地理空间中的空间关系表达和推理

针对地理空间中的应用,归纳了在空间关系的表达与推理中不同于人工智能领域研究的一些特点:在人工智能领域,更注重建立形式化的推理系统;而在地理信息科学中,则需更关注地理空间的特点以及地物的地理语义。该文基于地理空间和地理现象的本质且顾及地理空间认知,总结了地理空间中空间关系表达和推理的特点,具体包括空间的有限性、地球的球面特征、地物的地理语义、地物形状的复杂性、面状地物、特殊的空间关系、空间关系的层次性与尺度相应原则、不确定性、三维与时态特性九方面;进而介绍了地理空间关系表达的两个应用,即地理信息检索和基于对象的图像分析。该文的探讨可为地理信息科学中的相关研究提供方向性指导。

Representation and Reasoning of Spatial Relations in Geographical Space

QSR(Qualitative Spatial Reasoning) has become an important research topic in both artificial intelligence(AI) and geographical information sciences.In AI,spatial relations are formally defined such that an inferable algebra can be established.However,modeling geospatial relations should consider some special properties of geographical space and geographical phenomena.They make representation and reasoning of spatial relations more accordant with the way that human beings manage and process spatial knowledge.This paper summarizes nine aspects of characteristics of geospatial relations.They include finititude and spherality of geographical space;semantics,complex actual shape,and areal distribution of features;cardinalities and hierarchies of spatial relations;and uncertain,3-dimensional,and temporal phenomena.The discussions on those nine aspects and two associated applications,geographical information retrieval and object based image analysis,provide directions for future research.