论文部分内容阅读
对当前国际经典和前沿的6种代表性的端元提取算法进行比较研究,包括SPP-N-FINDR、VCA、SPICE、PCOMMEND、MVSA和MVC-NMF,通过理论和实验两种方式对这些算法进行综合性对比和分析,总结其优势和存在的问题。通过模拟和真实数据实验得出:SPP-N-FINDR算法的抗噪声能力不如其他5种算法;VCA和MVSA的稳定性较好;MVC-NMF和SPICE无需知道端元数目,且能直接得出丰度矩阵,自动化程度较高;PCOMMEND在真实高光谱图像中提取端元的结果最好,能直接得出丰度矩阵,但若端元数量为素数时精度会下降。研究成果将为今后围绕这些算法的相关研究提供必要的理论支持和参考。
The current international classic and cutting-edge six kinds of representative end-point extraction algorithms for comparative study, including SPP-N-FINDR, VCA, SPICE, PCOMMEND, MVSA and MVC- NMF, by both theoretical and experimental methods for these algorithms Comprehensive comparison and analysis, summarizes its advantages and problems. Through simulation and real data experiments, the SPP-N-FINDR algorithm has less anti-noise ability than the other five algorithms; the stability of VCA and MVSA is good; MVC-NMF and SPICE do not need to know the number of end- The abundance matrix is highly automated. PCOMMEND extracts the endmember best in the real hyperspectral image, which gives the direct result of the abundance matrix, but decreases if the number of endmember is prime. The research results will provide the necessary theoretical support and reference for the future research on these algorithms.