Conversion of 2D line drawings to 3D models

Abstract

This project aims to investigate the capability, speed, quality and stability of the three geometrical reconstruction systems, implemented based on three different methods, the cubic corner method, the optimization-based method and the hybrid method, which is a combination of the first two methods. Both accurate and inaccurate 2D input line drawings of a range of objects were tested and discussed. The objects range from simple geometrical objects to engineering objects with more complicated geometry. They can be made up of either all cubic corners or all non-cubic corners or a mixture of both. Inaccurate versions of a 2D drawing are created by modifying the shortest edge length connected to a vertex by a maximum percentage change (P) in a random direction. The P values selected in this report are 0.05, 0.1, 0.15, 0.2 and 0.25. In other words, the percentage change in the shortest edge length will be random, with the selected P values being the maximum.

Research has shown that both CCM and OBM each, and hence the hybrid method, has their own type of objects that cannot be recovered. In terms of the speed of these systems, generally, the results have shown that CCM is the fastest in recovering an object, followed by the hybrid method, then the OBM. In addition, both CCM and the hybrid method are generally able to consistently produce good quality results when the accuracy of the input drawing is good, but poor results when otherwise. However, the quality and stability of the OBM system are relatively poor. OBM can sometimes return very poor, or even unacceptable, results for inputs with good accuracy levels, but better results for more inaccurate input drawings.

However, it is important to note that the quality of the results can vary with the quality of the 2D input line drawings and the accuracy of the cubic corners used for CCM and the hybrid method. Also, a 2D line drawing may have more than one valid interpretation as a 3D object. This phenomenon is known as the Necker Cube illusion, which will be explained in this report.