Cross-Camera Multi-Object Tracking based on Person Re-Identification and Spatial-Temporal Constraints

Abstract

Abstract In order to reduce the influence of occlusion on the overall feature representation of tracks and improve the accuracy of track correlation between cameras, this paper proposes a cross camera multi-target tracking method based on person appearance and spatial-temporal constraints: a new cross-camera multi-object tracking framework is constructed. Then the person spatial-temporal probability model is established. Finally, the spatial-temporal probability model and the person appearance similarity are jointly measured and the person trajectory correlation under cross-camera is completed by using data correlation; Comparative experiments on the dataset proved that the method is effective.