Development of computer aided three-dimensional geometric modeler based on designer's idea in twodimensional

• Main Author: Haron, Habibollah
• Format: Monograph
• Language: English
• Published: Faculty of Computer Science and Information System 2006
• Subjects: ZA4050 Electronic information resources
• Online Access: http://eprints.utm.my/id/eprint/4386/1/74225.pdf; http://eprints.utm.my/id/eprint/4386/

In designing an engineering product during the conceptual stage, freehand sketch plays an important role in conveying engineers ideas. Solutions on the problems of representing, interpreting and reconstructing the sketch have been of research interest for several decades. The aim of this study is to propose a framework of geometric modeler in interpreting a single line drawing sketch and reconstruct the sketch to produce a two dimensional model. The methodology used in this study is based on the framework of a geometric modeler that contains two phases. In the first phase, two algorithms were proposed. The first is enhanced thinning algorithm that is used to thin the sketch while maintaining its topology and basic shape. The second is the conversion algorithm to convert the irregular line drawing into vertex chain code (VCC) and into modified Freeman chain code (FCC). These algorithm are used to produce chain code series that represented a thinned binary image of the sketch. In the second phase, a new computational corner detection algorithm of FCC and neural network classifier for corner detection of VCC are proposed to analyse the chain code series to produce two-dimensional geometry entities of the sketch in which the irregular line drawing is transformed into regular line drawing. This study has proposed better solutions in terms of an enhanced thinning algorithm, a modified Freeman chain code and Vertex chain code algorithm as picture description language, and a new computational algorithm and neural network classifier for corner detection. Performance comparison of the corner detection algorithms shows that neural network classifier is performed better while FCC is easier in representing thinned binary image.