基于HY-1C CZI影像光谱指数重构数据MNF变换的红树林提取

本文基于广西山口国家红树林生态自然保护区的HY-1C卫星的海岸带成像仪(Coastal Zone Imager，CZI)影像，分析了红树林与一般陆地植被的光谱特征及其光谱指数的相关性，采用归一化差值植被指数(Normalized Difference Vegetation Index，NDVI)、归一化差异水分指数(Normalized Difference Water Index，NDWI)、大气阻抗植被指数(Atmospheric Impedance Vegetation Index，ARVI)及利用CZI波段构建的光谱斜率比(CZI Visible Spectrum Slope Ratio，CVSSR)4个指数替代CZI原始波段形成重构数据，基于重构数据的最小噪声分离变换(Minimum Noise Fraction Rotation，MNF)结果分量，建立决策树并实现了红树林信息的自动提取。研究结果表明：结合本文所选光谱指数重构数据及MNF变换方法，能够有效增强CZI影像上红树林与一般陆地植被的光谱差异，基于MNF变换分量建立的决策树可有效提取红树林信息，经与专家解译结果比对，本文方法面积准确率达90%以上；经随机样本点验证，总体检测精度为88%。

Extracting mangrove information using MNF transformation based on HY-1C CZI spectral indices reconstruction data

In this study, we first used the spectral vegetation indices such as normalized difference vegetation index (NDVI), normalized difference water index (NDWI), atmospheric impedance vegetation index (ARPI) and visible spectrum slope ratio of coastal zone imager (CVSSR) to reconstruct the HY-1C coastal zone imager (CZI) image data of the Shankou mangrove national ecosystem nature reserve in Guangxi. And then, the minimum noise fraction rotation (MNF) was used to enhance the spectral difference between mangroves and general terrestrial vegetation on the reconstructed multi-band data set. We established a decision tree based on the MNF components to achieve automatic extracting mangrove information. The results show that the spectral indices reconstruction data and its MNF transformation can effectively enhance the difference between the mangroves and the general terrestrial vegetation on CZI images, the mangrove information can be effectively extracted by our decision tree. Compared with the experts’ interpretation results, the extracted accuracy of area of our method is over 90%. The overall detection accuracy is 88% after verification by random sample points.