地球物理磁数据中蕴含着丰富的数值信息和结构信息综合运用数值信息中的统计特征和结构信息，并结合地质和实际情况是磁异常解释的一条重要途径本文在前人工作的基础上〔１〕，尝试应用句法模式识别和聚类分析的方法，根据磁异常的结构和统计特征对磁异常进行图形化、机器自动地划分，为高精度、快速和可靠的磁测解释提供一种手段本文首先用句法模式识别方法对磁异常进行结构划分，找出基节点，提取基元，然后用“衍生树”〔２〕的方法形成基类；最后用聚类分析方法将各基类按距离相似性规则进行归类，从而达到磁异常分类解释的目的文章在理论分析和正演模拟的基础上，验证了方法的可行性；进一步结合三峡某地考古的实测磁数据证实了方法的实用性；取得了较为明显的效果，为模式识别技术在地球物理综合解释中的应用开辟了一条新路子，奠定了进一步应用研究的基础 Geomagnetic data contains a wealth of numerical information and structural information  The integrated use of statistical information and structural information in numerical information, combined with geological and practical conditions is an important way to explain magnetic anomalies  Based on previous work 〔1〕, try to use the method of syntax pattern recognition and cluster analysis, according to the structure and statistical characteristics of magnetic anomaly graphical abnormalities, the machine automatically divided for high-precision, fast and reliable magnetic test to provide a Means  In this paper, firstly, we use the syntactic pattern recognition method to divide the magnetic anomaly, find out the base nodes, extract the primitives and then form the base class by the method of “Derivation Tree” [2]; finally, use the cluster analysis method to classify each base class According to the rules of similarity of distance, the purpose of classification and interpretation of magnetic anomaly is achieved. The article verifies the feasibility of the method based on theoretical analysis and forward simulation. Combining the measured magnetic data of an archeology in Three Gorges Area, Practicality; made more obvious effect, for the pattern recognition technology in the comprehensive interpretation of geophysics With opened up a new path, laid the basis for further applied research 