Guava leaf disease detection using support vector machine (SVM)
Keshav Kumar Ray, Anshu Kumari, Sanjeev Kumar, Rajendra Machavaram, Imran Shekh, Sudhir Deshmukh, Prashant Tadge
Abstract
ABSTRACT This study aimed to develop a support vector machine (SVM) model for identify disease in guava crop namely, mummification, dot, canker and rust. The dataset was collected using a Raspberry Pi camera, capturing images of four common guava leaf diseases: Canker, Dot, Rust, and Mummification, along with healthy leaves. The images were pre-processed to enhance quality and convert from RGB to gray scale for feature extraction. Features such as Local Binary Pattern (LBP), Gray-Level Co-occurrence Matrix (GLCM), and HSV colour histograms were extracted to identify disease characteristics. The dataset was split into training (70%), validation (20%), and test (10%) sets. SVM models, using both linear and Radial Basis Function (RBF) kernels, were trained to classify the images based on the extracted features. Model performance was evaluated using metrics like accuracy, precision, recall, and F1-score. The SVM models were validated and tested on unseen data for reliable disease classification.The Linear kernel SVM model achieved an accuracy of 77.08% for detecting guava leaf diseases, with sensitivity, specificity, and precision values of 70.08%, 90.76%, and 93.68%, respectively. However, due to the non-linear nature of the dataset, the RBF kernel SVM model significantly outperformed the linear model, achieving an accuracy of 91.67%. The RBF model also showed improved metrics, including higher precision, sensitivity, and lower false positive and false negative rates, making it more suitable for leaf disease detection.