An elevation correction method for colored point cloud in building areas

摘要

Generating colored point cloud by the fusion of CCD images and point cloud data can exert both of their superiorities sufficiently, which has been a major method to obtain spatial information of the buildings for building reconstruction, object detection and other applications. Airborne LiDAR and CCD cameras are usually combined on one platform to carry out colored point cloud based on registration. In addition, there is also a new 3D imaging sensor that can acquire point cloud and CCD images with a stable relationship by the mechanism of common optical system, which could generate colored point cloud faster than the former. In the process of fusion, the colored point cloud is possible to absence some building information such as corners and boundaries. Interpolation is an optimistic method to solve the above issue. However, due to the unclear boundaries between building and ground in the point cloud data, the elevation error of the building area is large after interpolation. Therefore, a correction method for the elevation of colored point cloud in building area is proposed in this paper by combining point cloud contour extraction, image region merging and contour regularization. The new method can accurately obtain the edge of the building by the using of stable relationship, thus reducing the elevation interpolation error of the colored point cloud. The effectiveness of the method is validated based on the flight test data of 3D imaging sensor. The accuracy is improved by 33% after elevation correction.