A new method for orthographic video map construction based on video and GIS

Abstract

For the problem that the non-ground plane and dynamic object areas in surveillance videos have large deformation after geographic mapping, a new method for orthographic video map construction (OVMC) based on video and GIS is proposed. Firstly, the segmented video background (SegBG), dynamic object sub-image (SubObjPic) and its coordinates are extracted. Secondly, the homography matrix (H) is solved. An area (R) is set in a 2D map. based on H−1, the orthophoto map of the video background (OrviBG) is constructed. Finally, each SubObjPic are converted to geographic space and overlaid on OrviBG, then the ortho video map is generated. The results show that:(1) The accuracy of the semantic segmentation model for removing non-ground plane is 96.70%; (2) Compared with the traditional method, the amount of orthographic video is reduced by 22.9%; (3) The deformation of non-ground and dynamic object areas is eliminated. This method can promote the deep integration of surveillance video and GIS.