基于空间结构模式匹配的褶皱地貌类型自动识别

褶皱的自动化识别是构造解译、地貌类型划分、三维建模等工作的基础。目前,基于地形形态进行空间统计特征分析的地貌类型自动识别方法,难于满足以“对称重复”空间结构为特征的褶皱构造地貌的识别需要。本文以基于ARG的褶皱场景建模为切入点,利用形式文法定义不同褶皱类型的空间结构模式,形成了一种基于空间结构模式匹配的褶皱构造自动识别方法,进而有效地支持了褶皱地貌的自动化识别。该方法包括4个环节：（1）根据核部地层与横切剖面线提取原则,提取可能存在褶皱的场景条带;（2）基于邻接ARG建模方法对场景条带进行场景空间结构的建模与化简;（3）将化简后的邻接ARG模型表达成句子,并通过判别其能否由模式文法推导而来,实现针对不同褶皱构造模式的匹配识别;（4）基于识别出的褶皱场景进行山、谷的进一步判别,最终确定褶皱构造地貌类型。实验表明,本方法可以较为准确地识别出庐山北部山地区域的褶皱构造,并确定褶皱地貌类型。该方法基本解决了构造地貌类型的自动化识别问题,是对传统地貌形态划分方法的有效补充。

Automatic Recognition of Fold Landform Types Based on Spatial Structure Pattern Matching

The automatic recognition of fold structure is one of the bases of tectonic interpretation, geomorphology classification and three-dimensional geological modeling. At present, most of the automatic recognition methods used for landform classification are based on the regular statistical unit. These methods, although effectively extract the characteristic landform by using image or terrain data, cannot recognize the tectonic landforms which combined the structural feature and topographical feature. As one of the most general tectonic landforms, fold landform has featured a symmetric repetitive spatial structure, which can be used to recognize the fold. To realize the automatic recognition of fold landform types, this research provides a method based on the spatial structure pattern matching. This method focuses on building scene models of fold structures by using Attributed Relational Graph (ARG) and identifying the fold landform types by defining different spatial structure patterns through the formal grammar. The implementation process is presented as follows. Firstly, extract the long strip scene that may contain the fold structure according to the principles used in choosing fold cores and section lines. Secondly, build and simplify the spatial structure model of the long strip scene by following the ARG approach. Thirdly, convert the ARG model into sentences, and classify the fold types with respect to different grammatical inferences of the sentences. If the sentences cannot be inferred by Anticline Grammar and Syncline Grammar, then it is not a fold. Fourthly, determine the fold landform types by checking whether the terrain containing the fold is a mountain or a valley. The result shows that the proposed method is capable for automatically recognizing the fold landform types in the northern Lushan area. It basically solves the problem in the auto-recognizing of fold landform types for mountainous area, and can be a supplementary reference to the traditional methods used for landform classification.