A Detection Method of Pine Wilt Disease Based on Improved YOLOv11 With UAV Remote Sensing Images

Abstract

Pine wilt disease (PWD), also known as pine wilt nematode disease, is a severe forest disease caused by the pine wood nematode (Bursaphelenchus xylophilus), which spreads rapidly and causes severe ecological damage in China. However, the existing detection methods have low accuracy in identifying early-stage infected trees and are prone to missing detections. To address these issues, this study proposes YOLOv11-OC, a detection method of pine wilt disease based on improved YOLOv11 with UAV remote sensing images. The omni-dimensional dynamic convolution (ODConv) is employed to optimize the C3K2 module, thereby enhancing the feature extraction ability for small targets and improving the accuracy of identifying early-stage disease areas. Meanwhile, the context anchor attention (CAA) mechanism is introduced to improve the C2PSA module, enhancing the model's context awareness, effectively reducing missed detections, and improving detection performance in complex backgrounds. Experimental results demonstrate that the proposed YOLOv11-OC algorithm outperformed the original YOLOv11, achieving a precision of 94.2%, a recall of 83.1%, a F1-score of 88.3%, and a mean average precision (mAP) of 88.9%. Compared to the original algorithm, the improved version shows increases of 2.8% in precision, 1.0% in F1-score, and 1.4% in mAP. Furthermore, the algorithm also demonstrates good generalization ability on the public PlantDoc dataset.