Optimal road matching by relaxation to min-cost network flow
Hangbin Wu, Shan Xu, Shengke Huang, Junhua Wang, Xuan Yang, Chun Liu, Yunling Zhang
Abstract
Road matching is the prerequisite for integrating different data to obtain a complete, correct road network. The growth of multi-source data calls for the advancement of matching methods. There are many methods of road matching, ranging from traditional methods using threshold or weighted judgment to methods based on statistical and optimization models. This article provides an optimal matching method by relaxation to min-cost network flow. The matching problem is considered an optimization problem, i.e., finding a plan to match features in two datasets to minimize the total difference. The optimization-based matching method can overcome traditional method’s suboptimal nature. Referencing more road information than the other methods make our method superior in its flexibility and applicability to different matching relationships. Additionally, the iterative relaxation by angle, flow, and shortest path-based constraints is introduced to reduce ambiguous matching. We demonstrate the model in real-world experiments. Our method achieves a better result than the other four methods: it has a more stable performance for matching pairs of more complicated correspondence relations and easily confused situations. The average matching precision, recall, and f-score were 0.985 in two test sites.