STRTrack: multi-object tracking based on occlusion and trajectory forecasting

Occlusion between different objects is a typical challenge in Multi-Object Tracking (MOT). Especially in cases of severe occlusion and complete occlusion, they have not been effectively resolved before. To address this issue, we propose a method called Spatial Temporal ReTrack (STRTrack) to solve the problem of multiple occlusion. Our method comprehensively solves the problem of multi object tracking under local occlusion, severe occlusion, and complete occlusion conditions for the first time, achieving synchronous improvement of occlusion target detection and tracking performance. Simply put, we realize the synergy between position prediction and spatial relationship, which leads to strong robustness to occlusions. At the same time, we propose a hypothesis box continuous tracking mechanism for completely occluded objects, which fully utilizes object motion trajectory prediction information and combines the spatial position relationship between the object and the surrounding environment to improve the performance of occluded object detection and long-term tracking. The resulting approach achieves high accuracy for both detection and tracking. As a result, we achieve the best performances compared to the current state-of-the-art methods on the popular multi-object tracking benchmarks such as MOT16, MOT17 and MOT20.