The Organization And Retrieval Of PTZ Camera Surveillance Video Data Based On The Space-time And Content Features

Video Retrieval is widely applied in areas such as security monitoring.Text-Based Video Retrieval only use simple text attribute for the whole period of video retrieval,while Content-Based video retrieval can undertake depth retrieval of video content itself,which is a research hotspot in recent years.At present,the surveillance video retrieval based on content,general use only time-consuming visual content feature matching method,less use of geographical space and time information of the camera itself to optimize the retrieval process,difficult to realize the retrieval efficiency and quality of retrieval at the same time and is often not satisfactory in front of the massive video monitoring data.PTZ(Pan/Tilt/Zoom)Camera which has a fixed space position and computable space domain is widely used in monitoring field.In view of the many difficulties in monitoring video retrieval,this paper according to the space-time characteristics of the PTZ camera surveillance video content,global and local content characteristics,studied the surveillance video data and characteristics of optimization of organization,puts forward the retrieval speed and retrieval quality of give attention to two or more things PTZ camera surveillance video retrieval method.In this paper,the main research results are as follows:(1)The PTZ Camera Surveillance Video hierarchical organization modelAccording to the characteristics of the PTZ camera surveillance video data and retrieve oriented application demand,combined with the key value type quick access mechanism of database,database design based on the key type of surveillance video data hierarchical organization model,the monitoring data is divided into the camera,video,camera,key frames of four levels,analyzes the different hierarchical data type in the key values stored in the database structure,at the same time,according to the PTZ camera parameter information,calculate the space of the camera cover domain,according to the time characteristics of video and video segments based on interval time index of the tree,and in time,space,and set up a connection between content characteristics.(2)Video structured processing and feature extractionResearch Edge detection of the cameras,with the method of variable sliding window of video frame,and using color histogram lens shear point,edge profile gradient method to detect the camera.On the basis of shot segmentation,using an adaptive threshold key frame extraction,with the method of getting multiple Key-Frames represent the main content of the lens.To solve the retrieval target matching of the problems of variable size,will have a more granularity partition key frames and extract the global features and SIFT local feature representing the main content of the Sub-Frame,global features including color,shape,texture,using Gaussian normalization formula to normalization and multiple feature fusion.(3)PTZ camera surveillance video retrieval based on time-space-content featurePropose the space-time-content based retrieval method.first of all,fast and efficient pruning with covering time and space domain,and then using the vector dimension global features of relatively coarse query,key frame filter out the matching degree is low,according to the candidate set for precise query based on the using local character.In Changzhou city public security bureau to provide real video as experimental data source experiment,through the contrast experiment shows that global and covered due to the use of space and time domain features are pruned,reduced the need for time consuming the candidate set of local feature matching,improve the retrieval speed,effectiveness and local feature matching also ensures that the retrieval results.Experimental results show that the proposed based on the characteristics of space-time-content PTZ camera surveillance video data organization and retrieval methods take into account the quality of retrieval efficiency and retrieval and reached the anticipated goal.