Automated Classification of Bacterial and Fungus Infection in Crops

Abstract

The classification of diseases in the crops and plants are very interesting research area in the field of agricultural image processing. The diseases and the parasite infections present in the crops will multiply uncontrollably if left unnoticed. Hence the detection of such infection at the early stage is imperative for the benefit of the productivity of the crop yield. Automating this process is a cumbersome task as it needs intense mechanism and process to identify the type of infection (normal to abnormal). This paper introduces a new method using gray level co-occurrence matrix GLCM to extract the defects from the images of the crop. The proposed algorithm categorizes the images into two classes (normal and abnormal) and the extracted features are then modeled using support vector machines, decision tree and k-nearest neighbor algorithms. The experimental result showcased that the decision tree performed with 98%accuracy in classifying the abnormal crops and outscored the other two algorithms by a good margin.