Generative Point Cloud Registration

Methods for registering point clouds based on a series of 2D images are considered. An overview of existing methods for registering point clouds is given: the iterative closest points (ICP) algorithm, its modification using instead of the Euclidean metric, the Hausdorff metric. To find the corresponding points, it is proposed to use the method of optimization of the normal vector and particle swarm (NVP). The article discusses software for photogrammetry DF Zephyr Free, Meshroom on the AliceVision platform, VisualSFM, and the algorithm proposed by the authors implemented on the MATLAB platform. Approaches to assess the accuracy of the resulting models and visual quality are proposed. Using Virtual Scenes for Comparison of Photogrammetry Software. Using heatmaps. The proposed solutions were tested.