An Automated Grading Method of Pearl Roundness based on Optical Coherence Tomography

Pearl roundness is an important index in the pearl value evaluation and classification. The existing roundness classification schemes are usually based on 2D image or mechanical methods. However, the 2D image ignores the depth or 3D information, which leads to low classification accuracy, and mechanical classification is easy to cause surface damage. An automated grading method of pearl roundness based on optical coherence tomography is proposed. Based on the high-resolution and low-cost 3D imaging of the pearl surface, through noise reduction, binarization, segmentation, 3D reconstruction, surface extraction, and the surface of pearl is fitted as sphere. A new digital rating index named global local roundness fraction is designed for roundness evaluation, based on the length of the three axes and of coordinates of the fitted sphere. Four grades of pearls were automatically classified by our method, and the results showed that the accuracy of the proposed scheme reached 95%, indicating that the proposed method can be used as an effective means of automatic classification of pearl roundness.