The development of valid CAD models from point-cloud data

This paper is concerned with effective strategies to build valid representations of free-form features starting with point-cloud data obtained from a digital scanning system. Free-form features pose many unique problems for both design and manufacturing engineers. There is no accepted standard for the measurement and representation of such features, and the problems of rebuilding geometric models from measurement data have not been fully addressed. The major challenge in model construction is the detection of logical boundary curves for the segmentation process, and the establishment of appropriate continuity conditions at these boundaries. The automatic segmentation of point cloud data may result in arbitrary boundary curves bearing little resemblance to the natural topology or functional requirements of the model. It is therefore common practice to have some form of human intervention in the process of boundary curve generation. This may be done at the data capture stage by manually tracing along functional boundaries. Alternatively, it may be done at the modelling stage, by selecting points which lie on the chosen boundary and then creating a spline through these. This paper compares both methods.