A Novel Pedestrians Counting System Based on Infrared Imaging

摘要

A real-time system for pedestrian counting is proposed, which is based on the sequences of images acquired by a zenithal stationary camera. This system is constructed by an infrared lighting and an improved pedestrian segmentation and recognition algorithm, to count the pedestrians in the monitoring scene. The results show that the disturbances caused by the overlaps, shadows and illumination changes can be reduced, and the average accuracy is close to 90.1% for the bi-directional pedestrian counting.