An Efficient Method for Making High-accuracy Road True-Orthoimages by Fusing Vehicle-based Stereo Image and Laser Range Data

摘要

Orthoimages generated from aerial photography nowadays form a fundamental base layer in G1S. Moreover, they serve as a data source for the updating of topographic and thematic map data, while also being in many instances a more useful cartographic product than a line map. A true-orthoimage is usually made using both an image and a DSM (Digital Surface Model) of the same area. But it is very difficult to generate high-resolution and high-accuracy DSM for road ortho-image due to the difficulty of image matching for aerial imagery with a lot of moving vehicle in road surface.For more effectively making road ortho-image, this paper presents an integration method of vehicle-borne laser point cloud and stereo image to automatically generate high-precision true-orthoimage of road surface. First of all, the three-dimensional laser point cloud is used to automatically recognize road surface area and to remove laser points of vehicles on road surface, then to produce high-resolution and high-accuracy true DSM of road surface. Base on generated true DSM, stereo images not single images are used to produce true-orthoimages of road surface by eliminating those image pixels of vehicle not road with a stereo matching based efficient processing. The experiments prove that the method proposed in this paper can generate quickly and efficiently high-accuracy true-orthoimages of road surface, whose relative accuracy is 10mm and absolute accuracy is typically 20 cm with good GPS.