Abstract：A novel method for spectral similarity measure, which is called nonlinear spectral angle mapper, is presented. In this method, on one hand, nonlinear transformation and removing high relativity among bands are used through kernel PCA; on the other hand, the feature of both spectral reflection and absorption are combined in transformed space. At last, spectral angle mapper is used to measure spectral similarity. The authors′ experiments show that this method is effective in spectral similarity retrieval.
[1] Kruse F A，Lefkoff A B，Boardman J W et al. Remote Sensing Environment, 1993, 44: 145. [2] WANG Jin-nian, ZHENG Lan-fen et al(王晋年，郑兰芬等). Remote Sensing of Environment(环境遥感)，1996，11(1)：20. [3] Jia Xiuping. IEEE Transcation on Geoscience and Remote Sensing, 1999，37(1)：538. [4] Scholkopf Bernhard. Neural Compute，1998，19(5)：1299. [5] Clark R N，Roush T L. Journal of Geophysical Research, 1994, 89(7): 6329. [6] Clark R N, Swayze G A, Gallagher A J et al. The U. S. Geological Survey, Digital Spectral Library: Version 1: 0.2 to 3.0 microns, U. S. Geological Survey Open File Report 93-592, 1993，1340.
