Extraction of Surface Primitives from Noisy Large-Scale Point-Clouds

Abstract

Abstract3D shape acquisition of large facilities, such as industrial plants and power plants, has been receiving increasing attention for simulating complicated maintenance and repair tasks. It is widely recognized that model-based planning based on 3D CAD reduces the rework of maintenance tasks to a large extent. The state-of-the-art phase-based laser scanners are very promising to efficiently capture point-clouds of large facilities; because they can produce hundreds of millions point data in several minutes. However, point data captured from the phase-based scanner tend to include large noise components and quite a lot of outliers. In addition, very large memory space is required to generate solid models by processing very large point-clouds. Therefore, it is preferable to develop techniques that can extract surface primitives, such as planes, cylinders, cones, spheres, and tori from very noisy and large-scale point-clouds. This paper introduces a robust streaming smoothing operator for extracting surf...