Research On Pedestrian Flow Analysis Method Based On Machine Vision In Large Public Scenes

The detection and analysis of pedestrian status in public scenes is an important part of security work.With the development of image processing and computer vision technology,it's feasible to locate,identify and track pedestrian targets by camera images in public scenes.It can quantitatively record the important indicators reflecting the pedestrian status,such as the number of pedestrians,pedestrian trajectory,pedestrian density,etc.in the video surveillance system,and improve the service level of public places in terms of safety,humanization and intelligence.This paper analyzes the pedestrians in the video recording Shenzhen Baoan Airport check-in area,including pedestrian detecting and tracking.On this basis,it analyzes the operating efficiency of the check-in area of Baoan Airport.In the pedestrian detection section,this paper selects the deep learning regression network YOLO v3 to detect the pedestrians,and collects the specific datasets to train the test under the existing network framework of YOLO v3.In the tracking section,a multi-target real time tracking method based on the Deep-SORT algorithm is added to accurately track the detected pedestrian targets,thereby overcoming the shortcomings of YOLO v3 ignoring the association information between the associated frames in the target detection.It alleviates the phenomenon of "drop frame" in target detection,and effectively suppresses target occlusion.This paper evaluates the results of pedestrian detection based on YOLO v3 and multi-target pedestrian tracking based on Deep-SORT through experimental video in the airport complex indoor environment,like the results of recall,accuracy and overlap,which indicates the method used in this paper is feasible and effective.According to the pedestrian pixel position information and trajectory information obtained by pedestrian detection and tracking,the operating efficiency of a check-in area of Baoan Airport is analyzed,which mainly includes passenger flow analysis in different time of a day,estimating the queue length and pedestrian trajectory recovery in the check-in area.The pedestrian detection and tracking method used in this paper is better applied,by providing data support for passenger analysis in Baoan Airport.