Camera Agnostic Two-Head Network for Ego-Lane Inference

Vision-based ego-lane inference using High-Definition (HD) maps is essential in autonomous driving and advanced driver assistance systems. The traditional approach necessitates well-calibrated cameras, which confines variation of camera configuration, as the algorithm relies on intrinsic and extrinsic calibration. In this paper, we propose a learning-based ego-lane inference by directly estimating the ego-lane index from a single image. To enhance robust performance, our model incorporates the two-head structure inferring ego-lane in two perspectives simultaneously. Furthermore, we utilize an attention mechanism guided by vanishing point-and-line to adapt to changes in viewpoint without requiring accurate calibration. The high adaptability of our model was validated in diverse environments, devices, and camera mounting points and orientations.

I. Introduction

Identifying the location of an ego vehicle on the road plays a fundamental role in providing relevant information to the user within advanced driver assistance systems. In order to operate with limited onboard sensors, visual localization is commonly employed combined with low-cost Global Navigation Satellite System (GNSS). To extend the adoption of these systems to open road environments without assumptions about the camera, the robustness and installation flexibility have emerged as new challenges.