针对国产GF-1 WFV2传感器,提出一种缨帽变换系数的推导方法。利用全国范围内不同地区的18幅GF-1影像和5幅同步的Landsat 8影像,选择了大量覆盖不同地物类型的典型样本并分析光谱变化特征。在此基础上,用3幅Landsat 8同步影像对湿度分量回归,然后再用施密特正交化方法依次推导出两两正交的亮度分量和绿度分量。结果显示,推导的前2个分量能解释缨帽变换总变化量的97%以上,前3个分量能解释总变化量的99%以上。针对不同区域影像的实验结果表明,各分量都具有很好的稳定性,而不同分量之间的散点图也都有相似的分布特征,说明该系数有很好的适用性。与验证用的Landsat 8同步影像对比后发现,推导的缨帽变换结果与Landsat 8相应的分量之间都有很好的相关性和一致性,尽管2种传感器波段设置及光谱响应能力不同,但相关性仍高达0.9。对GF-1 WFV1,WFV3和WFV4传感器影像的实验结果表明,针对GF-1 WFV2传感器提出的缨帽变换系数,也可以很好地应用在WFV3和WFV1传感器上,但不建议在WFV4传感器上使用。

Aiming at GF-1 WFV2 sensor, a method for deriving Tasseled Cap Transformation (TCT) coefficients was proposed in this paper. Based on 18 GF-1 images and 5 synchronous Landsat 8 images of different regions throughout the country, a large number of typical samples covering different ground types were selected, and the spectral characteristics of these samples were then analyzed. On this basis, the wetness component of the TCT was retrieved by the regression using three Landsat 8 and WFV2 synchronous images. The orthogonal brightness and greenness components were then derived with the Gram-Schmidt orthogonalization method. The results showed that the derived first two components can explain more than 97% of total variation, and the first three components can explain more than 99% of the total variation. The experimental results from different regions showed that each component had good stability, and the scatter-plots of different components also exhibited similar distribution patterns. This indicates that the derived coefficients have good applicability. Compared with the validation images, the results of TCT are well correlated with corresponding Landsat 8 components. Although these two sensors have different band settings and different spectral responses, their correlation coefficient is still as high as 0.9. The application results for GF-1 WFV1, WFV3 and WFV4 showed that the coefficients derived from WFV2 sensor can also be used for WFV3 and WFV1 sensors, but they are not recommended for WFV4 sensor.