Research On Video Map And Its Generation Method

With the rapid development of the Internet and smart phones,applications based on positioning and maps have become increasingly infiltrated into people's lives.As a product under the support of computer screen display technology,GIS,computer aided graphics,and multimedia technology,electronic maps have the characteristics of greater interactivity,multi-scale expression,and macro overview.In order to display geographical entities in a more realistic and beautiful form on the map,more and more multimedia elements are integrated into the map space.Video as a real-time multimedia element with high degree of scene restoration has already penetrated into people's lives.As a major category of video,surveillance videos inherent the capability of expressing spatio-temporal.Since the monitoring device's geographical location is rarely changed once it is determined,it provides a stable data basis for the mapping and expression of video and maps.The PTZ camera is an important sort of video surveillance sensor.It can adjust the direction of the main optical axis of the sensor and the zoom lens according to the requirements,so as to increase the maximum visual range of the camera sensor.The PTZ camera and the 2-D map are expressed by fusion,and the pose of the sensor in each frame needs to be determined by means of feature detection and homography to complete the splicing of the video data on the map plane under different pose.There are two kinds of fusion schemes for electronic maps and videos.One is based on the mapping relationship between video and geographic coordinates,which makes it spatially dependent and meets the display and playback of video information in a multi-scale environment.The multimedia map is one of the most representative digital map scheme which simply links multimedia to a geographic entity.Another scheme uses the mutual mapping model of video and geographic scenes to enhance the spatial expression capability of video data,so that the video can be based on a projection coordinate system to achieve spatial measurement.The latter method is proposed for a shorter time,with lack of a complete framework and theory,and did not consider the map representation of dynamic targets.In order to strengthen the mutual mapping expression of video maps,this article defines the concept of video maps and discuss the theory and key technologies involved therein.Based on the above background and significance,the concept of video map is proposed in this paper.It focuses on the concept,characteristics,composition,core technology,and key technologies of video maps.It also discusses the separation of foreground and background of video images,PTZ camera pose tracking strategy,and the technology mapping optimization,the content of this article is as follows:1)Present the concept of video maps and its generation method.Defining the concept of video maps,and propose a tilt correction method based on projection plane transformation.The extracted moving target objects are abstracted as map symbols and expressed in different scales in the electronic maps in order to meet the requirements of multi-scale expression and macro view of video maps.2)Changeable-pose sensor tracking and reference frame model.Due to the possibility of pose changes in PTZ cameras and other PTZ surveillance system,an ORB algorithm based on feature points filtering and reference frame model is proposed to track poses.Feature points filtering can reduce feature mismatches generated by foreground objects and improve the accuracy of pose calculations.The reference frame model can be used to correct the error in time when the pose fails to track to improve the accuracy of the pose calculations.3)Background modeling of the video of the fixed-pose and changeable-pose-sensor.This paper proposes an improved Vibe algorithm for fixed-pose video sensor to improve the quality of sample background modeling.Based on the Vibe algorithm,a background pixel generation method is further proposed.For changeable-pose camera,the background modeling is based on the F2F strategy,improved ORB algorithm and Vibe algorithm to improve the quality of the background image and the accuracy of the foreground target mapping,and to strengthen the expression of the changeable-pose camera on the map.4)Generation of Effective sensing plane.Based on the relationship between the number of pixels before and after the projection transformation to crop video images,to reduce the impact of pixel redundancy and the declining of overall quality of the picture caused by pixel interpolation after projection transformation.Using the sensing intensity to generate effective sensing plane,to handle the effective clipping of video images under different pose.