Traceability Trajectory Data Extraction Based On Target Detection,Recognition And Coordinate Mapping

With the increasing demand for smart cities and public safety,the number of surveillance camera installations is increasing.Using video surveillance data to track traceability of traceable targets is a new idea in the field of traceability.At the same time,traceability data extracted based on surveillance video can provide more intuitive evidence,and such results are often more easily accepted by users.In video surveillance systems,a large number of cameras,a huge monitoring network,will instantly generate massive amounts of video data.Faced with such a large monitoring scale and the number of videos,the use of artificial video surveillance has been far from meeting the needs of practical applications.In order to extract the traceability data which contains the target motion trajectory from the massive video surveillance data,we need to perform fast and efficient processing on the traceable video.Target detection and target tracking are the main techniques for extracting traceability data.In this paper,the source monitoring video is deeply excavated.Based on the target detection,target recognition,target matching and target tracking algorithms in intelligent video surveillance technology,trajectory information for traceability is efficiently extracted from massive monitoring data,and multi-camera-based target tracking and continuous trajectory generation are also implemented.The main Tasks are described as following:1.From the perspective of traceability,based on the target detection algorithm,an effective frame extraction method combining non parametric background modeling and inter frame difference method is proposed to extract valid frames and sub-pictures with moving targets from surveillance video stream.Then analyzing the sub-picture of each target through target recognition algorithm based on deep learning,the boundary of each sub-picture containing the target is refined,and the information related to the target category and the target location in valid frames are further extracted.Finally,the extraction of effective information in the surveillance video is achieved.2.The data fusion operation is performed on the information related to the traceability target from the two angles of space and time.Find the mapping relationship between two-dimensional image space and three-dimensional real space for multiple cameras in current scene,and the pixel coordinates of the traceability target in twodimensional image space are mapped to three-dimensional real space based on the coordinate mapping equation.Combining class information,location information and image feature information to match the target in adjacent time nodes.Then connect the same target in the adjacent time node,and get the track data of target in the current monitoring scenario to realize data fusion based on coordinate mapping.