A Moving Passengers Recognition Algorithm Based On Color Image

At present, bus real-time forecasting, optimization and scheduling are some main issues in the research field of the intelligent transportation and the main questions of this research field focused on how to accurately capture real-time information about when and where the bus passengers get on and off. Automatic Passenger Counting (APC) technology is one of the highest feasibility ways to solve the problem. However, The APC technology based on infrared sensors or pressure sensor which in the use of foreign buses almost can not apply to the status that lots of passengers get off the bus with crowded and chaotic process, leading to low accuracy of identification and counts.According to environment of bus and the characteristics of the image from back door of bus, we proposed a new moving object recognition algorithm based on color space and dynamic background, which is not sensitive to the shadow and the change of light. Besides, we built a real-time passenger counting experiment system which using the network cameras to capture real-time image as input, general-purpose processor as the carrier and the monitor used to output the results. Also, we realized the proposed algorithm on the PC platform. The numerous experimental results show that the proposed algorithm based on color image can identify target effectively and meet the demand of real-time application.