摘要
通过研究凸面几何体理论,找出端元在高维空间中的分布特点,将对整幅影像的搜索转变为对影像中DN值最大和最小像元的分析,并将其应用到最大距离法初始端元的提取。考虑到高光谱影像在获取及处理过程中会产生误差,引入了距离阈值概念,计算距离原始端元小于距离阈值像元的平均光谱。实验证明,用平均光谱代替原始端元光谱,显著提高了光谱相似度,并用线性波谱分离,对Cuprite地区的AVIRIS数据进行丰度反演,取得了较好的效果。
By studying the theory of convex geometry,distribu tion characteristics of endmembers in the high-dimensional space is finded out,searching the whole image is changed from to analyzing maximum and minimum of the DN value,and it is applied to the the maximum distance method of the extraction of the initial endmembers.Considering the hyperspectral imag es in the process of acquiring and processing will produce the error,the distance threshold is introduced,and average spec trum of the pixels distance is calculated to original endmember to be less than the distance threshold.Experiments prove that using average spectrum instead of the original endmenber’s spectrum significantly improves the spectral similarity,by ap plying to abundance inverdion of AVIRIS data in Cuprite by Linear Spectral Unmixing,and achieving good result.
出处
《测绘地理信息》
2013年第4期42-44,共3页
Journal of Geomatics
关键词
最大距离法
端元坐标
凸面几何体
混合像元
距离阈值
maximum distance method
coordinate of end member
convex geometry
mixed pixel
distance threshold