Abstract
The goal of object matching is to identify objects representing the same real entity across multiple spatial datasets. This involves comparing and linking data from different sources using similarity measures, with the final matching decision made by combining these measures. Object matching is especially valuable for creating accurate and complete spatial datasets for underground networks, where data often come from various sources and may have imperfections like imprecision or incompleteness. The Dempster-Shafer (DS) theory, which uses mass functions to model data imperfections, is considered the best method for combining imperfect data. However, previous DS-based approaches produced highly conflicting results when many potential candidates for an object existed. In this work, we present an improved DS-based line matching approach for wastewater networks. Our key contributions include introducing candidate ranking, bidirectional measure combination, and mixed models to convert similarity measures into masses. We validated our approach through experiments on both synthetic and real-world datasets. The results demonstrate that our contributions significantly reduce conflict and improve the accuracy and correctness of the matching decision.