Abstract
Sketches, with their flexibility and suggestiveness, are in many ways ideal for expressing emerging design concepts. This can be seen from the fact that the process of representing early designs by free-hand drawings was used as far back as in the early 15th century [1]. On the other hand, CAD systems have become widely accepted as an essential design tool in recent years, not least because they provide a base on which design analysis can be carried out. Efficient transfer of sketches into a CAD representation, therefore, is a powerful addition to the designers' armoury. It has been pointed out by many that a pen-on-paper system is the best tool for sketching. One of the crucial requirements of a computer aided sketching system is its ability to recognise and interpret the elements of sketches. 'Sketch recognition', as it has come to be known, has been widely studied by people working in such fields: as artificial intelligence to human-computer interaction and robotic vision. Despite the continuing efforts to solve the problem of appropriate conceptual design modelling, it is difficult to achieve completely accurate recognition of sketches because usually sketches implicate vague information, and the idiosyncratic expression and understanding differ from each designer.