Pedestrian Crossing Detection Based On Monitoring Video

As the state pays more and more attention to social security,monitoring personnel will only visually identify abnormal behaviors such as pedestrian crossing and other lines that endanger the safety of social security assets.Both the accuracy and efficiency are not satisfactory.The traditional monitoring model has long been far from to meet the security needs of today's social environment,intelligent video surveillance systems have started to replace the outdated monitoring model and play an increasingly important role.The over-the-line detection-based monitoring system designed in this paper is an intelligent video surveillance system that can automatically identify lines that cross the line or have a tendency to cross lines.The system fully uses computer vision,digital image processing,and other methods to catch real-time monitoring of surveillance cameras.The video sequence is automatically analyzed and dealt with.The foreground detection and extraction are performed first.Then the pedestrians appearing in the surveillance video are marked and tracked accordingly.Finally,the line crossing behavior and the trend detection is performed.At the same time,more pedestrians appear in the video.When the line acts,an automatic alarm is granted.The system can well monitor the physical intrusion behavior of pedestrians,and can make an alarm in advance before criminals commit invading crime,effectively guaranteeing the security of monitoring sites.The system is developed in matlab2017 a software.There are mainly four function modules.Namely,the moving target detection module,the target identification and tracking module,the data processing module and the display module.In the moving target detection module,In this paper,an adaptive threshold strategy based ViBe moving object detection algorithm is proposed,which improves the accuracy of moving object detection module;In the target detection and tracking module,HOG features are used to train SVM support vector machines to identify pedestrians and kernel correlation filtering algorithms to achieve pedestrian tracking;The main function of the data processing module is to determine whether the target has the tendency to enter the warning area or to enter the warning area according to the motion characteristics of the target;The display module of the system is written in GUI of matlab2017 a,which provides a good man-machine interface.The result of the experiment shows that the system can meet the desired goals.