Detection of discolored trees infected by pine wilt disease based on improved YOLOv11 algorithm and UAV imagery

To improve the detection accuracy of pine wilt disease based on deep learning, an improved YOLOv11 algorithm was constructed by combining wavelet convolution (WTConv) with triplet attention mechanism (triplet attention) based on YOLOv11 algorithm. Based on UAV visible light remote sensing image, the algorithm was applied to the detection of pine wilt disease in Jinyun county, Zhejiang province. The detection effects of YOLOv8, YOLOv11 and improved YOLOv11 algorithm on infected and dead trees caused by pine wilt disease were compared. The results showed that the improved YOLOv11 algorithm had better performance. The average detection accuracy was 97.7%, the precision was 97.4%, and the recall was 95.4%, which were better than YOLOv8 and YOLOv11 algorithms. In the untrained area, the F1 score of pine wilt disease infected tree was 94.4 %, and the improved YOLOv11 algorithm model could still accurately identify the target in the untrained area. The research results provided a more accurate detection algorithm for pine wilt disease, and provided more accurate tool support for the location of pine wilt disease.