Hyperspectral remote sensing has been widely used for geological mapping and for identifying alteration minerals during mineral exploration. With the development of machine learning, the random forest algorithm has been widely applied to extract alteration mineral information in complex metallogenic settings and across diverse mineral types; however, the quality of training samples often affects recognition accuracy. Therefore, we first quantitatively compared different extraction methods and selected the SID-SGAtan algorithm as the most effective method for extracting reliable and representative training samples from hyperspectral remote sensing imagery. Subsequently, we conducted hyperspectral alteration mineral identification experiments using random forest classifiers to further validate the applicability of remotely sensed alteration information in geological prospecting. The results indicate that: (1) the SID-SGAtan algorithm extracts training samples that are more accurate and reliable than those produced by alternative methods; and (2) the overall accuracy of the random forest classifier based on SID-SGAtan reaches 78.46% (77.09% after dimensionality reduction), with a Kappa coefficient of 0.7053 (0.6863 after dimensionality reduction). These metrics represent an improvement in recognition accuracy over results obtained using SAM and other benchmark methods. The method examined in this study demonstrates strong potential for high-precision mineral identification in hyperspectral remote sensing imagery characterized by rich information content and complex backgrounds.
POURA B, HASHIMM, VAN GENDERENJ. Detection of hydrothermal alteration zones in a tropical region using satellite remote sensing data: Bau goldfield, Sarawak, Malaysia[J]. Ore Geology Reviews, 2013, 54: 181-196.
[14]
BEDINIE. Mineral mapping in the Kap Simpson complex, central East Greenland, using HyMap and ASTER remote sensing data[J]. Advances in Space Research, 2011, 47(1): 60-73.
ZHANGE, ZHANGX, YANGS, et al. Improving hyperspectral image classification using spectral information divergence[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(1): 249-253.
[18]
ROBILAS A. An investigation of spectral metrics in hyperspectral image preprocessing for classification[C]// Algorithms and technologies for multispectral, hyperspectral, and ultraspectral imagery XI. Bellingham: SPIE, 2005: 597-608.
[19]
DUY, CHANGC, RENH, et al. New hyperspectral discrimination measure for spectral characterization[J]. Optical Engineering, 2004, 43(8): 1777-1786.
CHENJ, PISONEROJ, CHENS, et al. Convolutional neural network as a novel classification approach for laser-induced breakdown spectroscopy applications in lithological recognition[J]. Spectrochimica Acta Part B: Atomic Spectroscopy, 2020, 166: 105801.
IORDACHEM, BIOUCAS-DIASJ M, PLAZAA. Collaborative sparse regression for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 52(1): 341-354.
SWAYZEG A, CLARKR N, GOETZA F H, et al. Mapping advanced argillic alteration at Cuprite, Nevada, using imaging spectroscopy[J]. Economic Geology and the Bulletin of the Society of Economic Geologists, 2014, 109(5): 1179-1221.