论文部分内容阅读
高空间分辨率的高光谱遥感数据不仅能够获取地物近似连续的光谱曲线,还具有丰富的空间信息.传统的基于单像元的光谱匹配方法无法将这两种特征很好地结合.针对该问题,提出将条件随机场(CRF)模型引入光谱匹配方法.CRF模型通过构造像元邻域描述空间信息,解决了基于单像元光谱匹配方法仅考虑光谱信息的不足,实现了聚类过程中光谱和空间信息的融合;然而,传统CRF模型基于欧氏距离和马氏距离等相似性测度,无法适应于高光谱遥感影像的数据特征,因此利用光谱相似性测度改进传统CRF模型的相似性测度准则.实验证明,所提出方法能够有效解决传统光谱匹配方法结果的噪声问题,较好地保留了地物的形状特征,分类精度得到提高.
Hyperspectral remote sensing data with high spatial resolution can not only obtain the approximate continuous spectral curve of the ground objects but also have abundant spatial information.The traditional single-pixel-based spectral matching method can not combine the two characteristics well (CRF) model is introduced into the spectral matching method.CRF model describes the spatial information by constructing neighborhoods of pixels, which solves the shortcomings of only taking spectral information into consideration based on the single-pixel spectral matching method, and realizes the clustering process However, the traditional CRF model can not adapt to the data features of hyperspectral remote sensing images based on similarity measure such as Euclidean distance and Mahalanobis distance. Therefore, the similarity measure of traditional CRF model is improved by spectral similarity measure The experiment proves that the proposed method can effectively solve the noise problem of the traditional spectral matching method, and retains the shape features of the features well and improves the classification accuracy.