Research On Subway High Density Passenger Flow Detection System

In recent years.China has entered the period of large-scale urban orbital traffic network construction and operation.the capacity of urban orbital traffic and increasing passenger flow has become the main contradiction.till the end of 2017.cities which has opened the subway in China has reached 35.by the end of February 2018,the operation that China put into subway has been a total mileage up to more than 3000 km.In 2015 Beijing subway has formed a three-ring,four-traverse.five-longitudinal,seven-radiation layout.the subway has become the most popular way for people to travel.At present,Beijing orbital traffic can support 12 million to 15 million passenger flow per day.Therefore,it has become a hot subject in society to solve the problem of people's safe travel.With the development of video surveillance equipment.pedestrian detection technology based on video and image has been widely concerned.It has vital social significance and value of research to detect passenger flow through image devices.At present.the commonly-used pedestrian recognition technologies based on video images are based on color images,which detect pedestrians by extracting pedestrians feature information from color images.However.the real world is three-dimensional.The color image often loses space information.thus it is not ideal to rely on the color image for detection and recognition.Especially in the scene such as subway who has complex and changeable background.easily variable brighness aslo high density pedestrian in unit area,it will be more difficult to detect passenger flow.Combining traditional color image and depth image with moving target detection,a passenger flow detection system that can run steadily under high density passenger flow environment is designed and implemented in this paper.Depth image and color image in monitoring area are acquired simultaneously by means of special depth camera,region of interest is obtained by segmenting foreground moving objects based on background differenc method.head and shoulder of the pedestrian within the region of interest are got through setting the threshold.The gradient histogram(HOG)combined with support vector machine(SVM)are used to detect head-shoulder region of pedestrians.Then the tracking algorithm based on the combination of Kalman and MeanShift is used to track the trajectory.The parameter of the passenger flow is extracted by the change position of the pedestrian path in the map.In order to adapt to the installation environment of the subway.this paper designed the system front-end equipment modularly.improvingthe stability of the system.After a long-term field test,the accuracy of the equipment can reach to 98.14%under the impact of the high density passenger flow in subway,and it can ensure stable operation for a long time.