矿产资源作为国家重要的物质生产基础，为经济繁荣、社会发展以及国防建设做出了重要贡献。地球化学方法作为一种直接找矿的方法，是矿产预测中关键的一环。通过地球化学方法识别出的异常，往往与矿床有着很强的相关关系，有的甚至直接指示了矿床的存在，因此，研究行之有效的方法提取地球化学异常，对找矿有着重要的意义。 多年的地质研究表明，地球化学异常存在一定的空间分布形态，如果能够通过识别出一些特殊的形态，进而推断地球化学异常，这将成为一种新的找矿思路。此外，地球化学元素组合异常往往由元素之间的相关关系来反映，且元素组合异常往往存在于某些特定地质背景下，如断裂带、某些特定岩石等。基于上述两点思路，本文将从空间形态和元素局部相关性两个角度，对地球化学数据进行分析与处理，具体研究内容如下： （1）从地球化学异常的特殊空间形态角度，提取地球化学异常。通过分析地球化学数据的元素值属性和空间位置，提出了基于空间邻域约束的聚类算法。该算法聚类的结果显示，在出现类似于环状、半环形、矩形等特殊形状的位置，与已知矿点位置相一致。在三片区域内分别进行了实验验证，证明该方法在提取地球化学异常方面的有效性。 （2）从地球化学元素之间的局部相关性角度，提取地球化学组合异常。鉴于已有的相关性计算方法大多基于全局范围内的相关性，忽视局部范围的相关性，本文提出一种基于局部相关系数的地球化学元素组合异常提取方法。该方法定义了一个计算窗口，落在窗口内的数据可计算出一个相关系数，窗口在全局范围内遍历，最终形成相关系数矩阵，可根据系数高低来寻找地球化学元素组合异常。在三片实验选区内分别进行了实验，结果表明，局部相关系数高的区域多出现在断裂带、岩性、水系附近等区域，且能够反映矿点的存在。 （3）融合空间邻域约束聚类和局部相关系数两种算法，提出邻域约束局部相关系数聚类算法，对崤山地区1:50000水系沉积物地球化学数据进行异常提取，结合地质信息和已知矿点信息，对该矿区进行找矿靶区的预测。

As an important material production base of the country, mineral resources have made important contributions to economic prosperity, social development and national defense construction. Geochemical methods, as a direct method of ore prospecting, are a key part of mineral forecasting. The anomalies identified by geochemical methods are often strongly related to mineral deposits, and some directly indicate the existence of mineral deposits. Therefore, it is of great significance to study geochemical anomalies by studying effective methods. Many years of geological studies have shown that geochemical anomalies have a certain spatial distribution pattern. If we can identify some special morphologies and infer geochemical anomalies, this will become a new idea for prospecting. In addition, the combination of geochemical elements is often reflected by the correlations between elements, and the elemental combinations are often present in certain geological contexts, such as fault zones, certain rocks, and so on. Based on the above two ideas, this dissertation will analyze and process geochemical data from two perspectives: spatial shape and elemental local correlation. The specific research content is as follows: (1) Geochemical anomalies were extracted from the perspective of special geomorphic geomorphology. By analyzing the element value attribute and spatial location of geochemical data, a clustering algorithm based on spatial neighborhood constraints is proposed. The clustering results of this algorithm show that in the appearance of special shapes like ring, half-circle, and rectangle, it is consistent with the position of known ore points. Experiments have been carried out in three regions to demonstrate the effectiveness of this method in extracting geochemical anomalies. (2) Geochemical combinatorial anomalies are extracted from the perspective of local correlation between geochemical elements. Since most of the existing correlation calculation methods are based on the correlation within the global scope and ignore the correlation of the local scope, this dissertation proposes a geochemical element combination anomaly extraction method based on local correlation coefficient. This method defines a calculation window. The data that falls within the window can be used to calculate a correlation coefficient. The window traverses in the global scope and finally forms a correlation coefficient matrix. The combination of geochemical element combinations can be found based on the coefficient's height. Experiments were carried out in three experimental constituencies, and the results showed that areas with high local correlation coefficients mostly appeared in fracture zones, lithology, and water bodies, and could reflect the presence of ore deposits. (3) Integrating two methods of spatial neighborhood constrained clustering and local correlation coefficients, a neighbor-constrained local correlation coefficient clustering algorithm was proposed to extract anomalously geophysical data from 1:50 000 river sediments in the Xiaoshan area, combining geological information with known minerals. The prospecting target area of the mine is predicted.