.png){kind=icon}
Return
A Global Position Estimation Algorithm for Multi-Camera Multi-Object Tracking
.png){kind=icon}
PRE
AI
DOI：10.1109/ACCESS.2026.3651249.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Traditional projection-based multi-camera multi-object tracking approaches typically project the bottom points of detected objects, such as feet, onto the ground plane to estimate global positions. However, frequent occlusions of the lower body hinder accurate projection onto global space. To address this limitation, we propose a global position estimation algorithm for multi-camera multi-object tracking. The proposed algorithm approximates the real-world position of an object by using its top point and height. For each camera, short-term tracklets, composed of the top points of the tracked objects, are projected onto the global space. These projected tracklets are used to estimate the global positions using the global position estimation algorithm. The estimated tracklets are grouped, and the average tracklets are computed from each group to form a corresponding global tracklet. These global tracklets are then matched with existing global tracks based on spatial and directional similarities. Compared to the bottom-point projection method, the proposed approach achieved a 124.5% improvement in MOTA and a 4.9% reduction in MOTP on the experimental dataset.
Keywords:
3D distance
homography
Kalman filter
multiple cameras
object tracking
position estimation
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W
Organization
Cited Papers
.png){kind=icon}
ShareSave
ShareSaveShareSaveShareSave
View more