Efficient Framework for Identification of Soybean Disease Using Machine Learning Algorithms

Abstract

This paper exhibits the detection and classification framework of soybean leaflet diseases. Identification and classification were performed utilizing an k-means algorithm and a multiclass support vector machine (SVM). Healthy and unhealthy leaflets infected by frogeye leaf spot, bacterial blight, and Septoria brown spot diseases were collected from a soybean field. The image database is developed by acquiring images with a constant background using a digital camera under the control environment. The image preprocessing techniques applied to the ROI of the raw image. After that, the partition of the diseased region is done using an k-means segmentation technique. The color and texture features were extracted from the segmented leaf region. The mean and standard deviation of RGB channels estimated to extract color features, and texture features were extracted using a (GLCM) method to define an image feature database. Finally, SVM was employed to identify soybean disease. The accuracy of the intended framework for classifying the defined disease was 90.20%.