基于列相关的热红外高光谱遥感图像非均匀性校正

由于探测单元之间响应不一致、电子增益和偏置发生变化、焦平面污染和损伤等因素，推扫式热红外成像光谱仪获取的图像常常表现为图像列之间不均匀，条带噪声严重，影响了热红外高光谱遥感图像的后续处理和应用。结合推扫式成像光谱仪非均匀性的来源和成因，以相邻地物的相关性为理论基础，提出了适用于热红外高光谱遥感图像的非均匀性校正方法。该方法的步骤是，首先逐波段对原始热红外高光谱遥感图像进行标准矩匹配校正，得到标准矩匹配校正图像；然后，以标准矩匹配校正图像为基础，选择相邻两列像元中相同的地物像元；最后，用两列中相同的地物像元，通过线性回归得到后一列的校正系数，并对其进行校正，顺次遍历一个波段的所有列，完成一个波段图像的非均匀性校正。按照此过程，遍历一幅热红外高光谱遥感图像的所有波段，完成一幅热红外高光谱遥感图像的非均匀性校正。将该方法应用于推扫式热红外光谱成像仪实际获取图像的非均匀性校正中。结果表明，相比矩匹配方法，在保证非均匀性校正效果的情况下，本文方法的各列均值和标准差更符合实际情况。

Non-uniformity Correction of Thermal Infrared Hyperspectral Remote Sensing Image Based on Column Correlation

Due to the factors such as inconsistent response between detection elements, changes in electronic gain and offset, and focal plane contamination and damage, the images obtained by a push-broom thermal infrared imaging spectrometer often exhibit the phenomena of non-uniformity between image columns and severe strip noise. This may affect the subsequent processing and application of thermal infrared hyperspectral remote sensing images. According to the source and cause of non-uniformity of the push-broom imaging spectrometer, a non-uniformity correction method for thermal infrared hyperspectral remote sensing images is proposed on the basis of correlation of adjacent columns. The steps are as follows: Firstly, standard moment matching correction is conducted for the original thermal infrared hyperspectral remote sensing image band by band to obtain a standard moment matching corrected image. Then, on basis of the standard moment matching corrected image, the pixels with same features in the adjacent two pixel columns are selected. Finally, by using the pixels with same feature in the two columns, the correction coefficient of the latter column is obtained though linear regression. The latter column is corrected and all columns in one band are sequentially traversed to complete the non-uniformity correction of one band image. According to this process, the non-uniformity correction of a thermal infrared hyperspectral remote sensing image is completed by traversing all bands of a thermal infrared hyperspectral remote sensing image. The method is applied to the non-uniformity correction of the image obtained by the push-broom thermal infrared spectroscopy imager. The results show that compared with the moment matching method, the average value and standard deviation of each column in the image corrected by the proposed method are more consistent with the actual situation if the non-uniformity correction effectiveness is ensured.