SLAM (Simultaneous Localization and Mapping) synchronous positioning and map construction is the key technology of intelligent perception of mobile robots. However, most of the existing SLAM methods are implemented in stationary environments, and when there are frequently moving obstacles in the environment, SLAM mapping will produce motion distortion, resulting in the robot being unable to accurately locate and navigate. Meanwhile, there are a large number of redundant 3D data points in the 3D point cloud data obtained by 3D scanning equipment such as LiDAR, and excessive redundant data not only wastes a large amount of storage space, but also affects the real-time performance of various point cloud processing algorithms. To address the above problems, a laser SLAM motion distortion removal method and a curvature-based point cloud data classification simplification framework is proposed in this paper. It optimizes SLAM motion distortion by laser interpolation, simplifying the classification of optimized point cloud data. It can improve the accuracy of SLAM mapping, and also well eliminate the redundant data points in the 3D point cloud data with unclear features, greatly improving the efficiency of computer operation.