摘要
以珠海市横琴岛为试验区,对CBERS-02星多光谱数据和SPOT5全色数据进行融合处理,融合方法包括HIS变换法、Brovey法、主成分融合法、小波融合法、SFIM法、Gram-Schmidt法、PANSHARP法,并使用定量指标对融合结果进行分析,通过计算各波段均值、信息熵、标准差、相关系数等指标对各融合结果进行分析评价,探讨适合两种数据的融合方法;对CBERS-02和SPOT5数据的PANSHARP法融合数据进行支持向量机分类,总体分类精度为82.55%,Kappa系数为0.7358;试验结果表明,通过融合CBERS-02多光谱数据与SPOT5全色数据,既可以增强空间纹理信息,又可以弥补SPOT5多光谱数据缺少蓝光波段的缺陷,具有丰富的光谱信息,可以在海岸带土地利用调查广泛应用。
In this article, taking H engqin island as experiment region, Normal image fusion algorithms （HIS transform,Brovey transform,PCA transform, wavelet algorithm, SFIM transform and PANSHARP trans- form） are used to fuse the multispectral image of CBERS-02 and panchromatic image of SPOT5. In order to analyze each band of fused images and find the best fusion method by criterion, four quantitative indexes such as the average value, the entropy, the correlation coefficient, the standard difference and the gradient are computed and analyzed. Based on the PANSHARP-fused image, the land use categories are extracted by support vector machine classifier,the total classify precision is 82.55% and Kappa index value is 0. 7358. The results show that by fusing the multispectral image of CBERS-02 and panchromatic image of SPOT5, both spatial and spectral resolutions can be improved,and the fused images are of wide application in vari- ous aspects such as land use monitoring and evaluation.
出处
《遥感技术与应用》
CSCD
北大核心
2009年第1期97-102,I0007,共7页
Remote Sensing Technology and Application
基金
国家"973"项目(2006CB701305)
我国近海海洋综合调查与评价908专项(908-01-WY06)项目
中国科学院知识创新项目(KZCXZ-yw-304)联合资助
