| Determination of the structure of crystals is a crucial subject for an industrially significant crystalline material. The initial motivation of this research is to identify a given crystal type using 3-D image processing and computer vision techniques. This feasibility study is needed as a first step for future research. Based on the property that crystals in nature are only of seven types, our proposed approach intends to classify the crystal structure by its two unique characteristics for all materials, which are the cell length ratio and cell interfacial angle. In particular, measuring the aspect ratio of cell length and interfacial angle in a 2-D image requires a uniform crystal orientation. This is a limiting factor since maintaining a uniform crystal orientation is very challenging in nature. The technique presented in this thesis aims to measure such geometry information in 3-D space so that such measurement is independent of a given crystal orientation. First, a 3-D reconstruction technique is investigated to acquire complete 3-D models of crystal structures. Second, we show that, based on the 3-D models of the crystals, a pattern recognition system using crystal cell length and cell angle as input features is feasible. We show that the proposed method is dependent on the accuracy of the 3-D reconstruction method used. 3-D reconstruction is a very challenging research area which is still in its infancy and without known generalized solutions. In this thesis, we show that in the case of crystals, since we are dealing with polyhedrals, corners, and edges, accurate 3-D reconstruction is more likely to be feasible than in the case of non-polyhedrals. Experimental results are presented to show the potential of this proposed method. |
