Enhancing Betel Nut Pest and Disease Identification in Hainan, China with Swin Transformer-YOLOv5: A Machine Learning-Based Approach for Improved Precision

Betel nut pests and diseases are one of the important factors affecting the yield of betel nut. As a major province of betel nut production in Hainan, China, betel nut pests and diseases have seriously lost the local economic development. In order to solve the problem of betel nut yield decline caused by betel nut plantation diseases and insect pests, this paper proposes a betel nut pests and diseases identification model based on improved YOLOv5 algorithm. Firstly, the images of seven betel nut pests and diseases were classified and collected, and then the nearest neighbor interpolation method was used to preprocess the images, and the image transformation method was used to enhance the images of betel nut pests and diseases, so as to increase the robustness of the model. Then, the YOLOv5 algorithm combined with Swin Transformer is improved. Finally, the model is trained on the pytorch-based deep learning framework. The experimental results show that the YOLOv5 algorithm combined with Swin Transformer is 1.6 percentage points higher than the unimproved algorithm in Presion, 1.7 percentage points lower in Recall, 6.1 percentage points higher in mAP, and the model converges faster. Finally, the method can successfully identify seven betel nut diseases and insect pests.