Design Three-dimensional Automatic Passenger Counting System

With the development of information technology, the importance of passenger information in public places is becoming increasingly prominent, such as shopping malls, supermarkets, scenic parks, stadiums, stations, airports and so on. Real-time, accurate information of passenger access and flow distribution can improve operational efficiency for managers and improve resource utilization. It can increase the power of coping with the peak passenger flow and dealing with the unexpected large passenger flow. The information can provide managers with the basis of making scientific decision to protect the safe operation of public places. And at the same time, it provides data support for construction and development of intelligent cities by establishing the information database of passenger information. There are several commonly used passenger counting systems on the market and the accuracy of the systems is about85%. Many environmental factors affect the systems adaptability and make them unable to meet the requirements of practical application.In this paper, computer stereo vision that is rapidly developing at present is applied to automatic passenger counting system. The system is a new embedded passenger counting system with the following characteristics:low-cost, real-time processing, integrated. It is named automatic passenger counting system based Xtion sensor. Taking Xtion sensor and ARM processor as platform, the system gets number of the passenger by analyzing the feature information of the video stream which got from the sensor, designing the algorithm of head’s recognition and the algorithm of target tracking based on adaptive threshold, judging preferred direction. System algorithm is designed to identify the head based on the three-dimensional spatial characteristics. It differs from the algorithm that need to establish pattern recognition training library to recognize the head based on color video streams, reducing the complexity of the algorithm. In addition, system algorithm reduces a number of factors that make the overall system performance low, such as illumination changes and mutual occlusion existing in the automatic passenger counting system based on color video streams. What is more, the system is from the effect of light and can work in all black environment because the sensor initiatives to launch near-infrared laser speckle. The system has practical value with96%accuracy and can be used.