Video Spatiotemporal Content Organization And Moving Target Retrieval Method In Surveillance Camera Network

In recent years,with the in-depth development of projects such as "Safe City" and"Smart City",a large number of surveillance cameras have been deployed in the main roads,secondary roads,and corners of roads and road intersections in each city.It is used to monitor moving targets such as pedestrians and vehicles.These surveillance cameras are continuously shot 24 hours a day,generating a large amount of surveillance video data.How to quickly and accurately query the moving targets in the massive monitoring video data has become a research hotspot and challenge.Efficient organization and retrieval of surveillance video data is at the heart of this problem.The surveillance video not only has rich content information but also a large amount of spatio-temporal information.According to the different content of the information,the existing monitoring video organization and retrieval methods are mainly divided into two categories:space-time-based and content-based.Although the retrieval speed and accuracy of surveillance video are high,these methods focus more on spatial retrieval at the video sensor or video segment level,and do not pay attention to the retrieval of targets in the video;content-based retrieval can achieve video The retrieval of the target of interest,but due to the huge amount of video data,difficulty in matching the moving target,etc.,it is not satisfactory in terms of retrieval efficiency and accuracy.Therefore,it is urgent to develop a unified monitoring video organization and retrieval method for space-time content to combine the advantages of the two methods,overcome the shortcomings of the two methods,and improve the practical application ability of the method.Based on the above research background,this thesis takes the surveillance video captured by the completely fixed surveillance camera as the research object,and focuses on the construction of the surveillance video data organization model of the unified time and space content,and proposes the monitoring video moving target extraction and matching method.The theory and method of video moving target retrieval under the constraints of time and space.The main research contents and results of this paper are as follows:(1)Multi-level monitoring video data organization method.This paper analyzes the existing monitoring video organization method,and designs the multi-level video organization structure for the temporal and spatial content retrieval of video moving targets.The multi-level spatial and temporal content organization model of surveillance video is proposed.The attribute description definition of multi-level video organization model is determined.A multi-level space-time content organization method for monitoring video is realized.(2)Multi-feature extraction and matching method for video moving targets.In this paper,based on ViBe algorithm and SSD algorithm,the video moving target localization method is designed.Based on the analysis of different image features with different emphasis on expressing image content,the target matching method based on multi-feature similarity fusion and matching result evaluation are proposed.standard.(3)The method of moving target retrieval under time and space constraints.In this paper,the geographic video coverage retrieval method and the motion information acquisition method of moving target direction and speed under time and space constraints are designed.The space-time constraint effect of the road network on the vehicle target is analyzed.Based on this,the target and the main optical axis of the surveillance camera are taken into account.The bearing relationship method under the road network constraint is proposed to effectively eliminate the invalid matching data,which improves the efficiency and accuracy of retrieval.(4)Monitoring video space-time content organization and retrieval system.Based on the existing research results,this paper designs and implements the "Monitoring Video Time and Space Content Organization and Retrieval System".This paper provides a new organization method for monitoring the spatial and temporal content of video surveillance,enriching and expanding the theory and method of surveillance video retrieval.From the application point of view,it can provide the target of monitoring video interest to the public security field quickly.Accurate retrieval to provide effective support for further analysis and decision making.