A Point Cloud Simplification Algorithm for Mechanical Part Inspection

A point cloud data set, a set of massive and dense coordinate data points sampled from the surface of a physical object, is emerging as a new representation format of 3D shapes. This is mostly attributed to recent advances in the range finding technology of high-speed 3D laser scanning. A typical laser scanned data set often contains millions of data points and this leads to significant computational challenges in processing the point cloud data for practical applications such as the high-speed laser inspection of mechanical parts. To reduce the number of the massive data points to facilitate geometric computation, this paper presents a simplification algorithm for the laser scanned point cloud data from manufactured mechanical parts, whose boundary surfaces include sharp edges. Due to the distinct feature represented by the points located on or near the sharp edges, these points are first identified and retained. The algorithm then repeatedly removes the least important point from the remaining data points until the specified data reduction ratio is reached. Quantification of a point’s importance is based on points in its neighborhood and it indicates the point’s contribution to the representation of the local surface geometry. The effectiveness of the proposed algorithm is shown through the simplification results of two practical point cloud data sets.