摘要
端元提取技术是混合像元分解中重要的步骤之一,传统的端元提取方法仅考虑了像元的光谱信息。本文将数学形态学算子扩展到高光谱空间,并应用到端元提取技术中,可以顾及像元的上下文信息。利用AVIRIS高光谱仿真数据对算法进行了实验验证,结果表明本文算法具有较强的抗噪能力和较高的可靠性。在此基础上,结合徐州地区的EO-1 Hyperion高光谱遥感图像,使用本文算法进行了端元提取应用研究,将实验结果与纯净像元指数、顶点成分分析方法做了对比分析和精度评价,证明本文算法是一种可靠的高光谱遥感图像端元提取技术。
Endmember extraction technology is a significant part of Spectral Mixture Analysis, the traditional endmember extraction algorithms only use spectral information. The proposed algorithm based on context could use both spectral and spatial information. AVIRIS simulated data was used to verify this algorithm' s performance, and the results showed that the proposed algorithm has strong antinoise ability and high reliability. On this basis, taking EO-1 Hyperion image of Xuzhou city in 2006 as experimental data, three algorithms (the proposed algorithm, Pixel Purity Index, Vertex Component Analysis) were used for endmember extraction. The results of comparison indicated that the proposed algorithm could be a reliable endmember extraction technology. Finally, abundance maps were obtained based ori Back-Propagation Neural Network.
出处
《测绘科学》
CSCD
北大核心
2012年第2期126-128,32,共4页
Science of Surveying and Mapping
关键词
端元提取
数学形态学
上下文信息
高光谱遥感
endmember extraction
mathematical morphology
context information
hyperspectral remote sensing