Alternative Method for Segmenting Digitized Data in Reverse Engineering

The study aims to propose an alternative method of segmentation of digitized data. It begins by layering the item's surface. It analyzes the nodes inside a point cloud to detect any consistent shape change. If one detected, then it holds it and looks for the next one which may describe a possible shape change. The points between those two are classified and marked as part of the shape change curvature line. The method remembers the marked points and holds them as central nodes which will later form reference regions. It uses normal vectors behavior methods to detect shape changes along X and Y axis. As any other direction would not be detectable by the bi-dimensional approach it then introduces a morphological parameter capable on its own to fully describe the curvature variation of a given item's surface by means of Gauss's curvature. To evaluate curvature variation, the method proposes that the central node's curvature should be compared with every found limit point. Because of the noise present in any points cloud it establishes a threshold value beyond which points may describe accurately any shape change. This procedure takes place for all analyzed reference regions and collects only those who have a greater value than the threshold one. These considerations may be extrapolated to other types of geometries as well, as it is the case with cylinders or cones.