Bus Passenger Flow Detection System Based On Monocular Vision Research

With the rapidly development of the economic society and the population in cites isincreasing, the computer video-based image processing technology for pedestrian countingsystem bring people’s daily life lots of conveniences. The pedestrian detection, targettracking and flow counting technology of the system have been widely used in entrancesof the stores, stations and roads, especially in the intelligent transportation system of thecity, the system could replace the operator to finish the daily management work anddistinguish the special events and so on. The paper applies the system to the buspassengers counting system, the counting system could formulate the operating planaccording to the real-time counting data and analysis results, and the system not only easesthe traffic congestion for the traveling of the citizens but also reduces the relevantoperators’ workload greatly. Because the bus condition is more complex than otherconditions, the paper puts forward an effective solution to overcome the passengerscounting influence in the bus condition, so we can apply the system to the other simpleconditions better. We can see that the system is obvious social and economic benefits.The paper advanced the relevant solution which could test the passenger flowinformation effectively. First, this paper compares a series of related algorithms, includingbackground subtract method, frame difference method, optical flow method and edgedetection method, the detecting result of edge detection method is more suitable for thebus condition. Then, we use the Hough transformation to extract the head feature with thegray judgment method and concyclic degree of confidence which could get rid of the faketarget curve. In the end, the detected head is tracked by Mean-shift algorithm based onKalman filter, the method could solve the inevitable blocking problem to track the head ofthe passenger continuously in the bus condition. Finally, the counting system gets thepassengers flow according to the tracking information automatically.