Research On Intelligent Monitoring System Of Staff On-the-job Status Based On Deep Learning

With the continuous development of science and technology,the monitoring industry has gradually changed from the original analog era to the digital era,the clarity of the monitoring video has been greatly improved,and the installation and maintenance has become more convenient.In order to realize public security and real-time supervision,many government agencies and enterprises purchase video surveillance equipment on a large scale.However,most surveillance video information still needs to be analyzed manually,which not only consumes a lot of human resources,but also has low supervision efficiency,and the analysis results are not objective.In recent years,with the development of machine vision and machine learning,the monitoring industry has gradually shifted from the digital era to the era of artificial intelligence.In response to the above problems and development status,this paper designs a set of intelligent monitoring system equipped with deep learning-based target detection and multi-target tracking algorithm for on-job personnel detection,so as to realize intelligent supervision on the working status of on-job personnel and visitors in government agencies and enterprises.Main research contents include:(1)In the aspect of target detection,the working principle and detection performance of target detection algorithms based on one-stage and two-stage models are studied and analyzed,and ILF-YOLOv3(Improve Loss and Feature-YOLOv3)algorithm for on-duty personnel state detection is proposed according to the requirements of target detection accuracy and real-time performance in the intelligent monitoring system.This algorithm is based on the one-stage model of YOLOv3 detection algorithm,the loss function and multiscale feature detection module are improved.The experimental results show that the average precision of the new algorithm model for the detection of the in-service personnel state is improved by 7.9%,and the recall rate is increased by 14%.(2)In the aspect of multi-target tracking,this paper studies and analyzes the multi-target tracking algorithm based on kalman filter,improves the matching measure of the algorithm’s motion state,and uses the fusion measure of the fusion three measurement models as the matching measure between the motion trajectory and the target detection.The experimental results show that the increase of the fusion measurement model not only alleviates the serious situation of IDSwitch situation caused by the use of Mahalanobis distance as the matching degree measurement alone,but also effectively avoids the failure of Mahalanobis distance measurement caused by camera movement.(3)In terms of system implementation,combined with the development idea of "Internet +",the front-end interface,functional modules,and data management of the onthe-job intelligent monitoring system are analyzed and designed respectively.Finally,the technical service of Easy UI + SSM + Mysql is used to implement Development of intelligent monitoring system based on B / S architecture.