面向GF-1数据不同融合方法对扰动地块的提取

以GF-1影像为数据源,采用HPF变换(High-Pass fusion,高通滤波变换)、Ehlers变换(空间滤波变换)、Brovey变换(彩色标准变换)、GS变换(Gram-Schmidt,正交化变换)、PC变换(Principle Components,主成分变换)等5种常用融合算法,根据视觉分析和数理统计分析对融合后的影像进行质量评价,并通过面向对象分类方法对扰动图斑信息进行提取,研究适用于生产建设项目集中区扰动地表信息提取的融合算法。结果表明:基于PC和GS算法的融合结果影像视觉效果好,细节清晰,色调自然,纹理增强明显,较好的融合了多光谱影像的光谱信息和全色影像的空间信息。通过定量分析可知,PC变换最大程度地保持原多光谱影像的光谱特性,GS融合算法在增益效果上具有优势,融合影像信息量最大,纹理最为细致。总体而言,PC变换和GS变换影像融合算法在生产建设项目扰动图斑信息提取上具有很大的优势,在煤炭覆盖区提取正确率均为100%,在裸露地表和在建建筑用地混合区大于80%,较原始多光谱影像分类精度提高了约10%。该研究为推进国产高分遥感数据在生产建设项目水土保持监管中的高效应用奠定数据基础,对提升生产建设项目水土保持监管效率和信息化水平具有非常重要的意义。

Comparison of Fusion Algorithms for GF-1 Data from Extracted of Distribution Information on Production and Construction Projects

Based on the GF-1 satellite images, the study used five different methods including high-pass fusion, Ehlers transformation, Brovey transformation, Gram-Schmidt spectral sharpening, and principle components spectral sharpening to obtain an optimal method for image enhancement of production and construction projects. To access the performance of these fusion methods, the quality of the fused image was evaluated by the visual and statistical analyses. Moreover, the object-oriented classification method was used to examine the accuracy of extracting the disturbance information on production and construction projects in the central area. Results showed that the principle components and gram-schmidt fusion algorithm achieved better fusion of multispectral spectral information and panchromatic image spatial information, and therefore, higher spectral fidelity, richer spatial information and better visual effect were obtained. Quantitative analysis showed that the principle components transform preserved the spectral characteristics of the original multispectral image to the maximum extent. The gram-schmidt fusion algorithm had the advantage of gain, and the fused image had the largest amount of information and the most detailed texture. Overall, the principle components and Gram-Schmidt image fusion algorithms had great advantages in information extraction of disturbance information on production and construction projects. The extraction accuracy in the coal-covering area was 100%, and greater than 80% in the mixed area of bare land and construction land and the classification accuracy was 10% higher than that of the original multispectral image. This research provide a solid foundation for promoting the efficient application of domestic high-resolution remote sensing data in water and soil conservation regulation of production and construction projects, and is of great significance to advance the efficiency and informationization of soil and water conservation supervision on production and construction projects.