TIE VOXEL METHOD AND ITS QUALITY FOR 3D AIRBORNE LIDAR BLOCK ADJUSTMENT

摘要

For achieving high-quality LIDARgrammetry and performing surface reconstruction by 3D point clouds acquired at different locations or in different coordinate systems, a novel method was proposed and is called tie voxel method, in which voxel denotes "volume element". The motivation, ideas and concrete formulation of the tie voxel method are to be given. Its potential applications are discussed. On the one hand, the theoretical quality of the tie voxel method is studied by using simulated data points. On the other hand, some tests are also done by using real airborne LIDAR points. One of diverse types of tie voxels is tie cuboid. This paper describes the tie cuboid method. This method can be used for multi-strips airborne LIDAR points registration. It is hoped that the blunder LIDAR points and the systematic errors could and must be detected somehow. The theoretical accuracy and reliability of the tie cuboid method is then studied by using simulated LIDAR points. Test results verify that the tie cuboid method provides good accuracy and reliability for airborne LIDAR point registration, even in the case of improper point distribution on a side plane. The method exploits all implicit geometric conditions such as coplanarity, colinearity and perpendicularity so that it can provide good geometric strength for 3D point cloud registration. Also, the test results using multiple tie cuboids between two neighboring LIDAR strips show that this method can determine an accurate vertical drift parameter Sz with the a posteriori standard deviation ±0.0494m⌒0.25←z, where←z=±0.20m denotes the a priori standard deviation of Z-coordinate of a LIDAR point. Moreover, some tests are also done by using real airborne LIDAR points with a priori standard deviations ±0.15m and ±0.30m in vertical and horizontal direction, respectively. The root mean square distance 0.213m of a LIDAR point to the corresponding plane is achieved. These test results verify the applicability of the proposed tie cuboid method.