基于移动最小二乘法的微型光谱仪标定方法

Research on Calibration Method of Micro Spectrometer Measurement Based on Moving Least-Squares

摘要:

针对现有微型光谱仪缺少一种统一、可靠的标定方法，提出了一种基于移动最小二乘法（Moving Least Squares, MLS）的微型光谱仪标定方法。首先，微型光谱仪分别获取汞氩灯和氖灯的标准光谱图；然后对含有高频噪声的原始光谱图进行小波去噪，之后通过峰值定位算法找到特征峰并筛选出需要参与拟合的特征峰所对应的像元序号，最后筛选出一定数量的标定点使用MLS进行拟合。选取一定数量的未参与标定的特征峰代入到拟合函数中进行精度验证。实验表明：基于MLS拟合标定后，标定集的误差标准差为0.136 nm，测试集的误差标准差为0.192 nm，高于传统的最小二乘法曲线拟合，该方法实现了快速、准确地对微型光谱仪进行标定，在实际工程应用中具有重要的指导意义。

Abstract:

In view of the lack of a unified and reliable calibration method for existing micro spectrometer, a calibration method for micro spectrometer based on moving least square (MLS) method is proposed. First, the standard spectrograms of mercury-argon lamp and neon lamp were obtained by spectrometer. Then, the original spectral image containing high frequency noise was de-noised by wavelet. After that, the characteristic peak was found through the peak positioning algorithm. The corresponding pixel number of the characteristic peak that needed to participate in the fitting was selected. Finally, a certain number of standard points were selected and fitted by moving least square method. A certain number of characteristic peaks not involved in the calibration were selected and substituted into the fitting function for accuracy verification. The experimental results show that the standard deviation of calibration set is 0.136 nm and the standard deviation of test set is 0.192 nm after calibration based on moving least square method, which is higher than traditional least square method. This method can realize fast and accurate calibration of micro spectrometer, and it has important guiding significance in practical engineering application.