Development And Application Of System Integrating 3D Real Scene Model And Surveillance Video

In recent years,the 3D real scene modeling technology based on tilt photogrammetry has developed rapidly.When it is applied to the field of video monitoring,it can build a B / S-end Web GIS video monitoring system with low cost,strong scalability,easy maintenance and simple operation.It can also realize some professional functions: for example,integrate the 3D real scene model with the monitoring video,make the monitoring video fit on the surface of the3 D real scene model,and the video scene becomes three dimensional,perspective and measurable,which is convenient for managers to comprehensively and truly understand the monitoring site;Combining the technologies of target detection,image matching,tilt photography and real scene modeling with the monitoring video to extract the target location and other information in the video,which is conducive to the immediate response and control of on-site emergencies.The main research work of this paper is as follows:(1)The basic technology used in system construction is introduced and clarified.It includes web front-end technology,video monitoring technology,cesium 3D Earth,nodejs technology,tilt photography technology,etc.,and expounds the role and advantages of these technologies in system construction.(2)The acquisition and release methods of system model data are studied.This paper expounds the key technology of tilt photography real scene modeling and the production process of the real scene model of the system,and carries out real scene modeling for some areas of Jiangning District and Pukou District of Nanjing;This paper analyzes the advantages of various web 3D data formats,uses the idesktop software of Super Map to convert the real model data that cannot be loaded on the web side into a better S3 M format that can be loaded on the web side,and uses Tomcat server to publish the real model data in S3 M format as a network model service for system call,so as to realize the rapid loading of model data.(3)Three key technologies of system fusion are studied.The advantages and disadvantages of various monomer methods of real scene model are compared in detail,and the monomer of survey area model is realized by using the dynamic monomer method with fast speed,good effect and strong function;The image matching algorithm involved in the projection of surveillance video to the real scene model is studied,and an improved RANSAC-SIFT image matching algorithm is proposed.Compared with the original algorithm,the average detection accuracy is improved by 13%,and the projection of surveillance video to the real scene model is realized;The target detection algorithm involved in the extraction of target location information in surveillance video is studied,and an improved YOLOv3 target detection algorithm is proposed.Compared with the original YOLOv3 algorithm,the average detection accuracy is improved by 4.25%,the detection time is reduced by 0.03 s,and the recall rate is increased by 4.01%.The recognition of small and medium-sized targets in surveillance video is strengthened,and the target location information is successfully extracted.(4)Develop a system integrating 3D real scene model and surveillance video.On the basis of demand analysis,the functions and design of the system are clarified,and the basic functions including browsing and zooming of Web GIS system,as well as the application functions such as measurement,plotting,video projection,video projection clipping,target location information extraction in video are realized;Taking the fourth middle school in Pukou District as an example,the functions of the system are displayed.The results show that each function of the system can be well applied.The system integrating 3D real scene model and surveillance video successfully realizes the 3D display of surveillance video and extracts the location information of the target in the video.