Research On Key Technologies Of Mobile Robot Intelligent Monitoring And Emergency Rescue

Video surveillance is a significant component of security and emergency rescue.It is widely used in various scenarios with its visualized presentation and diverse information transmission.The traditional method of video surveillance has three main lacks.One is the lack of intelligent video analysis.During the surveillance process,it is necessary to rely on synchronous manual observation of the surveillance images taken by multiple cameras for a long time,and people can merely depend on the surveillance content to give the subjective judgments,which leads to low efficiency.The second is the lack of flexible monitoring.Most of cameras are fixedly installed,which is prone to the blind spots in monitoring.At the same time,it is very easy to be limited by ambient light.It does not have good monitoring capabilities in more important low-light scenes or dark scenes compared to general scenes.Meanwhile,a large number of layouts or hardware upgrades will result in relatively high costs.The third is the lack of quick respond and efficient linkage operations of emergency rescue to video surveillance.Based on these issues,the main research work of this thesis has the following three points:Aiming at the problem that the existing video surveillance is easily limited by the illumination changes in real situation,the focus is on the intrusion detection of the video surveillance area under low illumination conditions.In this thesis,the brightness evaluation of the video surveillance and the image brightness enhancement method under low illumination conditions is designed to solve the problem of unclear targets under low illumination conditions,and the foreground segmentation algorithm under low illumination conditions is optimized to effectively reduce the interference of the shadow area and improve the accuracy of the contour extraction of invading targets.At the same time,two regional invasion schemes of full screen and regional delineation are proposed.Based on the determination of geometric relations,parameters for evaluating the degree of invasion are introduced,combined with the direction gradient histogram to extract features,to achieve accurate and efficient regional intrusion targets detection.In respect to the problem of incomplete coverage,blind spots and the infeasibility of vast arrangement of fixed cameras because of high cost,this thesis designs a remote control precision patrol detection method based on a mobile robot platform.By using lidar-based map construction and path planning algorithms,remotely controlled ground mobile robots can patrol autonomously and build fast linkage.At the same time,deep learning is used to realize the pedestrian detection of infrared video images in the dark environment to make up the blind spot of the monitoring viewing angle.In order to realize the rapid linkage between video surveillance and emergency rescue,this study builds an intelligent video surveillance and emergency rescue software platform based on the browser and server architecture,and realizes the functions of accident alarm,emergency rescue linkage,and accident information integration.A quick-response linkage mechanism between video surveillance and emergency rescue has been constructed to improve the flexibility of accident handling while promote the efficiency of surveillance.