Road Networks Matching Using Multiple Logistic Regression

Identifying corresponding objects is crucial in the process of heterogeneous road network matching. This paper proposed a road network matching method based on a multiple logistic regression algorithm. First, three dissimilar characteristics integrating both spatial and non-spatial features were used to describe the difference of the corresponding pairs of road objects;the minimum angle of the orientation, the mixed median Hausdorff distance, and semantic discrepancy. Using these three characteristics as variables of multiple logistic regression, we built a basic multiple logistic regression matching model. Samples to train the final road matching model were acquired to obtain matching results by predicting probability of each candidate road matching pair. Experimental results show that this method needs no exact feature weights and thresholds, and can solve the matching result problems stemming from over-reliance on single variable. This method has good adaptability, with higher precision and recall rates.