| Passenger flow statistics technology is the focus of the development of urban intelligent public transport.With the development of big data,how to accurately obtain passenger flow data has important research value.Through the statistics of the real-time passenger flow between stations and lines,this paper analyzes the "origin-destination" situation of citizens’ travel,and dynamically guides the public transport management department to formulate the operation plan,so as to realize the optimization,scheduling,prediction of the public transport network and the maximization of the utilization of public transport resources.However,due to the special scene environment in the bus,there are many problems such as non rigidity of passengers,interference of light and serious body jitter,so there is no perfect solution.In view of the above research difficulties,this paper uses video image processing technology to count the bus passenger flow.In this experiment,firstly,the captured image sequence is preprocessed.Considering the serious image distortion and body jitter,the local motion estimation block matching algorithm is used to stabilize the image.After selecting the matching criteria and search strategy,the global motion vector is obtained,so as to carry out motion compensation and retain the key information of the video.Secondly,according to the analysis of camera shooting angle in the car,the feature information of the passenger’s head is relatively complete,so the improved Hough transform algorithm is used,and the color probability distribution map is used to iteratively get the position of the passenger’s head,so as to detect the passenger and reduce the influence of occlusion and light interference.Thirdly,in view of the serious occlusion and the deformation caused by passenger movement,in the passenger tracking link,the "CAMSHIFT TLD" algorithm is used,and the TLD detection module is constantly initialized and the learning module is constantly updated,so that the target template is updated to track the passengers stably.Finally,the number of people on and off the bus is counted by setting the counting line,so as to achieve the purpose of passenger flow statistics.In this paper,through the actual bus scene video,we analyzed the improvement of the selected algorithm on video image stabilization,passenger detection and passenger tracking performance,and counted the accuracy of bus passenger flow statistics in different environments.The experimental results show that the selected algorithm achieves the video de jitter effect,improves the accurate detection and stable tracking of passengers,and maintains the statistical accuracy of the number of passengers in the bus above 80%,which can meet the actual needs of bus route planning,scheduling and scheduling. |
