道路网多特征匹配优化算法

同名道路匹配技术是道路数据集成、更新和融合的重要前提。道路网匹配在智能交通(intelligent transportation system,ITS)与位置服务(location-based service,LBS)等方面具有重要的研究价值和应用意义。本文提出了一种道路网多特征匹配优化算法:首先从形状、距离、语义3方面分别设计了基于面积累积的形状差、综合中值Hausdorff距离和全局加权属性项距离3种相似性度量,以更准确地描述道路待匹配对之间的特征差异;然后通过SVM对相似性特征样本集训练,以构建道路网回归匹配模型;最后利用此模型对未知匹配结果道路待匹配对进行匹配结果预测。大量试验结果表明,本文算法对非线性偏差明显的道路网数据能够实现较高的匹配准确率和召回率,能有效地用于包含多重匹配关系的道路网匹配。

An Optimization Algorithm for Multi-characteristics Road Network Matching

Identifying homonymous road objects is a crucial prerequisite to the integration,updating and fusion of road data.Road networks matching is of great theoreticalresearch value and practical significance in aspect of intelligent transportation system and location-based Service.This paper proposed an optimization algorithm for multi-characteristics road network matching.Designed from shape,distance and semantics aspects,three similarity characteristics—shape differences based on area accumulated, mixed median Hausdorff distance and distance with global weighted attributes,described candidate corresponding pairs more accurately.Then,the matching regression model could be then constructed by training the similarity samples set through SVM algorithm.Finally,the constructed model can be used to predict whether the road matching pairs were matched.A great number of experiments show that the algorithm achieves a robust matching precision and recall even for road networks data with apparent non-rigid deviation.And the proposed method can be effectively applied for road networks matching with multiple matching relationship.