부산대학교 도서관

To address the challenges in accurate identification of coal shearer drum teeth, we propose an improved YOLOv8 algorithm for coal shearer drum teeth detection. First, we introduce the C2f-DCNv3 module into the backbone feature extraction network to address the difficulty in capturing small target features. Second, a shuffle attention (SA) mechanism is added to the neck section to help the model more effectively integrate features from different levels, enhancing the accuracy and generalization of the model. Experimental results show that the proposed method achieves precision of 90.6%, recall of 86.8%, and AP@0.5 of 91.7%. Compared to YOLOv8, the detection precision, recall, and AP@0.5 are improved by 2.5%, 1.9%, and 2.0%, respectively. This indicates a significant improvement in the accuracy of detection for coal shearer drum teeth in underground mines.