Pest Detection in Crops Using Deep Neural Networks

Pests play a significant impact in crop destruction. Currently, crop yields are being reduced as a result of pest- infested crops, resulting in a reduction in output rate. The possibility of identifying the plant disease in an unfavorable environment is not pursued further. The key problem is to reduce pesticide use in the agricultural field while increasing the assembly rate's standard and quantity. This paper is laboring to explore the plant disease prediction at an untimely action. In spite of traditional convolution neural network (CNN)-based approaches, new method for detecting and recognizing insect pests have been created to produced adequate results. To optimize various parameters, CNN-based approaches necessitate a huge dataset. As a result, CNN-based two-stage recognition and identification process for insect pests is proposed. A region suggestion network for insect pest detection using YOLOv3 and a re-identification method are also presented in this paper. To train these models, knowledge augmentation method using image processing is proposed. YOLOv3 provides 92.11% accuracy when compared to CNN during pest detection.