Accurate Sphere Marker-Based Registration System of 3D Point Cloud Data in Applications of Shipbuilding Blocks

This paper presents a sphere marker-based registration system of 3D point cloud data (PCD) considering the demanding requirements of high accuracy in shipbuilding industry. In our system, we formulate the problem of aligning pairs of point clouds as Euclidean distance squared minimization energy function. Sphere centers of artificial markers are extracted and used for computing rigid transformations among pairs of point clouds. Sphere fitting and outlier removal algorithms are presented. Principal Component Analysis (PCA) can robustly transform coordinates without the need of multiple iterations by the Iterative Closest Point (ICP). Our proposed unsymmetrical triangular configuration of three sphere markers largely improves our time efficiency. In the end, we proposed an innovative quantative marker-based evaluation method and applied it in real applications of shipbuilding blocks to demonstrate the high performance of our registration system by comparing with commercial PCD processing software.