Detection of Rice Sheath Blight Leaf Based on Improved YOLOV8 Algorithm

Traditional YOLOv8 algorithm fails to achieve the desired detection accuracy in object detection. Therefore, this paper proposes an improved YOLOv8 algorithm by introducing the BiLevel Routing Attention mechanism and the Wasserstein loss function, which significantly improves the accuracy of the YOLOv8 algorithm. Results show that compared to YOLOv8, the proposed system exhibits improved overall recognition accuracy for the detection of withered rice, as evidenced by various evaluation metrics. The overall recognition accuracy reaches 98.7%.