1. Institute of Space Sciences, Shandong Provincial Key Laboratory of Optical Astronomy & Solar Terrestrial Environment, Shandong University, Weihai, Weihai 264209, China
2. State Key Laboratory of Transient Optics and Photonics, Chinese Academy of Sciences, Xi’an 710000, China
Abstract：Multi-source data fusion is a powerful method to combine data from multiple sources to improve the potential values and interpretation performances of the source data. Multi-payload collaborative analysis is regularly used to detect the same target in planetary exploration. Therefore, it is of great significance and potential application to use spectral fusion to establish a more accurate and robust clustering analysis model for Martian minerals identification. In this paper, the spectral characteristics of the main Martian-related minerals were analyzed by using both visible near-infrared (Vis-NIR) reflectance spectroscopy and Raman spectroscopy. And some data pre-processing methods such as baseline correction, Savitzky-Golay smoothing, standard normal variate (SNV) scaling were used to produce a high-quality representation of the spectral data. Firstly, the information-rich spectral bands with higher signal-to-noise ratio and less overlapping were selected (i. e., Vis-NIR：430～2 430 nm; Raman：130～1 100 cm-1) for the clustering analysis. Secondly, soft independent method of class analogy (SIMCA) and principal component analysis-K-nearest neighbor (PCA-KNN), were respectively built based on selected Vis-NIR, Raman and two kinds of their fusion data(i. e., coaddition fusion and concatenation fusion), respectively. The accuracy of SIMCA model was enhanced from 72.6% (Vis-NIR) and 90.7% (Raman) to 96.3% (coaddition fusion) and 98. 1% (concatenation fusion). The accuracy of PCA-KNN model was improved from 68.9% (Vis-NIR) and 72.9% (Raman) to 80.3% (coaddition fusion) and 92.6% (concatenation fusion), respectively. The results indicate that the fused Raman/Vis-NIR data can improve the classification model’s accuracy of Martian-related minerals which will lay the foundation of quick rock classification for future Mars exploration.
Key words：Visible-near infrared spectroscopy; Raman spectroscopy; Spectral fusion; Soft Independent Method of Class Analogy; Principal Component Analysis-K-Nearest Neighbor
[1] WU Xiao-li, LI Yan-jun, WU Tie-jun(武晓莉, 李艳君, 吴铁军). Chinese Journal of Analytical Chemistry(分析化学), 2007, 35(12): 1716.
[2] LI Zhi-gang, PENG Si-long, YANG Ni, et al(李志刚, 彭思龙, 杨 妮, 等). Chinese Journal of Analytical Chemistry(分析化学), 2016, 44(3): 437.
[3] JIAO Ai-quan(焦爱权). Journal of Food Science and Biotechnology(食品与生物技术学报), 2016, 35(4): 357.
[4] Casale M, Sinelli N, Oliveri P, et al. Talanta, 2010, 80(5): 1832.
[5] Bi Y F, Zhang Y, Yan J W, et al. Plasma Science and Technology, 2015, 17(11): 923.
[6] Wu Z, Wang A, Ling Z. Earth and Planetary Science Letters, 2016, 452: 123.
[7] Wang A, Freeman J J, Jolliff B L. Journal of Raman Spectroscopy, 2015, 46(10): 829.
[8] Ling Z C, Wang A, Jolliff B L. Icarus, 2011, 211(1): 101.
[9] Buzgar N, Buzatu A, Sanislav I V. Analele Stiintifice de Universitatii AI Cuza din Iasi. Sect. 2, Geologie, 2009, 55(1): 5.
[10] Haaland D. Vibrational Spectroscopy, 1991, 1(3): 325.
[11] Miyake M, Minato I, Morikawa H, et al. American Mineralogist, 1978, 63(5-6): 506.
[12] Harri A M, Genzer M, Kemppinen O, et al. Journal of Geophysical Research: Planets, 2014, 119(9): 2132.
[13] Rivera-Valentin E G, Chevrier V F. Icarus, 2015, 253: 156.
[14] Liu Y, Wang A, Freemen J J. Lunar and Planetary Science Conference，2009, 40: 2128.
[15] WANG Wen-xiu, PENG Yan-kun, XU Tian-feng, et al(王文秀, 彭彦昆, 徐田锋, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2016, 36(12): 4001.
[16] Moros J, Laserna J J. Talanta, 2015, 134: 627.
[17] Duca D, Mancini M, Rossini G, et al. Energy, 2016，117: 251.
[18] Zhu X, Suk H, Shen D, et al. NeuroImage, 2014: 91.
[19] Zhu X, Suk H, Shen D, et al. Computer Vision and Pattern Recognition, 2014: 3089.
