ICP-Based Registration Method between Point Cloud Data and CAD Models for Structural Manufacturing Error Analysis

초록

Three-dimensional point cloud data are increasingly used for analyzing manufacturing errors in machined parts and assemblies. Since point cloud data is defined in the coordinate system of the measurement equipment, whereas CAD data is defined in the model coordinate system, the iterative closest point (ICP) algorithm is commonly used to align the two datasets. However, large-scale point clouds containing significant noise, non-uniform point distributions, and overlapping measurements, combined with lightweight models composed of hundreds of thousands of triangles, result in high computational cost and susceptibility to local optima. In this study, we developed an efficient method for registering large-scale point cloud data to the CAD coordinate system. A grid-based representative point sampling strategy and a grid map for an efficient neighboring triangle search are introduced. Approximate registration was first performed from multiple initial positions, considering translations and rotations along the x, y, and z axes, using a few sampled points. Subsequently, precise registration is conducted for candidate positions with low errors, and the final registration was determined by selecting the position with the minimum error.

키워드