The traditional cavity monitoring measures have many shortcomings, including of obtaining little of data, being difficult to monitor the unmanned cavity, and not calculating the volume of the cavity accurately. The three-dimensional (3D) laser scanner for cavity can scan the cavity to obtain the 3D point cloud data effectively and roundly, but it is trouble to use these point clouds which will exist a lot of noise that is formed by the dusty, moisture and the 3D laser scanner, and the first point cloud and the second point cloud will be misaligned because the ground may have the emergence of deformation. To solve these problems, this paper puts forward the point cloud denoising algorithm based on KD Tree and registration algorithm based on characteristics of point cloud. The experiments show that these algorithms are effective to remove the noise in the point cloud and realize the registration of point cloud, the point cloud will provide the data basic for mine to use in the future.