A method for fast and high-precision extraction of spectral confocal signal peak wavelength based on mean-shift

Abstract

The invention provides a fast high-precision peak wavelength extraction method based on the Mean-Shift algorithm. The method comprises the following steps: collecting a spectral confocal signal of a spectral confocal displacement sensor, performing a normalization process, and cutting off the signal after normalization process based on a set light intensity threshold to obtain a point wavelength sequence corresponding to a light intensity sequence; setting an iterative initial wavelength and carrying out calculation to obtain a Mean-Shift vector corresponding to the current peak value of the point wavelength sequence, and moving the current point wavelength sequence along the obtained Mean-Shift vector to obtain a new peak wavelength and a new point wavelength sequence; calculating 2-normof the obtained Mean-Shift vector; and determining whether the 2-norm of the obtained Mean-Shift vector is smaller than a given allowable error. With the provided method, the peak wavelength extraction speed and accuracy can be improved obviously.

Description

technical field [0001] The invention belongs to the field of fast and high-precision spectral confocal signal peak extraction, in particular to a fast and high-precision peak wavelength extraction method based on the Mean-Shift theory. Background technique [0002] The spectral confocal displacement sensor is a displacement sensor based on wavelength displacement modulation. The measurement accuracy can reach submicron, and the working frequency can reach several thousand Hz. It is widely used in medical treatment, optical detection and displacement measurement. In the sensor system, the polychromatic beams are focused on different positions of the optical axis due to the effect of dispersion. The beams focused on the surface to be measured meet the confocal conditions. Gaussian curve of the spectral confocal signal. After the mapping relationship between signal peak wavelength and displacement is established, the displacement of the measured object can be reversed accordin...

AI Technical Summary

Problems solved by technology

[0003] Commonly used peak wavelength extraction algorithms in the prior art include maximum value method, centroid method, and fitting method. The maximum value method directly selects the wavelength value with the largest light intensity as the peak wavelength, which is simple to operate, but is easily affected by noise and has low precision; The centroid method is simple to calculate and fast, but the non-uniformity of the sampling interval will lead to systematic errors and low accuracy

A.K.Ruprecht et al. derived the centroid method compensation formula based on the error transfer formula. This method needs to know the FWHM of the signal. In actual measurement, the calculation of the FWHM of the discrete signal takes a lot of time. Fitting methods such as parabolic fitting and Gaussian fitting Fitting into a parabolic and Gaussian curve, and then finding the peak according to the characteristics of the curve, with high precision, but complex calculation and slow speed

The accuracy of the fitting method proposed by Tan Jiubin has improved compared with the parabolic fitting method and the Gaussian fitting method, but the efficiency has not been greatly improved.

Images