Critical points correspondence between vector homonymous features based on string matching method

Automated matching of critical points from vector homonymous features is important in the field of spatial data fusion and conflation. In view of complicated shape characteristic of map features, string matching method was improved and used to solve the critical points correspondence problem in this paper. Firstly, map features were simplified base on overlay similarity by analyzing how similar they were between simplified map feature and the original map feature, and thus critical points could be extracted automatically. Secondly, traditional differential invariant was alternated to robust local feature invariant defined in this paper, and lastly the best critical points correspondence relationship can be found by minimizing the value of edit cost function with dynamic programming method. Many experiments indicate that the matching result of the proposed method were not only invariant to similarity transformation, but also were insensitive to noise and scale difference, and therefore the proposed method can be applied in the spatial data conflation efficiently which lay a solid foundation for point accuracy improvement.