Abstract: Multiple solutions are inevitable in the interpretation of a single geophysical prospecting method, especially in the complex geological conditions. Traditionally, the data of different methods of the same survey line are interpreted separately. And then based on the interpretation result, we conduct comprehensive analysis and mutual verification, which is a simple combination analysis method. Although the data characteristics of different methods are take into account, the deeper characteristics of the data are not mined from the data level, and the interpretation results are merely multiple data profiles with poor visualization. To solve this problem, this paper proposes a multi-method engineering geophysical data fusion method based on Sparse Auto Encoders (SAE). SAE is a kind of deep network algorithm, which automatically mines the deep features contained in the data through continuous learning. The fused data integrates the physical parameter features contained in various geophysical data, fully excavates geological information from the data, effectively reduces the ambiguity of interpretation, and provides more visualized display, which reflects the characteristics of geological anomalies more comprehensively.

Key words: sparse auto encoders, normalization processing, data fusion, comprehensive interpretation

摘要： 单一物探方法在解释时不可避免地存在多解性,尤其是在复杂地质条件区。通常对同一测线不同方法的数据分别解释,再基于解释成果,综合分析,相互佐证,是一种简单的组合分析法。虽然考虑了不同方法的数据特征,但未能从数据层级挖掘其中更深层次的特征,解释成果是多个数据剖面,显示也不直观。为此,文中提出一种基于稀疏自编码器(Sparse Auto Encoders,SAE)的多方法工程物探数据融合方法。SAE是一种深度网络算法,通过不断学习,自动挖掘蕴含在数据中的深层次特征。融合数据兼备了多种物探数据中蕴含的物性参数特征,充分挖掘了数据中的地质信息,有效降低了解释的多解性,并能做到更直观地显示,可以更加全面地反映地质异常体的特征。

关键词: 稀疏自编码器, 归一化处理, 数据融合, 综合解释