Brassicaceae Leaf Disease Detection using Image Segmentation Technique

Abstract

In the present time, leaf disease is one of the major problems of Brassicaceae vegetables in the agriculture domain as it affects the quality and quantity of the vegetables. The most common leaf disease of these vegetables is downy mildew, blight leaf disease, and leaf spot. This paper mainly considers identifying the diseased region of the leaves using the image segmentation technique and presents experimentation of the desired number of clusters k. To address the objectives, image acquisition, pre-processing, segmentation, and emphasizing the affected portion of the leaves are all part of the process of the proposed method. The images were transformed into grayscale and removed from the background using Otsu’s thresholding method. K-means clustering algorithm was applied to segment the different regions of the sample images. Finally, the clustered images were then analyzed using a median filter to emphasize the region of interest of the affected leaves. With the different number of clusters k used, k = 4 was successfully segmented the diseased portion, and it was confirmed by the elbow method. Further, the infected area of the sample images was presented in different colors. Also, the proposed method provides a 96.90% accuracy compared to other image segmentation techniques. Image segmentation has become an effective tool in various applications in the agricultural sector.