A neural network algorithm for simplifying the representation of 3D point clouds

Point cloud registration is a task that aligns two or more different point clouds by estimating the relative geometric transformation between them. It is a well-known problem that plays an important role in many applications such as SLAM, 3D reconstruction, mapping, positioning, and localization. In this paper, we propose a method for thinning point clouds based on an autoencoder. When applied to the problem of two point clouds alignment, the autoencoder allows you to determine a set of points that can be removed from both clouds without significantly reducing the accuracy of the registration. This problem is relevant because the dimension of the clouds significantly affects the speed of the algorithm. Computer simulation results are provided to illustrate the performance of the proposed method. © 2025 SPIE. All rights reserved.