Qingjun Wang, Tong Liu and Mujun Zang
Agricultural pest and disease detection is of great significance in ensuring crop yield and quality. In order to achieve real-time and accurate pest and disease identification, this paper proposes a real-time pest and disease detection and recognition algorithm based on deep learning. By introducing convolutional neural network (CNN) and feature pyramid network (FPN), this method improves the recognition ability of complex pest and disease images under multi-scale conditions. The system realizes remote management, data collection and transmission of monitoring equipment in farmland through high-definition cameras, 4G or 5G networks. Data preprocessing technology is used for image enhancement and denoising to improve the feature extraction ability of the model. The trained model is applied to real-time detection in the actual environment to form a historical data set to provide support for the prediction of pest and disease development trends. This study provides efficient and low-cost technical support for pest and disease prevention and control in smart agriculture, and has good applicability and promotion value.
Agricultural pest detection, Deep learning, Real-time recognition, Image processing, Neural network algorithm