为了利用遥感影像进行更加精确的找矿预测，本文选择新疆东天山尾亚地区ASTER数据进行矿化蚀变信息提取方法研究。为了提高信息提取精度，本文提出了结合主成分分析（PCA）、多尺度分割和支持向量机（SVM）的遥感矿化蚀变信息提取方法。首先，分析ASTER数据的特征，选取各矿化蚀变信息的特征波段，对组合波段进行主成分分析，获得主分量图像；然后，对各主分量图像进行多尺度分割，并获得分割之后的均值图像；接着，提取训练样本，利用SVM对训练样本进行训练，采用试验方法求得最优核参数和松弛变量，构造最优SVM模型；最后，运用最优SVM模型完成矿化蚀变信息的提取。进行主成分分析时，铁染蚀变信息选择Band 1、2、3、4组合，Al-OH基团蚀变信息选择Band 1、4、6、7组合，OH和CO₃^2-基团蚀变信息采用Band 1、2、8、9组合。在运行SVM时采用了序列最小优化算法（SMO）进行求解，速度提高了12%。实验结果表明，与波段比值法、主成分分析法及基于光谱角和SVM的方法等3种方法相比，本文方法提取铁染蚀变信息、Al-OH基团蚀变信息及OH和CO₃^2-基团蚀变信息的总体精度可达到87.98%、 90.01%及88.93%，Kappa系数分别为0.8011、0.8134及0.8023，与成矿区带、已知矿点和已有不同地质背景成矿特征相关性较好。

In order to accurately locate the deposit, the ASTER data of Weiya area in eastern Tianshan Mountain of Xinjiang is selected to study the extraction method of mineralization alteration information. To improve the accuracy of ASTER data mineralization alteration information extraction method, a method based on Principal Component Analysis (PCA), multilevel segment method, and Support Vector Machine (SVM) is proposed in this study. First, the special band of alteration information is selected after analyzing the ASTER data, and the principal component image is acquired by PCA. Then, the mean image is obtained after the principal component image is segmented. Subsequently, the training samples are trained by SVM after the training samples are extracted. Moreover, the optimal model is constructed using the optimal kernel parameters and flabby variable obtained by repeated testing. Finally, the optimal model is used to accomplish the extraction of alteration information from ASTER data. The abnormal ferric contamination is extracted using 1, 2, 3, and 4 bands, the alteration anomalies with AL-OH groups are extracted from 1, 4, 6, and 7 bands, and the alteration anomalies with OH, CO₃^2- groups are extracted by 1, 2, 8, and 9 bands. SMO is adopted to improve operation efficiency. Thus, the speed is increased by 12%. A comparison with band ratio method, PCA method, spectral angle mapper and SVM method is conducted. The degree of the abnormal ferric contamination, the alteration anomalies with AL-OH groups, and the alteration anomalies with OH and CO₃^2- groups are 87.98%, 90.01%, and 88.93%, respectively. The corresponding Kappa coefficients are 0.8011, 0.8134, and 0.8023. The extraction results of anomaly information are consistent with metallogenic belt, the known mineralization points, and the mineralization characteristics of different geological conditions.