Making the Most of Using Depth Reasoning to Label Line Drawings of Engineering Objects

Abstract

Automatic creation of B-rep models of engineering objects from freehand sketches would benefit designers. A subgoal is to take a single line drawing (with hidden lines removed), and from it deduce an initial 3D geometric realisation of the visible part of the object. Junction and line labels, and provisional depth coordinates, are important components of this frontal geometry. Most methods for producing frontal geometry use line labelling, but this takes little or no account of geometry. As a result, the line labels produced can be unreliable. Previously,we proposed an approach which inflates a drawing to produce provisional depth coordinates, and uses these to make deductions about line labels. Even a naïve implementation can outperform previous line labelling methods in certain cases. In this paper, we further enhance this approach. We extend the algorithm to non-isometric-projection drawings, consider improved ways of realising some of the concepts, and also consider how to combine this approach with other labelling techniques to gain the benefits of each. We test our approach using to be drawings of what we consider representative samples of engineering objects; these exemplify difficulties not considered in many previous papers on line labelling. Our results, based on this test set, show that the enhancements result in significant benefits.