The Technology of Kiwifruit Pest and Disease Identification Based on Deep Learning

[Purpose] To achieve fast and accurate identification of kiwifruit pests and diseases in real environments, a kiwifruit pest and disease identification method based on improved YOLOv7 is proposed.[Methods] Using the YOLOv7 network model as the base network, the feature extraction and feature fusion capabilities of the network model are improved by introducing the proposed multi-scale feature fusion module to improve the recognition accuracy of various types of kiwifruit pests in a balanced manner; The kiwifruit planting bases in South China were used as sample collection points to construct a kiwifruit pest and disease dataset; then, the collected dataset was enhanced by introducing data enhancement methods such as Random Noise, Mixup, and Cutout according to the characteristics of its sample distribution, which enabled the deep learning model to learn the visual features of pest discriminatory power from a deeper dimension; The YOLOv7 network is improved and a multi-scale neural network model based on feature pyramid is constructed to enhance the recognition accuracy of pest and disease features.[Results] Based on the improved YOLOv7 for kiwifruit pests and diseases, the average accuracy reached 94.43%, which is an increase of 6.51 percentage points in accuracy and 6.73 percentage points in mean average accuracy compared to the base network YOLOv5s.[Conclusion] The method can quickly and accurately identify kiwifruit pest and disease targets in real environments. So it can provide a reference for the intelligent management of kiwifruit plantation industry.