Research on Banana Leaf Disease Detection Based on the Image Processing Technology

Abstract

The banana diseases directly affected banana quality and yield, so a detection method for banana leaf diseases based on image processing is proposed. Firstly, the color segmentation method is used to remove the green background such as healthy leaves, and weed from original banana leaf disease image collected by smart phone. And the segmented image is converted to V component gray image of YUV color space and then the Ostu segmentation which uses the minium intra-cluster or the maximum inter-cluster gray variance to segment the image, and the area threshold method are used to remove the non green background such as soil, and dead grass, thus extracting a complete disease regions. Next, these color features such as the color means, color variances, and color skewnesses of R, G, B, Y, U and V component in the disease regions, are extracted and secleted to construct a set of feature vector of each banana leaf disease with standard image samples. Finally, the minimum euclidean distance classifier is used for banana leaf disease recognition. The results show that the method can effectively segment the disease regions from the original banana leaf disease image and has high recognition rate with 91.7% for gray leaf spot disease and 90% for sigatoka leaf spot disease. Therefore, it is proven that this method has high accuracy and reliability in banana leaf disease recognition, having better promotion and application value.