NDVI and 2D-CNN Plant Leaf Classification

This repository contains MATLAB implementations for classifying plant leaves based on NDVI and deep learning using 2D-CNN (AlexNet). It includes MATLAB App Designer integration, image pre-processing for training, and classification of healthy and diseased plant leaves.

Table of Contents

• Introduction
• Script Code
  • Image Resizing
  • NDVI Classification Using MATLAB App Designer
  • 2D-CNN Classification Using AlexNet
• Dataset
• Results and Analyses
• Usage

Introduction

This project aims to classify plant leaves into healthy and unhealthy categories using two main methods:

1. NDVI-based classification through MATLAB App Designer.
2. Deep learning classification using a 2D Convolutional Neural Network (CNN) based on AlexNet.

Key Objectives:

• Perform NDVI-based image classification using near-infrared (NIR) and RGB images.
• Use transfer learning on AlexNet for classification of diseased and healthy plant leaves.
• Integrate NDVI and CNN classification in a MATLAB App Designer interface.

NDVI Classification Using MATLAB App Designer

The MATLAB App Designer interface allows users to input two sets of images (NIR and RGB) for NDVI classification. It outputs the NDVI image and classifies objects in the image.

Main Features:

• Input NIR and RGB images.

• Generate NDVI images.

• Classify vegetation based on the NDVI values.

NDVI Application Developed in MATLAB

The NIR_RGB.m script implements an NDVI calculator.

Here is the result for the NDVI classification:

2D-CNN Classification Using AlexNet

The DL_classifier.m script implements a deep learning-based classification using AlexNet. It trains a model on plant leaf images and classifies them as healthy or diseased.

Dataset

The following species were considered for classification, each having images of healthy and unhealthy leaves:

Species Name	Healthy Leaves	Unhealthy Leaves	Image Size (pixels)
Alstonia Scholaris	179	179	227 x 227
Mango	170	170	227 x 227
Lemon	159	159	227 x 227
Guava	142	142	227 x 227
Jamun	279	279	227 x 227
Pomegranate	272	272	227 x 227
Pongamia Pinnata	276	276	227 x 227

Dataset source: Chouhan, Siddharth Singh, Uday Pratap Singh, Ajay Kaul, and Sanjeev Jain. "A data repository of leaf images: Practice towards plant conservation with plant pathology." In 2019 4th International Conference on Information Systems and Computer Networks (ISCON), IEEE, 2019, pp. 700-707.

Each image was resized to fit CNN architecture. Code to resize pictures is also provided.

Results and Analyses

Summary of Deep Learning Classification Accuracy

The accuracy results for the classification of each species are summarized below:

Species Name	Accuracy (%)
Alstonia Scholaris	91.4
Mango	100
Lemon	82.7
Guava	100
Jamun	97.8
Pomegranate	97.1
Pongamia Pinnata	96.7

From the results, we can conclude that the combination of vegetation index (NDVI) and artificial intelligence techniques like 2D-CNN can effectively classify healthy and diseased plant leaves.

Jamun Result

Here is the result for the Jamun species classification:

The classification accuracy and other relevant details are discussed above.

References

1. Brown, M. and Süsstrunk, S., “Multi-spectral SIFT for scene category recognition.” In CVPR 2011, IEEE, 2011, pp. 177-184.

2. Chouhan, Siddharth Singh, Uday Pratap Singh, Ajay Kaul, and Sanjeev Jain. "A data repository of leaf images: Practice towards plant conservation with plant pathology." In 2019 4th International Conference on Information Systems and Computer Networks (ISCON), IEEE, 2019, pp. 700-707.

3. https://www.mathworks.com/help/wavelet/ug/classify-time-series-using-wavelet-analysis-and-deep-learning.html

4. Chamara, N. (2021). Development of an Internet of Things (IoT) Enabled Novel Wireless Multi Sensor Network for Infield Crop Monitoring.

Script Code

Image Resizing

The Image_Resizev2.m script resizes large plant leaf images to a resolution of 227 x 227 pixels, making them suitable for input into AlexNet for 2D-CNN classification.

2D-CNN Classification Using AlexNet

The DL_classifier.m script implements a deep learning-based classification using AlexNet. It trains a model on plant leaf images and classifies them as healthy or diseased.

Citing This Repository

If you find this repository useful for your research, please cite it as follows:

Adarsha. (2024). NDVI-and-2D-CNN-Plant-Leaf-Classification. GitHub. https://github.com/adarsha30735/NDVI-and-2D-CNN-Plant-Leaf-Classification

Full Code:

DatasetPath = ('E:\\Test\\Pongamia Pinnata (P7)'); % Choose your own path
images = imageDatastore(DatasetPath, 'IncludeSubfolders', true, 'LabelSource', 'foldernames');
numTrainFiles = 230;
[TrainImages, TestImages] = splitEachLabel(images, numTrainFiles, 'randomize');

net = alexnet;
layersTransfer = net.Layers(1:end-3);
numClasses = 2;
layers = [
    layersTransfer
    fullyConnectedLayer(numClasses, 'WeightLearnRateFactor', 20, 'BiasLearnRateFactor', 20)
    softmaxLayer
    classificationLayer
];

options = trainingOptions('sgdm', 'MiniBatchSize', 20, 'MaxEpochs', 8, 'InitialLearnRate', 0.0001, 'Shuffle', 'every-epoch', 'ValidationData', TestImages, 'ValidationFrequency', 10, 'Verbose', false, 'Plots', 'training-progress');
netTransfer = trainNetwork(TrainImages, layers, options);

Ypred = classify(netTransfer, TestImages);
YValidation = TestImages.Labels;
accuracy = sum(Ypred == YValidation) / numel(YValidation);

% Plot confusion matrix
plotconfusion(YValidation, Ypred);


#### Full Code: Image Resizing
```matlab
output_address = ('E:\\UNL\\semester 1\\CSI 8300 cv and ıp\\PA_B\\leaf\\Test\\Pongamia Pinnata (P7)\\Unhealthy Plants');
current_folder = dir('E:\\UNL\\semester 1\\CSI 8300 cv and ıp\\PA_B\\leaf\\Test\\Pongamia Pinnata (P7)\\diseased\\*.jpg');

for i = 1 : length(current_folder)
    current_image = current_folder(i).name;
    Img = imread(fullfile(current_folder(i).folder, current_image));
    j = imresize(Img, [227, 227], 'bilinear');
    imwrite(j, fullfile(output_address, current_image));
end