DT-Loc: Monocular Visual Localization on HD Vector Map Using Distance Transforms of 2D Semantic Detections

Chi Zhang,Hao Liu,Hao Li,Kun Guo,Kuiyuan Yang,Rui Cai,Zhiwei Li

DOI: https://doi.org/10.1109/iros51168.2021.9636419

2021-01-01

Abstract:Localizing a vehicle on a prebuilt HD vector map is a prerequisite for many autonomous driving applications. Existing visual localization approaches usually require a separate local feature layer to function. The separate localization layer suffers from the robustness issue inherited from the local features. Also, it could be difficult to create a feature layer that aligns perfectly with an existing vector map. In this paper, we propose a monocular visual localization method that exploits the vector map directly as the localization layer. The method detects semantic traffic elements from the images and matches them with the vectors in the map. To deal with the harmful problem of false matches, we propose to align the vector map to the distance transforms of the semantic detections, which enables a non-explicit and differentiable data association process. The system is able to achieve centimeter and sub-meter accuracies in lateral and longitudinal directions, respectively.