |
| Waghmare et al. | Detection and classification of diseases of grape plant using opposite colour local binary pattern feature and machine learning for automated decision support system | |
| CN112101271B (en) | Hyperspectral remote sensing image classification method and device | |
| Doh et al. | Automatic citrus fruit disease detection by phenotyping using machine learning | |
| Raut et al. | Plant disease detection in image processing using MATLAB | |
| Arivazhagan et al. | Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features | |
| Syazwani et al. | Automated image identification, detection and fruit counting of top-view pineapple crown using machine learning | |
| Iniyan et al. | Plant disease identification and detection using support vector machines and artificial neural networks | |
| Patil et al. | Grape leaf disease detection using k-means clustering algorithm | |
| AlShahrani et al. | Automated system for crops recognition and classification | |
| Liu et al. | Identifying immature and mature pomelo fruits in trees by elliptical model fitting in the Cr–Cb color space | |
| Noon et al. | Handling severity levels of multiple co-occurring cotton plant diseases using improved YOLOX model | |
| Khan et al. | Machine learning-based detection and classification of walnut fungi diseases | |
| Huang et al. | GCS-YOLOV4-Tiny: A lightweight group convolution network for multi-stage fruit detection | |
| Fiona et al. | Identification of ripe and unripe citrus fruits using artificial neural network | |
| AU2021101496A4 (en) | A Process for grape leaf disease identification using machine learning techniques | |
| Asriny et al. | Transfer learning VGG16 for classification orange fruit images | |
| SANJU et al. | An automated detection and classification of plant diseases from the leaves using image processing and machine learning techniques: A state-of-the-art review | |
| Umar et al. | Precision Agriculture Through Deep Learning: Tomato Plant Multiple Diseases Recognition with CNN and Improved YOLOv7 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |