Matching method combined hough transform and RANSAC for spectrometer wavelength calibration

Abstract

Wavelength calibration is carried on before the spectrometer being used normally. The usual calibration process is to measure a standard light source of known wavelength, such as the mercury-argon lamp. The spectral lines formed by the standard light source is divided to find the peak value. It's the critical process of calibration to match the found peak values with the wavelengths of the standard source. The common method in use at present all need artificial participation. To make the calibration procedure automatic, this paper proposed a matching method combined Hough transform and RANSAC, which transforms the problem of finding correct pixel-wavelength matching pairs into finding the right linear model. This paper utilized the Hough transform to detect the linear model, and use the RANSAC to find the optimal linear model. The experimental result indicated that the matching method combined Hough transform and RANSAC can solve the matching problem in wavelength calibration efficiently without artificial work. In addition, both the average precision (96.74%) and the average recall (87.05%) exceed that of the conventional artificial matching methods.