Research And Realization Of Intelligent Video Surveillance System And Multi-view Target Tracking Function In Production Scene

With the continuous improvement of productivity scale,in recent years,the concept of smart factory has quietly emerged in various manufacturing industries.Intelligent factory aims to upgrade the existing infrastructure by using various modern technologies,and realize automatic operation in production process,supervision and management,administrative office and other aspects,so as to achieve the goals of reducing labor costs,reducing human errors and standardizing enterprise management.At the same time,intelligent factories combine artificial intelligence technology to develop and apply corresponding functions such as intelligent analysis,risk early warning and decision-making assistance,and cooperate with production managers to complete the overall decision-making work.Among them,the intelligent transformation of the existing video surveillance system is the only way to realize the intelligentization of the factory.Nowadays,manufacturing enterprises have covered all important scenes in the production site with the video surveillance of cameras.Perfect video monitoring network plays a significant role in ensuring production safety,but enterprises still use traditional manual monitoring methods in the analysis and management of video information.It is difficult for production managers to screen useful information in massive video data information,which not only increases the labor intensity,but also has some problems such as human oversight.At the same time,the existing video surveillance networks use different models of cameras and different video transmission protocols,which makes it difficult to manage video data uniformly.In this paper,a distributed intelligent video monitoring system is designed and implemented according to the characteristics of production site environment and the deployment of camera monitoring network.At the same time,using machine vision technology on this platform,the multi-angle human target tracking scheme is studied and tested.The main work of this paper is as follows:(1)Single-view target tracking has incomplete acquisition of target three-dimensional space information due to limited viewing area.Therefore,when the target reappears in the monitoring screen after being blocked or disappeared,the target tracking process will be interrupted.Based on the camera imaging model,this paper explores the mapping relationship between pixel coordinate system and 3D coordinate system,and proposes a cross-view target association strategy based on spatial constraints to realize multi-view personnel target tracking.(2)Analyze the system requirements and design the intelligent video surveillance system.Among them,the system integrates Dubbo service and ZooKeeper registration center,decouples each module,improves the compatibility and expansibility of the system,and realizes the distributed design requirements of the system.At the same time,aiming at the high concurrent reading and writing scenario of system business data,this paper explores and compares different data updating strategies,and finally adopts the strategy of updating the database first and then updating the cache to ensure the data consistency of the system.(3)According to the production scene and the system design scheme,the intelligent video surveillance system is deployed and implemented on site.Among them,multi-view target tracking module is its basic module.By calibrating the camera on the spot and applying the cross-view target association strategy,the multi-view target tracking process is realized,which verifies the feasibility of the system platform design and multi-view target tracking algorithm.