Abstract: Aiming at the problem that the spatial details and spectral features of PCA-Brovey algorithm could not be improved simultaneously in fusing multi-spectral remote sensing images (MS) and panchromatic images (PAN), a remote sensing image fusion method based on KPCA-Brovey was proposed. Firstly, by utilizing the characteristics of non-linear spectral data mining based on Kernel Principal Component Transform (KPCA) algorithm, three principal components including KPC1, KPC2 and KPC3 with the largest amount of information in multispectral images were extracted. Then, three principal components and panchromatic bands were normalized by Brovey algorithm to enrich the spatial and spectral information in the fusion results. Finally, the experimental results with two satellites data, from Landsat5-TM (Wuhan District) and QuickBird (Foshan District), indicated that the proposed method obviously outperformed the PCA-Brovey method in spectral and spatial information extracting. The spectral information extracting rates were increased to 96.6 % (Wuhan District) and 96.1% (Foshan District) respectively. The spectral correlation coefficients were increased to 0.9079（Wuhan District）and 0.8979(Foshan District) respectively.

Key words: environment engineering, image fusion, panchromatic image, multi-spectral image, KPCA, Brovey

摘要： 针对PCA-Brovey算法在融合多光谱遥感影像（MS）和全色影像（PAN）方面存在的空间细节特征和光谱特征不能同时得到较大提高的问题，提出一 种基于KPCA-Brovey的遥感影像融合方法。利用核主成分变换（KPCA）算法的非线性光谱数据挖掘的特性，提取多光谱影像中信息量最大的3个主分量KPC1 、KPC2和KPC3；然后利用Brovey 算法对3个主分量和全色波段进行归一化融合运算，使融合结果中的空间信息和光谱信息更丰富。利用武汉区的 Landsat5-TM和佛山区的QuickBird影像进行实验，与PCA-Brovey方法相比，该方法在空间信息和光谱信息方面有明显提高，其中光谱信息提取率分别提高 到96.6%(武汉区)和96.1%（佛山区），光谱相关系数提高到0.907 9（武汉区）和0.897 9（佛山区）。

关键词: 环境工程, 影像融合, 全色图像, 多光谱图像, KPCA, Brovey