Automatic passenger counting based on multi-objects recognition using dynamic images
Dihua Sun, Min Zhao · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2005
Based on image sequences analysis, a new automatic counting method of passengers getting on and off a bus is proposed. In particular, it addresses the problem of counting when bus is crowded and illumination changes. By extracting and tracking passengers' heads appearing in image sequences and their moving direction, passengers boarding and alighting a bus are counted respectively. The multi-objects recognition and tracking is the key task of passenger counting. After foreground is separated from background subtraction, true objects are extracted through analysis on connected binary foreground image and their features are also extracted. Then, block matching algorithm is presented for the tracking multi-objects represented by regions and moving direction is judged according the result of matching. The experiment results show that the proposed method can realize counting passenger getting on and off a bus in high density correctly and improve the accuracy of counting.