Comparative Study of Major Algorithms for Pest Detection in Maize Crop

Abstract

Early-stage control of maize plant pest is a big issue in precision agriculture. Farmers can prevent pest only if they identified as early as possible and can avoid economic losses also. This type of process can minimize the usage of pesticides and can bring healthy crop. Here algorithm features are takes place for comparative study to identify the training time, training data, merits, demerits and accuracy. In this study, we found Faster R-CNN is good in accuracy but time delay was high. Faster R-CNN convolutional neural network is desirable to develop user friendly system for the farmers that can automatically identify pest. The object detection method can be further improved by adopting the proposed model. We proposed new algorithm to replace the drawback and to provide best result in both accuracy and time delay. IoU method supports to find the prediction object from the real object, hence the pest can identify from the crop. The pseudocode can use to develop the real time system to bring out the process in effective and also efficient manner. Some of the samples have initialize to check the performance metrics of the proposed algorithm.