Research On The Key Technology Of Pearl Detection System Based On Machine Vision

Traditional detection of pearl relies on manual labor, which affected by subjective factors. Along with progress of product process and technology, pearl automatic detection system takes the place of traditional detection gradually.Theoretical study and experimental analysis are done for pearl automatic detection system based on machine vision and techniques of image processing this paper.Research status and detection of pearl automatic detection system, design of measurement hardware and choices of image preprocessing was also involved,especially the theoretical modeling and simulation about hardware of lighting system.On the side of pearl color classification, several main pearl color were selected then the eigenvalues of color model between RGB and HSV were extracted. The H and S of HSV color model were also analyzed, and then the different color of pearl was transform into HSV color model. We extracted the pearl color HSV as the eigenvalues, assort the pearl color use neural networks. The experiments showed that measurement error of this system was below 8%. Total fluctuation range was satisfied the experiment requirements, which indicated that the detection of pearl color was feasible.Then the detection of pearl size and shape was introduced. First compared image preprocessing with the method mentioned this paper, did the excision for the image processed and extracted the edge contour information. This paper mainly introduced how to get the path length in the polar coordinates and decide if the situation of the pearl edge profile aberrations occurs. The results showed that the deviation was below 0.200 mm, the deviation of shape judgment was under 8%,which satisfied the requirements of range of variation. Scheme for shape and size detection of pearl is feasible.At the last part, the texture classification was researched, using different form and texture, based on the GLCM theory. Then the selection of impact factor for pearl image was talked about and dimension reduction of the characteristicparameters was done using principal component analysis in the part of data analysis.Two principal component was got as testing standards, and the contribution rate was as high as 90%. Then the texture of pearl was analyzed by Mahalanobis distance method to calculate the Mahalanobis distance of different sample pearls. The final result showed that the Mahalanobis distance of same-level pearl was shorter.