An Android Application For Pest Detection and Classification Using Regional Convolution Neural Network

Abstract

Different species of pest affect the agricultural crops in the fields, which causes less yield in crops production. Identification of pest in the crop field require manual detection and classification, which consumes large amount of time and manual power. In addition, existing pest identification and classification process based on simple convolution neural networks, which requires high data set collection and high requirements. To overcome these problems an effective image classification technique for CNN based process in deep leaning concept by creating an android application for real time detection in the crop field. It uses the selection search method to detect the important regions from the given image. RCCN algorithm is used to train and testing of the data-sets. The classification of the pest uses the support vector machine (SVM) that is fused with the CNN. The accuracy of the pest detection is increased using the region of interest (RoI) technique for providing the necessary pest regions for detection and classification process. Pest features and detection score has the advantage to provide the identification score of the detected pest in real time using the bounding box regression method.