Ego Lane Estimation Using Visual Information and High Definition Map.

This paper presents a new ego lane estimation method which makes full use of visual information and high-definition map. Compared with existing matching-based method, our method realizes the fusion of multiple road elements using all observed visual information and high-precision map and builds filters to apply map constraints, which achieves higher lane matching accuracy. Lane change information is also used to implement lane tracking for the robustness of our method. The experiment results show that an integrated navigation system based on this method can achieve a lane matching accuracy of more than 94%.