GradsSORT: Research on Multi-Object Tracking Algorithm Based on Pseudo-Gradient
DOI:
https://doi.org/10.54097/9mg4z136Keywords:
HybridSORT, Multi-object tracking, Pseudo-gradientAbstract
To address the issue of low tracking accuracy for occluded targets in complex scenes, we propose a multi-object tracking algorithm (GradsSORT) based on pseudo-gradient. First, targets are categorized into different gradient layers according to their degree of occlusion. Then, a hierarchical association strategy is adopted during data matching, prioritizing the association of detection boxes in lower gradient layers. This approach helps remove occluders for occluded targets, making it more effective for tracking tasks in crowded occlusion scenarios. To verify the effectiveness of the proposed algorithm, we implemented improvements based on the HybridSORT tracker and conducted experiments on the public datasets MOT17 and MOT20. The experimental results show that on the MOT20 dataset, the overall tracking performance metrics—HOTA, tracking accuracy (MOTA), and tracking stability (IDF1)—reached 64.8 (+0.5), 76.1 (+0.6), and 79.6 (+1.6), respectively. On the MOT17 dataset, the three key reference metrics, HOTA, MOTA, and IDF1, also improved to 63.5, 78.7, and 78.9, respectively. Our algorithm provides an effective solution for tracking targets in crowded occlusion scenarios.
Downloads
References
[1] Vandenhende S, Georgoulis S, Van Gansbeke W, et al. Multi-task learning for dense prediction tasks: A survey[J]. IEEE transactions on pattern analysis and machine intelligence, 2021, 44(7): 3614-3633.
[2] Bewley A, Ge Z, Ott L, et al. Simple online and realtime tracking[C]//2016 IEEE international conference on image processing (ICIP). IEEE, 2016: 3464-3468.
[3] Ge Z, Liu S, Wang F, et al. Yolox: Exceeding yolo series in 2021[J]. arXiv preprint arXiv:2107.08430, 2021.
[4] Wojke N, Bewley A, Paulus D. Simple online and realtime tracking with a deep association metric[C]//2017 IEEE international conference on image processing (ICIP). IEEE, 2017: 3645-3649.
[5] Aharon N, Orfaig R, Bobrovsky B Z. BoT-SORT: Robust associations multi-pedestrian tracking[J]. arXiv preprint arXiv:2206.14651, 2022.
[6] Chen L, Ai H, Zhuang Z, et al. Real-time multiple people tracking with deeply learned candidate selection and person re-identification[C]//2018 IEEE international conference on multimedia and expo (ICME). IEEE, 2018: 1-6.
[7] Cao J, Pang J, Weng X, et al. Observation-centric sort: Rethinking sort for robust multi-object tracking[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 9686-9696.
[8] Zhang Y, Sun P, Jiang Y, et al. Bytetrack: Multi-object tracking by associating every detection box[C]//European conference on computer vision. Cham: Springer Nature Switzerland, 2022: 1-21.
[9] Yang M, Han G, Yan B, et al. Hybrid-sort: Weak cues matter for online multi-object tracking[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2024, 38(7): 6504-6512.
[10] Du Y, Zhao Z, Song Y, et al. Strongsort: Make deepsort great again[J]. IEEE Transactions on Multimedia, 2023, 25: 8725-8737.
[11] Liu Z, Wang X, Wang C, et al. Sparsetrack: Multi-object tracking by performing scene decomposition based on pseudo-depth[J]. arXiv preprint arXiv:2306.05238, 2023.
[12] Qin Z, Zhou S, Wang L, et al. Motiontrack: Learning robust short-term and long-term motions for multi-object tracking[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 17939-17948.
[13] Zeng F, Dong B, Zhang Y, et al. Motr: End-to-end multiple-object tracking with transformer[C]//European Conference on Computer Vision. Cham: Springer Nature Switzerland, 2022: 659-675.
[14] Quach K G, Nguyen P, Duong C N, et al. Depth perspective-aware multiple object tracking[M]//Engineering Applications of AI and Swarm Intelligence. Singapore: Springer Nature Singapore, 2024: 181-205.
[15] Liu Q, Chen D, Chu Q, et al. Online multi-object tracking with unsupervised re-identification learning and occlusion estimation[J]. Neurocomputing, 2022, 483: 333-347.
[16] Liu Y, Zhang X, Zhang B, et al. Multi-camera vehicle tracking based on occlusion-aware and inter-vehicle information[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022: 3257-3264.
[17] Milan A. MOT16: A benchmark for multi-object tracking[J]. arXiv preprint arXiv:1603.00831, 2016.
[18] Dendorfer P. Mot20: A benchmark for multi object tracking in crowded scenes[J]. arXiv preprint arXiv:2003.09003, 2020.
[19] Luiten J, Osep A, Dendorfer P, et al. Hota: A higher order metric for evaluating multi-object tracking[J]. International journal of computer vision, 2021, 129: 548-578.
[20] Bernardin K, Stiefelhagen R. Evaluating multiple object tracking performance: the clear mot metrics[J]. EURASIP Journal on Image and Video Processing, 2008, 2008: 1-10.
[21] Ristani E, Solera F, Zou R, et al. Performance measures and a data set for multi-target, multi-camera tracking[C]//European conference on computer vision. Cham: Springer International Publishing, 2016: 17-35.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Journal of Computer Science and Artificial Intelligence
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
