1. Motion-guided and occlusion-aware multi-object tracking with hierarchical matching.
- Author
-
Zheng, Yujin, Qi, Hang, Li, Lei, Li, Shan, Huang, Yan, He, Chu, and Wang, Dingwen
- Subjects
- *
SPACE perception , *NETWORK performance , *MOTION , *TRACK & field , *SPATIAL variation , *DETECTORS - Abstract
In the field of multi-target tracking, the widely embraced tracking-by-detection paradigm has rapidly progressed with the refinement of detectors and matching techniques. However, the paradigm of joint detection and tracking is relatively limited, and it is difficult to model complex scenes, such as the complexities introduced by camera motion and occlusion. In this work, a hierarchical joint detection and tracking framework is proposed, namely MSPNet. From a temporal concern, a motion-guided feature aggregation module is proposed to address the complexities of multi-frame variations. From a spatial concern, an occlusion-aware head and hierarchical spatial association are proposed to handle the challenges of occlusion. Extensive experiments on MOT challenging benchmarks demonstrate that the MSPNet can effectively reduce false negatives and improve the accuracy of tracking while outperforming a wide range of existing methods. • An enhanced MOT model through motion-guided spatial perception is proposed, namely MSPNet, which optimizes detection and tracking jointly. • To capture temporal variations and ensure spatial motion consistency of instances, a motion-guided cross-temporal feature aggregation module (MFA) is proposed. • To handle the cases of occlusion, occlusion-aware head (OAH) and spatial hierarchical association (SHA) are proposed. The OAH and SHA improves the tracking performance of the network for challenging samples, specifically occluded targets. • Incorporating the power of MFA, OAH, and SHA, the proposed MSPNet, demonstrates notable improvements compared to previous models. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF