Lane-level orthophoto map generation using multiple onboard cameras

Map application has played a more and more important role in today's autonomous drive and navigation techniques. However, most of the existing maps can only reach road-level (meter scale) accuracy, which cannot meet the needs of driving with no hands. Based on this situation, a method to generate lane-level (centimeter scale) orthophoto maps is proposed. This method utilizes a multiple onboard cameras capture scheme to capture images of the road surfaces that the vehicle have passed through. After that, camera calibration and joint calibration are done to integrate all these road surface images into orthophoto images. To get lane-level orthophoto maps, the orthophoto images need to be projected to the given positions. The Mercator projection is employed to project GPS coordinates to pixel coordinates. The whole lane-level orthophoto map is divided into several local segments. In each local segment, all the orthophoto images belonging to this local segment are projected according to their pixel coordinates. Finally, for the convenience of map management and release, all the map segments are put together and displayed in the layered map form. Experimental results show that the proposed method can generate a promising lane-level orthophoto map, which verifies the effectiveness of such an approach.