Airborne lidar planar roof point extraction using least-squares fitting supervised by a posteriori variance estimation

Abstract

The least-squares fitting method can be used for planar roof point extraction from airborne lidar points; however, it cannot avoid the impact of non-planar roof points (blunders) due to lack of robustness. Therefore, this study has developed a least-squares plane fitting based on a posteriori variance estimation, as proposed by Li in 1983, to reduce the weights of non-planar roof points. Additionally, least absolute deviation (LAD) was integrated into the first step of this improved Li method, to increase blunder detection. For simulated data, the proposed approach increased the blunder detection rate by up to 6% compared to the original Li method. Test results with real data showed that the proposed approach demonstrated robustness, applicability and effectiveness.