Crop Disease Classification Through Image Processing and Machine Learning Techniques Using Leaf Images

Abstract

The role of the agriculture sector in global economic development is important. The development of agriculture and growth in production is essential for achieving global food security. Diseases in plants/crops are the responsible agents for the loss of agricultural production globally. Most of the diseases in plants initially strike the leaves of the plant, later its symptoms are evident on all parts of the plant. The diseases significantly affect the quality and quantity of total crop production. Typically, plant diseases are identified through visual observation or laboratory investigations by phytopathologists, but this is very challenging for farmers or non-specialists. Image processing and machine learning together can play an important role in helping farmers to identify the diseases in crops. The major steps in such methods generally include image acquisition, image pre-processing, image segmentation, feature extraction, and disease classification. This paper presents an analysis of some of the major classification techniques used in such methods. The experiments are carried out for two important crops Apple (Malus domestica) and Blackgram (Vigna mungo) to analyze baseline classifiers such as decision tree, naive Bayes, logistic regression, k-nearest neighbor, linear discriminant analysis, support vector machine, and random forest using plant leaf images. The leaf images of apple are taken from a benchmark PlantVillage dataset, while images of blackgram (urdbean) leaves are obtained from a self-prepared dataset. In both datasets, the leaf images contain a simple eliminated background.