Parking-vehicles recognition using spatial temporal data (A study of mobile robot surveillance system using spatial temporal GIS part 2)

Abstract

The unique omni-directional motion stereo method featuring robust epipolar estimation and hybrid use of the feature/area based matching, and the change-region recognition technique which uses dense color-textured depth map and 3D-GIS data for segmentation, are presented. The dense stereo imaging data, which is acquired by the coupled use of an ODV (omni-directional vision) and a GPS/INS (Inertial Navigation Systems) by the motion stereo method, is classified in "change region" or "registered region" by the D-GIS's geometric model of the building. Hence, the changeable region like a parking-vehicle on the road is modeled as a hexahedron through surface recognition process, the position of vertexes and the image texture of three surfaces are measured, and are additionally registered in the spatial temporal GIS (Geographic Information System) as new object data. The proposed method can be applicable to a mobile robot surveillance system which is used in a site immediately after a disaster.