Abstract: Automated matching for road networks plays an important role in map registration, map update and change detection, and this paper carries out research on the problem of matching different levels of road network automatically. Because the great majority of the existing methods for matching road network are based on finding optimal solution locally, which will cause mismatch often occurs when homonymous road features differ greatly in their locations and there is 1:1 or 1:N corresponding relationship between them, so in order to solve such a problem, this paper constructs global optimization model of matching road network through analyzing locations, shape and topological structure of road junctions and road arcs, and this model not only considers local similarity between road nodes but also takes matching situations of their adjacent nodes into account. The optimal solution is found by using probability relaxation method, and matching relationship between road nodes and between road arcs can be obtained finally. In probability relaxation method, topological structure similarity between nodes is chosen as the initial probability, and shape similarity between road arcs is chosen as compatibility function. Many experiments indicate that our method can ensure that matching results agree with each other globally, and has a higher accuracy as a result, furthermore, as a strong constraint condition, global optimization can ensure that our method still gets a satisfactory result even when homonymous features differ greatly in their locations and there is 1:N matching relationship between roads, and avoids the difficult problem of setting proper weights to some degree.