Detection and Classification of Cotton Leaf Diseases: A Comprehensive Approach

In India, Leaf diseases are a main cause of reduced crop productivity. In this paper, we focus on the detection and classification of cotton leaf diseases. We captured the infected leaf diseased images from the agricultural field Anand. We considered totally 105 images of three diseases namely Alternaria leaf spots (35 images), Bacterial blight (35 images), and Nitrogen deficiency (35 images). The captured images were used as image pre-processing. We considered the different background of the images such as cream, sky blue, with shadow and without shadow etc. So, the background of the images was removed from the images as required for the image pre-processing step. Modified TSAI technique was used for background removal. After images were pre-processed, image segmentation step was performed. In image segmentation, color image segmentation and binary segmentation techniques were performed. Using image color segmentation, green pixels were masked and omitted from the image. Otsu method was used for binary image segmentation. A binary image was used to mask the image, and the mask image was used to image and color texture features. The features were extracted and classified using a library support vector machine—LIBSVM multiclass SVM toolbox. With a cost value of 12 and gamma value of 0.03 the technique provided 96% training and 93% testing accuracy.