Early identification of Tomato Plant Leaf Diseases using Clustering and Neural Networks

Abstract

Agriculture aims at cultivating crops to gratify the human needs. A study claims that nearly 50% of crops are being affected with diseases in early stages. Even though there exists many traditional approaches to eradicate the early stage diseases, the implication of digital technology lays a foundation in a efficient manner. Early diagnosis not only reduces time and money, also avoids spread of diseases to other crops, in turn improving the overall production. By considering this fact in the mind (as the focal area), the paper proposes an approach to categorize diseases in the tomato plant leaf at an early stage(s). The proposed method adopts clustering technique to cluster the disease affected regions and extract the features from those regions subject to classification process. The techniques adopted for clustering, feature extraction and classification are K-Means, Discrete Wavelet Transform (DWT) followed by Gray Level Co-occurrence Matrix (GLCM) and Neural Networks (NN) respectively. The proposed approach is assessed on the standard dataset images from Kaggle repository. Among various classes of leaf categories, randomly 50 images are chosen from each category to evaluate the proposed method. Experimental findings unfold that the proposed disease identification plays a vital role in early stages that the existing state-of-the-algorithms.