Plant Leaf Diseases Identification using Convolutional Neural Network with Treatment Handling System
Koay K. Leong, Lim Lian Tze
Abstract
Agriculture is a very crucial industry to Malaysia where it would bring a huge impact on the country's wealth. All plant species, regardless of cultivated or wild, are prone to diseases and it is often inevitable. In the older days, identification of plant diseases was done by the experts of the field through observing the plants manually which can be rather tedious, time consuming and often inaccurate. With the advent of image processing technologies, automatic detection of plant disease can be done by capturing and processing the image of the plant leaves. In this paper, the K-means clustering and color thresholding are applied for image segmentation of the plant leaf. Both algorithms are compared in the study to evaluate which serves as a better segmentation algorithm on a target dataset to segment the region of interest (ROI). Moreover, two feature extraction approaches of ResNet-50 Convolutional Neural Network (CNN) and Gray-Level Co-Occurrence Matrix (GLCM) are also studied to evaluate their performance as a feature extractor. With applying the CNN to the support vector machine (SVM) classifier, it is investigated that an average classification accuracy of 96.63% can be achieved to perform the classification of the leaf diseases. A graphical interface for the system is also developed to provide the treatment handling method for the detected diseases.