LAUSR

Surface classification is an effective way to assess the surface quality of parts. During the last decade, the assessment of parts quality has gradually changed from simple geometries to complex three-dimensional (3D) surfaces. Traditional quality assessment methods rely on identifying key product characteristics of parts, e.g., the profile of surface. However, for point cloud data obtained by high-definition metrology, traditional methods cannot make full use of the data and lose a lot of information. This paper proposes a systematic approach for classifying the quality of 3D surfaces based on point cloud data. Firstly, point clouds of different samples are registered to the same coordinate system by point cloud registration. Secondly, the point cloud is divided into several sub-regions by fuzzy clustering. Finally, a novel parallel classification network method based on deep learning is proposed to directly process point cloud data and classify 3D surfaces. The performance of the proposed method is evaluated through simulation and an actual case study of the combustion chamber surfaces of the engine cylinder heads. The results show that the proposed method can significantly improve the classification accuracy of 3D surfaces based on point cloud data.