A multitarget multi-camera tracking method for terminal scene

This paper focuses on Multi-Target Multi-Camera tracking task in terminal scene. The task presents significant challenges attributable to multiple factors, including substantial occlusion and variations in appearance resulting from diverse camera angles.we propose a practical framework for dealing this task. Splicing of trajectories acquired by single camera under temporal and spatial constraints, which can solve the problem of frequent changes of ID. Define matching area and construct the similarity function matrix, associates all candidate trajectories to solve the problem of crosscamera trajectory matching. This framework is capable of rectifying tracking inaccuracies resulting from occlusions and missed detections within cross-camera trajectory, thereby generating a comprehensive global trajectory for each target in all cameras. Experiments results on the surveillance video from an airport terminal. In comparison to alternative methodologies, the accuracy achieved by proposed method is 90.61%.