3D measurement of diamonds based on stereo vision

Abstract

Three dimensional (3D) measurement for diamond has great significance in diamond industry, because the 3D model of diamond can bring a lot of advantages and convenience for size measurement, cut grade evaluation, display and even indexing. On the other hand, the rapid technological development in lenses and CCD sensors magnifies the advantages and potential of stereo vision technology in measurement. Due to special property of diamond, stereo vision is more suitable to be utilized for 3D measurement. However it is new to apply this technology to diamond, and no stereo vision system with such high accuracy has been well developed yet. The main objective of this thesis is to develop a stereo vision system which aims at conducting 3D measurement for small objects, including but not limited to diamond, with high accuracy around 5 μm. Several issues are investigated and adequate methods are proposed to successfully apply the stereo vision technology to diamond. Since there are drawbacks in the popular existing 2D image based calibration techniques, a more reasonable 3D based mechanism for multi-camera calibration is proposed, which searches the optimal parameter values by minimizing the reconstruction errors in 3D space. A precise apparatus is also designed to generate accurate 3D dense points for calibration. Besides, a modified two-stage approach is proposed and generalized to extract the linear edges in both crown and pavilion images of diamond. Moreover, a simple algorithm is developed to solve the mapping relationship between world coordinate systems, thus the separately reconstructed parts can be combined to form the whole 3D model for diamond. A stereo vision system is designed and fabricated for conducting 3D measurement for small objects including diamond. Both computer simulation and real experiments were carried out, and the obtained results verify the methods developed in this thesis. Experiment results show that the accuracy of the developed stereo vision system can be higher than 5 μm, and the performance of 3D measurement for diamond meets the requirements in practice.

Date of Award	2014
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology