基于格式塔识别原则挖掘空间分布模式

面向空间群目标的分布模式识别是空间数据挖掘比较关注的问题。本研究基于空间认知原理与视觉识别格式塔完形原则并结合空间聚类方法对该问题进行研究,提出用于描述实体间差异的"视觉距离"概念,其定义综合考虑视觉识别中的位置、方向、大小差异,通过Delaunay三角网计算几何构造建立该距离计算的模型。在实验基础上提出基于最小支撑树MST的聚类方法,获得与视觉认知相一致的结果。研究试图表明一个观念,即通用性的数据处理模型在GIS实际应用时,需要根据GIS作为"空间认知"科学的原理,作技术方法上的改进,需要考虑认知主体在感知、辨析、识别、推理不同思维过程中的认知心理原则。

Polygon Cluster Pattern Mining Based on Gestalt Principles

The pattern recognition of spatial cluster object is an active issue in the field of data mining.This study attempts to investigate the application of spatial cognition principles in this question combing the Gestalt principles and spatial clustering method.To replace the traditional Euclidean distance,a new distance concept,namely visual distance is built to represent the difference between neighbor entities,which considers the difference not only in geometric position but also in size and layout orientation.Because these factors greatly affect the visual judgment in spatial cognition.Based on the Delaunay triangulation the study presents the geometric construction similar to Voronoi diagram to compute the visual distance.The clustering method based on MST structure shows the building group recognition is consistent with the manual identification.This study tries to state that the utilization of common mathematic model in GIS domain has to take into account the impacts of spatial cognition to design the improved model which from the perspective of recognition considers the psychological process,such as perception,identification,recognition and reasoning.