Airborne Lidar point cloud data filtering is one of the difficulties and emphases in the study of Lidar data post-processing, and also the primary problem needed to be solved. Least squares adjustment is used to fit block terrain in traditional moving curved fitting filtering method, but this method is sensitive to outliers. Aiming at solving this disadvantage, robust estimation theory is introduced to fit block terrain more reasonably, and self-adaption threshold is set to distinguish between ground and non-ground points automatically. The test data provided by International Society for Photogrammetry and Remote Sensing (ISPRS) are adopted for experiment. Compared with 8 classical filtering methods, experimental results prove that, robust estimation can provide more reasonable fitting curve, which means that the proposed method is practical with reliable filtering results and has strong adaptability to various terrains.