基于氧气A波段临近空间大气温度反演的贝叶斯与最小二乘法对比

基于氧气A波段的临边辐射模拟数据进行临近空间大气温度廓线的反演，分析比较了贝叶斯和最小二乘两种不同反演算法的特点.80km以下，信噪比为66～337时:基于贝叶斯理论反演的三条谱线761.59，762.2，764.05nm的反演误差平均值分别为5.52，3.94，4.73K；采用最小二乘法的反演误差平均值分别为10.57，7.04，8.80K.信噪比为6～34时:基于贝叶斯理论反演的三条谱线的反演误差平均值分别为18.27，12.18，18.27K；采用最小二乘法的反演误差平均值分别为103.18，68.79，85.98K.研究结果表明，基于贝叶斯理论的反演方法，利用先验信息对反演结果进行约束和修正，在有噪声的情况下获得了更合理的解，从而提高了反演精度和抗干扰能力.这为星载探测临近空间大气温度的算法研究和开发提供了参考，也为提高光谱仪器信噪比并进而提高温度反演精度提供了理论基础. 

Bayesian and Least Square Method for Temperature Inversion of Adjacent Space Atmosphere Based on Oxygen A-band

This paper shows that the atmospheric temperature profile in near space is inverted based on the simulated data of adjacent radiation in oxygen A-band. Based on the inversion results, the characteristics of two different inversion algorithms, Bayes and least square, are analyzed and compared. Below 80km, the mean inversion errors of the three spectral lines based on Bayes inversion at 761.59, 762.2 and 764.05nm were 5.52, 3.94 and 4.73K, respectively, after adding the noise with a signal-to-noise ratio of 103. The mean inversion errors of the least square inversion were 10.57, 7.04 and 8.80K, respectively. The mean inversion errors of the three spectral lines based on Bayes were 18.27, 12.18 and 18.27K, respectively, after adding the noise with a signal-to-noise ratio of 102. The mean errors of the least square inversion were 103.18, 68.79 and 85.98K, respectively. Research results show that the inversion method is based on Bayes theory, the inversion results to make use of a priori information constraints and correction, in the case of noisy a more reasonable solution is obtained, which improves the inversion precision and anti-interference ability. It lays a solid foundation for the research and development of the algorithm for detecting the adjacent space atmosphere temperature on board and provides theoretical guidance for increasing the signal-to-noise ratio of spectral instruments to improve the inversion accuracy of temperature.