Fine-Grained Recognition of Crop Pests Based on Capsule Network with Attention Mechanism

Abstract

Crop pest detection and recognition in the field is one of the crucial components in pest management involving detection, localization in addition to classification and recognition which is much more difficult than generic object detection because of the apparent differences among pest species with various shapes, colours and sizes. A crop pest identification method is proposed based on capsule network with attention mechanism (CNetAM). In CAN, capsule network is used to improve classification performance of the traditional convolutional neural network (CNN) and an attention module is added to reduce the noise influence and speedup the network training. The experimental results on a pest image dataset demonstrated that the proposed method is effective and feasible in classifying various types of insects in field crops, and can be implemented in the agriculture sector for crop protection.