基于ICA的遥感蚀变信息提取方法的研究  

Extraction of alteration information from remote sensing data based on ICA

作　　者：李明明[1,2] 周可法[1] 孙莉[1] 秦艳芳[1,2] 孙雷刚[1,2] 王桂钢[1,2] 

机构地区：[1]中国科学院新疆生态与地理研究所,新疆乌鲁木齐830011 [2]中国科学院研究生院,北京100049

出　　处：《干旱区地理》2011年第2期309-316,共8页Arid Land Geography

基　　金：国家自然科学基金(40871204;40871028);国家自然科学青年基金项目(40601103)联合资助

摘　　要：在研究比较了ICA(独立成分分析)和PCA(主成分分析)方法的差异后,借鉴PCA提取遥感蚀变信息的应用方法,应用ICA的原理,提出了一种基于ICA的遥感蚀变信息提取方法。以ICA方法替代传统的PCA方法,进行遥感矿物蚀变信息提取实验。并以哈图金矿矿区为例,分别用ICA和PCA方法对ETM+影像提取羟基蚀变信息和铁染蚀变信息进行了对比研究。通过实地考察和地质资料对蚀变结果进行了验证,并进行了蚀变信息的结果评价,结果表明:应用ICA提取的蚀变信息能更好的反映矿区的蚀变状况,可以为找矿提供了更为准确的信息,减少找矿成本。同时也证明了独立成分分析方法的有效性。This paper analyses the algorithm of ICA(Independent Component Analysis) and compares its differences with PCA(Principal Component Analysis) algorithm,then presents a method to extract alteration information of miner from remote sensing data based on ICA.The method which is different from the traditional method used by PCA is used by ICA to extract alteration information.PCA,used widely in geological prospecting industry,is multivariate data processing method based on second order statistics methods on signal.The principal components obtained are unrelated and no duplication or redundant information among the components.So PCA is fully utilizedofinformation extraction in the alteration exception.However,ICA algorithm is multivariate data processing methodbased on high order statistics,which can eliminate secondorder multivariable correlation between the data,also can eliminate the high order correlation amongimage data.So it may be more useful for extracting the alteration exception information.In order to test the ICA method,the paper took Hatu mineralization area in Xinjiang province as anexample,the Hydroxylate anomaly information and Ferric contamination information was extracted from the ETM+ images by ICA and PCA algorithm respectively,and the resultswasvalidatedby combining regional geological information with filed confirmation.Finally,the results extracted by the two algorithms were compared,which indicated the information extracted from the ICA was more suitable to thatresearch area,and the method based on ICA was effective.The accuracy of Hydroxylate anomaly information and Ferric contamination information extracted by ICA increased nearly 21% and 39%.As a result,the method based on ICA provides more exact information to reduce the cost of searching mine.

关 键 词：独立成分分析 主成分分析 蚀变信息 遥感 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]