航空高光谱的大气校正是进行高光谱定量反演的基础, 但通过空地同步对比分析航空高光谱大气校正的研究较少, 论文主要研究了Hyspex高光谱遥感数据不同的大气校正方法。 在现有的几种大气校正方法的基础上, 提出了一种大气校正的新方法: 首先, 采用果蝇-鲍威尔优化算法反演光谱的性能参数(中心波长和半波高度的偏移量), 对光谱重定标。 然后在光谱重定标的基础上, 采用MODTRAN模型对Hyspex高光谱数据进行大气校正, 得到地表反射率数据。 利用同步采集的五种典型地物的地面实测ASD数据将提出的新方法与现有的几种大气校正方法(快速大气校正、 经验线性法大气校正、 基于6S模型的大气校正、 基于FLAASH模型的大气校正、 基于MODTRAN模型的大气校正)的大气校正结果进行对比分析, 并采用决定系数(R2)和均方根误差(RMSE)来比较各种大气校正方法的精度。 结果表明: 提出的果蝇-鲍威尔优化MODTRAN模型的大气校正结果最好, 决定系数在80%以上, 均方根误差在15%以内; 基于MODTRAN模型、 FLAASH模型、 6S模型的方法的校正结果稍次于本文提出的新方法, 结果比较稳定, 决定系数在70%以上, 均方根误差在20%左右; 快速大气校正与经验线性法的校正结果不稳定。 可以得出结论: 本文提出的果蝇-鲍威尔优化算法有效可行, 可以精确的反演出中心波长和半波高度的偏移量, 其大气校正的精度优于现有的多种大气校正方法。

Atmospheric correction of airborne hyperspectral is the basis of quantitative retrieval of hyperspectral remote sensing. However, the comparison analysis of aerial and field synchronous data was relatively rare, and it was mainly studied in this paper that the different atmospheric correction methods are compared with the fieldwork spectral of Hyspex hyperspectral remote sensing data. Based on the existing several atmospheric correction methods, a novel atmospheric correction method was proposed in this paper: Firstly, we used Fruit fly-Powell optimization algorithm, spectral performance parameters, that is, shift at the center wavelength (σλ) and Full Width of Half Maximum (σFWHM) are retrieved, so the original spectral is recalibrated. We used the spectral of recalibration, and MODerate spectral resolution atmospheric TRANsmittance algorithm (MODTRAN) was applied for atmospheric correction. Ground synchronous measured reflectance data of five types of typical objects was used, and it was then evaluated the accuracy of the method proposed in this paper and other five generally used atmospheric correction methods: QUick Atmospheric Correction (QUAC), Empirical Line Correction (ELC), Second Simulation of the Satellite Signal in the Solar Spectrum(6S) atmospheric correction, Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) correction and MODTRAN atmospheric correction. Determination coefficient (R2) and root mean square error (RMSE) were introduced to evaluate the accuracy of the atmospheric correction results. Accuracy evaluation results showed that the proposed MODTRAN optimized based on Fruit fly-Powell algorithm in this paper was the best, with R2 above 80%, and RMSE within 15%; the results of MODTRAN, FLAASH and 6S atmospheric correction methods were closer to the proposed new method, and also the accuracy of the three atmospheric correction results were relatively stable, with R2 above 70%, RMSE around 20%. Moreover, QUAC and ELC methods were instable. It is concluded that Fruit fly-Powell algorithm is effective and feasible to estimate σλ and σFWHM, and thus the accuracy of the novel atmospheric correction method is better than the existing various atmospheric correction methods.