Research on Multi-Object Tracking of Overhead Transmission Line Components Based on Improved YOLOv8+ ByteTrack

Currently, most inspections of overhead transmission line components primarily rely on object detection. To enhance the data dynamic analysis capability of inspection tasks and the algorithm's resistance to adverse factors such as non-uniform motion of unmanned aerial vehicles and component occlusion, this paper proposes a multi-object tracking method for overhead transmission line components based on improved YOLOv8 with ByteTrack. The aim is to improve the accuracy and efficiency of the algorithm for object detection of overhead transmission line components. Firstly, this paper adds a small object detection layer to the head of YOLOv8 and incorporates an EMA attention mechanism. Secondly, based on the original ByteTrack, global motion compensation is introduced, and Kalman filtering is improved. Experimental results show that the proposed method achieves a mAP50 of 91.1% and mAP(50-95) of 65% in detector performance, representing an improvement of 6.5% and 5.3%, respectively, compared to the original YOLOv8s. In terms of tracker performance, the MOTA is 57.1% and IDF1 is 73.6%, which is an increase of 7.5% and 8.5%, respectively, compared to the original YOLOv8s with ByteTrack. This illustrates the precision and effectiveness of the proposed approach in identifying components on overhead transmission lines.