Identification of Maize Plant Diseases Based on Linear Vector Quantization with Neural Network

Abstract

Agricultural output is a critical prerequisite for economic success in any country, and it offers raw materials, work, and food to many citizens. A variety of factors contribute to anticipated crop production changes across the globe. Chemical fertilizer usage is one of the most prevalent, but others include the presence of contaminants in water sources, unpredictable rainfall patterns, and changing soil fertility. Apart from these concerns, one of the most common challenges around the world is the destruction of a large portion of productivity due to crop diseases. After contributing effective resources to the farms, the occurrence of pests in the plants grown reduces a significant portion of the production. As a result, reliable techniques for plant disease detection are becoming increasingly important. Using a deep convolutional neural network (CNN), this research paper presents an interesting technique for predicting maize leaf diseases from images of different disease stages. The Linear Vector Quantization (LVQ) augmented CNN is recommended for maize leaf identifying diseases because it improves the accuracy of maize leaf diseases while lowering network parameters. Improved models are utilized to train and evaluate various types of maize leaf images, thereby improving maize leaf diseases accuracy by reducing convergence iterations, potentially improving model training and identification efficiency. The proposed CNN-LVQ is compared with two other contemporary machine learning models, such as VGG-16 and ResNet-50, to analyze its performance.