Accurate contactless measurement of triangular objects composed of 3D point clouds of low quality and unclear rotation

Image processing has become an important tool in various fields of human activity. One of its important applications lies in the determination of the size and dimensions of an object. Thus, in the present paper, we deal with the contactless measurement of triangular components, which was found to have certain limitations once performed on a production line in real time. These are (i) low quality output with unclear rotation in the space of a 3D scanner (missing accurate Cartesian coordinate system rotation data) and (ii) fully autonomous measurement for the selected shape of the component (no manual data tweaking as possible in the laboratory). Besides measurement, the application program contains preprocessing, such as segmentation or removing outliers. But we focus on just the key problem of measurement, i.e., finding the best-fitting hyperplane through three preprocessed point clouds and measuring the sides of the triangle projected onto the hyperplane. The problem is solved by orthogonal linear regression utilizing the Principal Component Analysis approach.