A data fusion-based method for pedestrian detection and flow statistics across different crowd densities

Abstract

Accurate tracking and statistics analysis of pedestrian flow have wide applications in public scenarios. However, the conventional tracking-by-detection approaches are prone to missing individuals in densely populated or poorly lit environments. This study introduces a pedestrian detection and flow statistics method based on data fusion, which effectively tracks pedestrians across varying crowd densities. The proposed method amalgamates object detection strategies with crowd counting technique to determine the locations of all pedestrians. By observing the coordinates of pedestrians' foot points, this approach assesses the interaction dynamics between the movement trajectories of pedestrians and designated spatial areas, thereby enabling the collection of flow statistics. Experimental results indicate that the proposed method identifies 2.7 times more pedestrians than object detection methods alone and decreases false positives by 58% compared to crowd counting techniques in crowded settings. In conclusion, the proposed method exhibits considerable promise for achieving accurate pedestrian detection and flow analysis.