Optimal Segmentation with Back-Propagation Neural Network (BPNN) Based Citrus Leaf Disease Diagnosis

In recent years, plant diseases in agriculture are mainly accountable for production loss, which leads to a huge economical loss. Citrus plants are utilized as an important part of nutrients such as vitamin C. At the same time, `Citrus' disease seriously affected the growth and quality of citrus fruits. Recently, computer vision models have been commonly employed to detect and classify plant diseases. In this study, an optimal segmentation with classification model for citrus plant disease is presented. The presented model involves a set of various sub-processes. The citrus lesion spots undergo extraction using a weighted segmentation approach that takes place on improved test image. Next, different characteristics undergo codebook construction. In addition, the optimal characteristics are chosen by a hybridization model. The chosen features are provided to Back Propagation Neural Network (BPNN) for classifying citrus diseases. The presented method is validated on Citrus Disease Image Gallery Dataset. The presented model shows extraordinary results over the compared methods and also attains superior results on the applied test images.