Study Of Key Techniques On Image Processing Of Double Optical Fiber Spherical Coupling Sensor

A new measurement method based on principle of aiming and triggering by double optical fiber spherical coupling is proposed, in which the light is transmitted reversely between different optical fiber through coupler, the transverse displacement of the fiber probe is transformed into the displacement of laser, which transformed into a larger displacement of a CCD by lens. The relative displacement between the transmitter moments of triggering signal of the probe is measured by the length measuring device, and then the measurement of micro-cavity dimension is accomplished.Take the measurement of micro hole in diameter as example, for the purpose of acquiring the accurate sighting information, the algorithms including image filtering, image segmentation, edge detection in subpixel and location of light spot center are used to process the image signal that is transformed by light signal of double optical fiber spherical coupling sensor, thereby the transmitter moment of triggering signal is ascertained accurately and quickly, finally the measurement accuracy of micro hole in diameter is improved.This paper starts from the energy distribution of the optical fiber end, analyze the particular characteristics of image. According to the features of image and distribution of noise, a method combining denoising by adding multiple image and adaptive median filter is used to process the image. Aiming at the non-uniform energy distribution of image, a segmentation algorithm based on gray stretching and morphology is proposed, which is the foundation of the later analysis. In order to improve the edge location accuracy, and then higher center location accuracy of light spot center could be obtained, a new subpixel edge location method based on High-order Orthogonal Jacobi-Fourier moment(OJFM's) and error compensation is proposed. The detailed information of the edge is fully extracted based on the rotation invariance and special features of OJFM's in edge detection. Subpixel location of the edge is obtained according to the relationships among different orders of the OJFM's in vertical directions. Compensation of errors caused by templates with different sizes and the actual edge model are also provided to achieve higher edge location accuracies. In order to verify the effectiveness of OJFM's, Direct Least Square Circle Fitting is combined to contrastively analyze the subpixel edge location accuracy of Spatial moments, Zernike moments and Orthogonal Fourier-Meilin moments. Grubbs'Criterion is used to estimate and eliminate the Gross Error. Experimental results show that the spot center location accuracy obtained by fitting subpixel edge of OJFM's is higher and up to 0.13 pixels compared to Spatial moments, Zernike moments and Orthogonal Fourier-Meilin moments. A filtering algorithm based on Two-sided approximation is used to process the spot center obtained by fitting edge of OJFM's, the uncertainty of the repeated locations is improved from 0.13 pixels to 0.04 pixels, which realize the high location accuracy of spot center and finally improve the measurement precision of the micro hole in diameter.