Recent Advancements in Image-Based Prediction Models for Diagnosis of Plant Diseases

Abstract

India plays a significant role in the world as a major contributor to the overall food industry. Farming being a major occupation, crop protection from plant diseases has become a serious concern. The occurrence of plant diseases is hugely dependent on environmental factors which are uncontrollable. Ongoing agricultural research and the advanced computational technologies can be coalesced to determine an effective solution, which can improve the yield and result in better harvests. The objective is to minimize the economic and production losses. Researchers have developed several modeling techniques, viz. Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Deep Learning for prediction and preferably early detection of diseases in plants. Mostly, images of the diseased plant are given as input to these prediction models. Early detection leads to minimizing the pesticide usage, hence, resulting in lowering the expense along with the ecological impact. Also, it is essential that the prediction techniques are nondestructive in nature. This paper delves into the Machine Learning based plant disease prediction models, along with a brief study of diverse imaging techniques, viz., RGB, multi- and hyperspectral, thermal, fluorescence spectroscopy, etc., catering to different features/parameters of a plant, useful in the prediction of diseases. Around 35 research papers from reputed peer-reviewed journals were studied and analyzed for this systematic review. It highlights the latest trends in plant disease diagnosis as well as identifies the future scope with the application of modern era technologies.