GroupTrack: A Feature-aware Enhanced Group Tracking Method

Group target tracking in sequential images is crucial for applications like public security. This paper proposes GroupTrack, a novel group multi-object tracking method based on temporal awareness and multi-modal association. The framework comprises a Feature-aware Enhanced Group Detection Network (FEGDN) and a Multi-Dimensional Similarity-based association strategy (MDS). FEGDN enhances robustness through multi-scale perception and bidirectional feature fusion, while MDS achieves stable trajectory association by fusing appearance, IoU, and scale consistency. Experiments on the dataset from the "National Big Data and Computational Intelligence Challenge"show that GroupTrack achieves a state-of-the-art average score of 88.18%, significantly outperforming advanced baselines. It secured the first prize in the competition, demonstrating its advancement and practical utility.