Framework for Classification of Chilli Leaf Diseases

Abstract

The detection and classification of chilli leaf diseases is a major concern in agriculture. Farmers need to track crop fields and recognize signs of disease at early as possible. Image processing is an aid in the identification and classification of leaf diseases. Cercospora leaf spot, Chilli mosaic, leaf curl, powdery mildew chilli leaf diseases are harmful to chilli production. For the identification of leaf disease, there are basic image features, i.e., texture and color. This study presents feature selection as an optimization problem, and a manual feature selection strategy based on permutation and combination is provided to improve chilli leaf disease classification outcomes. Experiments on a common benchmark reveal that the permutation and combination formula for feature selection is a viable tool for improving classification results. These features are Contrast, Energy, Correlation, Entropy, Cluster shade, Cluster Prominence, Kurtosis, and skewness are top priority features entered in SVM, KNN, and CNN. The selected features provide input for the identification of a disease and are extracted using the GLCM algorithm. There are 5078 samples of four diseases and one healthy leaf collected in this work, out of the 3500 samples are trained and 1579 data are tested. The k-fold method is used to train the model for accuracy evaluation. CNN provides 93.53 % accuracy, SVM 85.07 %, and KNN provides 89.92 % accuracy. newlineSix experiments are carried out on the distance and stages of diseases using chilli leaf images of chilli leaf disease to assess the effectiveness of the designed framework, using training data and classification of the given image dataset. According to experts, the research identifies disease names and provides chemical solutions to diseases. The findings are evaluated in terms of disease name and chemical treatments, and the system code is written in MATLAB. newline newline