中文摘要：

从人在识别同名道路时的视觉注意机制入手,提出一种基于层次路划结构关系约束的城市路网自动匹配方法。该方法遵循从突出到一般的原则,利用道路自然延伸形态模式构造比＂路段＂粒度更大的要素单元＂路划＂,首先计算较大粒度的路划结构间的映射关系,继而结合概率松弛迭代优化机制建立结点之间的细部关系。试验结果表明：使用路划为单位进行相似性计算时,候选路划与最大参照解之间的差异率比候选路段与其最大参照解之间的差异率更为显著。在首层路划选取比例为10%时,能够保证首层路划匹配关系的稳健性,通过分层稳定参照空间与层次路划结构关系提高了对匹配关系的辨识能力,在局部结构相似而全局路划结构差异显著的区域具明显优势。

英文摘要：

From the visual attention mechanism when people are looking for the sameroad,this paper proposed a road network matching method based on hierarchical stable strokes.The method adopts the principle of matching from obvious to vital,and constructs larger matching unit＂stroke＂instead of＂road section＂.First,the correspondence of strokes is calculated;Then,the detailed node correspondence is further identified.The results show that：when calculating similarity,the differences rate between candidate strokes and the largest reference solution is more significant than that of road segment.In this example,10 percent of the first layer strokes can ensure the stability of the matching relationship.Layered stable reference and hierarchical strokes improves the ability to identify matching relation,and it has obvious advantages in the region with similar local structure and obviously different global strokes.