Research on Sub-Pixel Positioning and Detection Accuracy of Optical Interference Fringes in Methane Detectors

The optical interference methane detector is a portable, high-precision instrument. However, the accuracy and stability of methane ($\text{CH}_{4}$) concentration detection are significantly influenced by stripe positioning accuracy and environmental factors. To enhance detection accuracy, a sub-pixel stripe localization method based on wavelet transform was proposed. Firstly, Gaussian smoothing is applied to the optical interference fringes to eliminate noise. Subsequently, by selecting appropriate wavelet bases and decomposition levels, wavelet coefficients are calculated at different scales and positions. The envelope of the pattern detection optical interference fringes is then extracted based on these wavelet coefficients. Gaussian fitting and the least squares method are used to fit the envelope, and the maximum value of the Gaussian curve is taken as the sub-pixel localization result of the zero-order fringes. Experimental results demonstrate that the envelope localization method using wavelet transform achieves a positioning standard deviation as low as 0.066, which is twice as accurate as direct Gaussian fitting localization. The coefficient of determination ($R^{2}$) for the calibration curve fitting is close to 1. For $\text{CH}_{4}$ concentrations ranging from 0% to 1%, the maximum error of cross-validation is 0.066%, with an average maximum error of 0.039%. Therefore, the zero-order stripe localization method based on wavelet transform exhibits high localization accuracy and meets the detection accuracy requirements of the MT 28-2005 standard.