Abstract: The topology relationship of laser trace lines was analyzed. The abnormal points were divided into two categories:dead points and noise points. The environmental factors that affect the dead point error in laser detection were analyzed. The 3D point cloud data were obtained as XYZ format data files， and the data amendment algorithms for dead points were proposed based on 4point interpolation. Moreover， the noise filtering algorithms for the minimal angle and stringheight ratio of noise points based on G2continuity were proposed. The validation shows that the volume， the exposed roof area and roof height are consistent with the reality after reconstruction. For example，the volume revised about 20~70m3， and the height of roof increased about 1~2m. The algorithms is simple logically and less timeconsuming， thereby providing a new method for dealing with goaf point cloud data.

Key words: complex environment, deep goaf, laser scan, point cloud, data amendment