.png){kind=icon}
.png){kind=icon}
Return
Multi-Object Tracking with Prediction Association of Motion Behavior
.png){kind=icon}
PRE
AI
L
DOI：10.1587/transfun.2025EAP1089.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The main challenge of multi-object tracking (MOT) is how to maintain a continuous trajectory of each object. Existing tracking methods typically rely on linear prediction models and intersection over union (IOU) values between objects in adjacent frames to predict associations. However, in dense scenes, complex motions and extreme occlusions tend to undermine the reliability of linear predictions and the discriminability of IoU matching. In this paper, we propose a simple yet effective multi-object tracking method that achieves more accurate prediction and association by leveraging the motion behavior information of the objects. We first design a Union Unscented Kalman Filter (UUKF) to capture the non-linear motion and interaction between objects. It is more suited for the non-linear movement patterns of pedestrians and captures the interaction between moving objects, enabling more accurate prediction of their positions. Moreover, to enhance the ability of data association in complex scenes, this paper proposes a bounding box scaling strategy based on orientation consistency. By dynamically adjusting the size of the bounding boxes, the interference caused by highly overlapping objects is reduced. Additionally, orientation consistency is used to achieve more accurate data association, thereby lowering the probability of mismatch. At last, in order to fully validate the performance of the proposed method, we have conducted extensive experiments and ablation studies on MOT16, MOT17, and MOT20, respectively. The experimental results show that it achieves superior performance.
Keywords:
multiple object tracking
key multiple object tracking
motion behavior
Kalman filter
data association
Journal
IF:
0.4
Papers:
182
Citations:
1.3K
Organization
Cited Papers