Study On Intelligent Extraction Algorithm Of Video Moving Target Trajectory And Webgis Integration

In the context of artificial intelligence,surveying and mapping puts forward higher requir ements for the management and analysis of geospatial information,and people need to be mor e intelligent in perceiving the world.As a part of spatial information,moving target trajectory information is essential for building smart transportation and smart city.It can play an importa nt role in emergency handling,urban public management,traffic accident identification and p ublic behavior detection.As the process of urbanization continues to deepen,video data beco mes an important means of acquiring spatial information.The video can show the scene of the moving target and the most intuitive trajectory of the moving target in real time.Therefore,h ow to effectively manage the video data and obtain the trajectory information of the moving ta rget from the video becomes a problem to be studied.This paper relies on the national key R&D project “Integrated Disaster Reduction Intelli gent Service Research and Demonstration Application Project,2016YFC083100”,based on th e need for the extraction,query,statistics and analysis of the moving target trajectory of the tr affic accident disaster mitigation service in the project,based on convolution The deep learnin g method of neural network extracts moving target trajectories from video.A method based on Faster R-CNN for video moving target trajectory is proposed.With the advantage of WebGIS system in spatial information management,a moving target trajectory extraction system based on WebGIS is developed.Through the integration of WebGIS and moving target trajectory ex traction algorithm,the management,query and analysis of video data in key areas are realized and the trajectory extraction of moving targets in video is realized.It provides system and algo rithm support for the follow-up research of the project,and realizes the system integration of t rajectory extraction and WebGIS video data management.The main research contents and results of this paper:(1)By studying the basic principles of convolutional neural networks and regional convo lutional neural networks and the computational processes and functions of various parts of the network,the principles and training methods of convolutional neural networks and region-bas ed convolutional neural networks are mastered.The relationship and development process of r egion-based convolutional neural networks R-CNN,Fast R-CNN and Faster R-CNN are analyzed.The feasibility of applying Faster R-CNN network to the extraction of moving target traje ctories is demonstrated.(2)A method based on Faster R-CNN for video data moving target trajectory is propose d.Using Faster R-CNN,the principle of fast target location and feature extraction can be achie ved,and the target area is quickly determined.Based on this,the distance-threshold limit meth od is used to perform target association.The method utilizes the feature that the convolutional neural network can quickly extract the feature of the region,and does not require the Faster R-CNN model to have higher target frame regression accuracy.While ensuring the accuracy of the trajectory extraction,the time and data determined by the target region are reduced.Quanti ty;and by uniformly extracting the method associated with the target reference point,the time of the target association is further reduced.The experiment compares this with Camshift-base d region extraction method.The results show that the method greatly improves the accuracy o f target region determination under complex background and complex texture conditions,and makes the trajectory extraction rate more.high.(3)Designed the related WebGIS system and integrated the algorithm proposed in this pa per into the WebGIS system.The requirements of the WebGIS trajectory extraction system fo r database construction are discussed,and the necessity of spatial database construction is expl ained.The applicable scope of different spatial indexes is studied,and the spatial index databa se required by this paper is constructed by using PostGIS database technology.A Node-based WebGIS-based moving target trajectory extraction and utilization system has been developed,which realizes video data management and retrieval,trajectory information extraction,map se rvice,spatial query and operation.