3D point cloud data splicing algorithm based on feature corner model

Abstract

Three-dimensional point cloud data splicing technology has always been a research hotspot and it is also a difficulty in reverse engineering, computer vision, pattern recognition, surface quality detection, and photogrammetry. Taking reverse engineering as an example, three-dimensional digital technology is the first link in reverse engineering. In the actual measurement process, due to the limitation of the geometric shape and measurement method of the measured object, the measurement device needs to perform multiple positioning measurements on the object from different viewing angles. Then, the splicing of multiple views is performed on point cloud data measured from different perspectives. The 3D point cloud splicing technology is also called repositioning, registration, or splicing technology on different occasions, whose essence is to coordinate transformation of data point clouds measured under different coordinate systems. The key to the problem is the determination of coordinate transformation parameters (rotation matrix) and translation vectors. Therefore, in the field of engineering, how to carry out non-contact reverse measurement of large-size parts and complex surfaces, how to digitize parts with high efficiency and high precision, and how to effectively realize the three-dimensional reconstruction of large-size and complex parts based on the digitized results of the part surface are key issues. This paper proposes the novel perspective of dealing with the above-mentioned challenges, and the experiment shows its effectiveness.