Embedded gpu based agricultural pest classification using machine learning and deep learning techniques

Abstract

Detection of insects is a major challenge in the field of agriculture. newlineTherefore, effective and intelligent systems should be designed to detect the newlineinfestation in minimizing the use of pesticides. newlineDeep Learning (DL) is a common machine learning algorithm used newlinein various applications. There are many deep learning techniques and newlinearchitectures, including Radial Function Networks, Multilayer Perceptrons, newlineSelf-Organizing Maps, Convolutional Neural Networks, and more. Among newlinethem, Convolutional Neural Network (CNN) is frequently used for the newlinerecognition and classification tasks. The CNN has the design to extract a high newlinenumber of features from any given image. Various applications, including newlineplant disease diagnosis, ripening stage of crops and fruits, weed identification, newlineand crop pest identification have utilized CNN for recognition and newlineclassification. newlineIdentifying pests from farmland is tedious. Though researchers newlinehave shown several methods for the recognition and classification of insects, newlinestill several issues and improvises must be addressed. To overcome the newlinebarriers to pest identification and classification, an efficient and memoryconstrained newlinearchitecture is required for a fast classification process in a crop newlinefield. This thesis aims to develop an intelligent insect classification system newlinethat would be capable of detecting and classifying the types of most common newlineinsects in the agriculture field. newline