The Pest and Disease Identification in the Growth of Sweet Peppers Using Faster R-CNN and Mask R-CNN

Abstract

Early-stage control of plant pests is a crucial topic in modern agriculture. If plant pests and diseases can be identified as early as possible, farmers can prevent their occurrence in advance and avoid economic losses. Early identification of pests and diseases can minimize the cost of pesticides. However, correct identification of pests and diseases requires knowledge and corresponding expertise, and this knowledge accumulation requires time. Therefore, in this study, Faster region-convoluted neural networks (R-CNNs) and Mask R-CNNs were used to develop a knowledge-based system that can automatically identify plant pests and diseases. The Faster R-CNN exhibited a regional recognition accuracy of 89%, and the Mask R-CNN exhibited an area recognition accuracy of 81%. A pest and disease identification system was developed in this study. The developed system can be further improved by adopting the proposed reinforcement model construction flow.