基于空间约简的高光谱影像端元检测研究

Research on End Element Detection of Hyperspectral Images Based on Spatial Reduction

端元检测是高光谱遥感影像处理的关键技术,为了提高对高光谱影像混合像元的有效处理,提出基于空间约简的端元检测方法。该方法首先将高光谱影像采取分割处理,并设计相应的像元权重计算与空间信息模型,有利于量化计算。然后基于空间信息设计了超球处理,为防止隐性端元的误删除,同时尽量去除其中的无效数据,根据光谱信息与距离计算,分别设计了球心与半径的确定规则。最后利用设计的纯度指数,完成高光谱影像的端元检测。通过实验,分别从端元检测的效率与精度两方面进行对比验证,证明提出的空间约简方法能够有效提高高光谱影像端元检测的速度和精度。

End element detection is the key technology of hyperspectral remote sensing image processing.In order to improve the effective processing of mixed pixels in hyperspectral images,an end element detection method based on spatial reduction is proposed.Firstly,hyperspectral images are segmented,and corresponding pixel weight calculation and spatial information model are designed to facilitate quantitative calculation.Then,hypersphere processing is designed based on spatial information.In order to prevent the error deletion of recessive terminal elements and remove the invalid data as far as possible,the rules for determining the center and radius of sphere are designed according to the calculation of spectral information and distance.Finally,the end element detection of hyperspectral image is completed by using the designed purity index.Experiments show that the proposed spatial reduction method can effectively improve the speed and accuracy of hyperspectral image endmember detection by comparing the efficiency and accuracy of endmember detection.