Research On Intelligent Video Surveillance Technology For Operation Security

Safety work at the workplace is very important.Traditional security surveillance system based on manual duty is costly,inefficient,and has poor real-time performance.At present,intelligent video surveillance technology,which can achieve target detection and tracking,behavior analysis and so on,is developling rapidly,but it requires high performance of the algorithm.This thesis applies intelligent video analysis technology to the workplace surveillance system,studies the intelligent detection,tracking and behavior analysis of the workplace workers and abnormal alarms.The main tasks are:1)The entire scheme of the intelligent video surveillance system.According to the actual situation of the workplace,this thesis analyzes the overall demand of the intelligent video surveillance system,and analyzes the functions of the system,including video images acquisition,moving target detection and tracking,and worker violation alarms.Furthermore,the entire scheme of intelligent video surveillance system for job security is proposed,including the overall framework and technical route.2)Detection and tracking of multi-moving target.According to the real-time tracking requirements of workplace workers,firstly,the GMM background modeling method is used to detect the image after the median filtering noise reduction.And the foreground image is post-morphologically processed to obtain the complete foreground and mark the moving target.Secondly,in order to improve the tracking accuracy and reduce the amount of calculation,the KCF kernel correlation filter tracking algorithm is used to track the target.Finally,a multi-target tracking scheme is proposed to track multiple moving targets that appear in the surveillance videos.3)Typical violation detection and identification.For the illegal intrusion detection,the system demarcates the warning area in the video image,and detects whether there is illegal intrusion through the location information of the worker.For improper wearing of helmets,head area positive and negative sample library of worker who wear helmets wearing is self-made.The classification model is obtained by HOG+SVM classification training,and then the worker in the video is identified whether wearing a helmet.When the worker has a violation,an alarm is issued.4)System integration and testing.This thesis is based on OpenCV and QT to realize the software design of intelligent video surveillance system.In order to verify the feasibility of the system,this thesis conducts multi-target detection and tracking,illegal intrusion detection and helmet wearing condition identification test in the user interface.Finally,the helmet wearing recognition function is tested and analyzed.The wear recognition rate is 79%,the unworn recognition rate is 84%,and the single frame processing time is 51 ms,which mostly meets the real-time requirements of the system.The research significance of this thesis is as follows: intelligent instead of manual surveillance,increasing surveillance efficiency and reducing cost,is conducive to the safe operation of the workplace.Intelligent surveillance is realized in the workplace,which has reference value for the use of intelligent video surveillance technology in other areas of reference.