曲波变换的高光谱遥感图像融合方法在土地利用调查中的应用

Image fusion algorithm of the hyper-spectral remote sensing appling to the survey of land utilization based on a curvature-wave transformation

对两个有关土地利用状况调查区域的高光谱和Quickbird数据进行了数百次算法对比实验,最终确定Sym4和Db2是曲波变换中的最佳小波函数,提出两种基于曲波变换的高光谱遥感图像融合方法,即加权法和选择法,获得了更精确的高光谱遥感融合影像。在一个研究区域,通过对高光谱数据的子带分解、脊波变换、一定融合以及它们的逆变换等曲波变换过程,表明融合的高光谱遥感影像能提供许多有利于土地类型划分和影像识别的精确土地信息。在另一个研究区域,除进行曲波变换外,还通过二次多项式计算实现了高光谱与Quickbird数据遥感影像匹配,最终从高光谱与Quickbird数据获得的遥感融合影像经目视及信息熵、相关系数等评价表明,曲波变换融合图像优于小波变换融合图像,与Brovey融合图像相近,但比PCA融合图像稍差。客观和主观评价证实,曲波变换的遥感融合方法能为高光谱遥感图像在土地利用状况调查提供更多更准确的信息。

In this paper, after several-hundred times of comparing calculation of the hyper-spectral and Quickbird data in two land utilization survey areas, The Sym4 and Db2 are determined to be the best wavelet function in the curvature-wave transformation at last. Two hyper-spectral remote-sensing image fusion algorithms, i. e. , weighted method and selected method, are given based on the curvature-wave transformation. And then the more precise fusion images are provided. In one research area, the hyper-spectral remote-sensing fusion image is benefited for the land typing and image discrimination after the images to resolve son, ridge-wave transformation, some fusions, and their inversion transformation of the hyper-spectral data. In the other research area, except for the curvature-wave transformation, the remote-sensing image matching of the hyper-spectral and the Quickbird data is finished by calculation of quadratic equation, and the fusion image of the curvature-wave transformation is better than that of wavelet transformation, and similar to the Brovey fusion image and slightly inferior of PCA fusion image after the appraising by observation, information entropy and relation analysis. It is suggested that the hyper-spectral remote-sensing image fusion algorithm can provide more precise information for the land utilization survey based on the curvature-wave transformation.