建筑群组合直线模式识别的模板匹配方法

The template matching approach to combined collinear pattern recognition in building groups

建筑群空间分布模式识别对制图综合、多尺度表达及空间数据挖掘具有重要意义。针对局部异质性明显的建筑群直线模式识别提取问题,本文提出基于模板匹配的建筑群组合直线模式识别方法。首先分析研究组合直线模式的认知特征和定义;然后,利用建筑物的空间邻近性、尺寸和方向约束进行聚类,获得建筑物间的空间邻近关系和扩展对齐关系;以放大的建筑物最小面积外接矩形作为初始匹配模板;最后,在确定模板分布间隔及方向的基础上,考虑组合直线模式的直线性、相似性和局部异质性约束条件,通过连续构建模板进行搜索匹配,识别出组合直线模式。试验表明本文方法能有效识别出组合直线模式,其识别结果符合人类认知特点。

Spatial distribution pattern recognition of buildings is significant to cartographic generalization,multi-scale representation and spatial data mining.This paper presents an approach to recognize combined collinear patterns with local heterogeneity.Firstly the cognitive characteristics and definitions of combined collinear pattern are analyzed.Then neighborhood relationship and extension alignment between buildings are obtained by clustering based on the proximity,size and orientation.Take the enlarged building’s smallest minimum bounding rectangle as the initial matching template.Finally,considering the constraints of straightness,similarity and local heterogeneity,the combined linear patterns are extracted by searching and matching buildings through continuous templates constructed based on distribution spacing and direction.Experiments show that the proposed method is effective for combined collinear pattern recognition with the agreement of human spatial cognitive characteristics.