| In recent years,with the rapid development of urbanization in China,the population density of major cities is also increasing.At the same time,convenient and fast urban subway has become an important way for most citizens to travel.In citizens travel peak,it gets very crowded part of the subway station in the crowd,this not only affects people’s travel speed,and also can give the pedestrian public safety brought great hidden danger,therefore,based on the video monitor station pedestrian safety risk prevention and control and the construction of the emergency disposal system has already been paid high attention by local.In this context,this paper aimed to extract the walking path and speed of the individual crowd in the subway station monitoring video and analyze the threshold value of abnormal movement,so as to realize the monitoring and warning of abnormal movement of individual crowd.To be specific,HOG-SVM algorithm is firstly used to extract pedestrian sample features in the scene of subway station,conduct classification training through features,and then mean-shift algorithm is used to track the identified pedestrians,so as to obtain the movement data of pedestrians.In the process of tracking,pedestrian tracking will fail due to the blocking of pedestrians and obstacles.In view of this problem,this paper adopts the method of predicting the coordinate position between the two adjacent frames of the pedestrian and determining the final coordinate position of the pedestrian through the kalman filter algorithm.In the process of calculating the pedestrian speed,the path coordinates of each frame of the pedestrian need to be extracted.However,since each camera will have the deviation of the camera point in the monitoring process,the direct use of two-dimensional pixel coordinates to calculate the pedestrian speed will have a great deviation.In this paper,the space resection equation of single phase film in space photogrammetry is improved to be applicable to the scene of subway station.After establishing the 3d coordinate system,the improved spatial resection equation is used to convert the 2d pixel coordinates of the pedestrian path into 3d coordinates,so as to eliminate the image point deviation caused by the camera’s own defects and calculate the more accurate instantaneous pedestrian velocity.Finally,this paper USES the above technology and algorithm to extract and analyze the pedestrian path and speed in the monitoring video of Wuhan optical valley square subway station.The analysis results show the feasibility of the related techniques and algorithms. |
