Classifying Rice Leaf Diseases through CNN for the Sustainable Agriculture

Abstract

Abstract A major source of revenue and a means of sustenance in India is agriculture. Rice is a staple meal that is farmed in the major areas of India. It has been discovered that diseases significantly harm rice harvests, causing considerable costs for the agricultural sector. Plant pathologists are searching for an accurate and reliable method of identifying the illness afflicting rice plants. One effective use of machine learning in crop remote sensing is the categorization of agricultural illnesses. A major area of research right now for detecting agricultural diseases is deep learning. In this study, an effective Convolution Neural Network (CNN) based technique for detecting leaf disease in rice plants was developed. The major subjects of this study are the three well-known rice illnesses brown spot, hispa and leaf blast. This approach for diagnosing and recognising rice plant disease is based on the size, shape, and colour of lesions in the leaf picture. In Otsu’s global thresholding, the background noise is removed from the picture by binarizing it. The Histogram of gradient image edges and features are then displayed to see whether the image contains vectors that can identify sick regions. This model is used to learn the features after the vectors have been validated. It then divides the rice leaf images into four categories: healthy, brown spot, hispa, and leaf blast. The necessary deep learning toolkit is constructed in MATLAB to use the CNN based rice leaf detection algorithm. The Support Vector Machine (SVM) based classifier is also examined for comparison.