高光谱影像端元提取算法研究与分析

Research and Analysis of End Member Extraction Algorithm for Hyperspectral Image

高光谱影像混合像元分解技术将遥感分类问题深入到了亚像元级别,端元提取是混合像元分解中的重要步骤。本文选择了基于体积的N-FINDER算法、基于投影和变换的VCA、OSP算法、基于最优化的MVSA算法,结合SPP算法对数据进行预处理,利用模拟数据与真实数据分别进行实验,对比分析实验结果,总结端元提取算法的优点与缺陷以及各自适应的条件。

The high spectral image mixed pixel decomposition technology has extended the problem of remote sensingclassification to the sub-pixel level, and end member extraction is an important step in mixed pixel decomposition. Thisarticle chose the N - FINDER algorithm based on volume, VCA, OSP algorithm based on projection and transformation,MVSA algorithm based on optimization, combined with SPP algorithm for data preprocessing, the simulation data and realdata were used to conduct experiments, the results were compared and analyzed. The advantages and disadvantages of theendmember extraction algorithms and the suitable conditions are summarized.