GPS-guided Vision-based Lane Detection for Autonomous Vehicles

Vision-based lane detection methods often rely on lane markings in image streams as the only source of information to determine lane boundaries. However, different road conditions such as strong shadows and lighting changes present significant challenges in extracting lane marking features, which may result in inaccurate or inconsistent lane boundary detections. This paper proposes GPS-Guided Lane Detection (GPS-LD) as a robust lane detection method that uses conventional image processing techniques for feature extraction and global route data as prior information to determine lane boundaries. The proposed method uses the route from the mission planner to determine the shape of the road in front of the vehicle. The road shape extracted from the route data is then used as a reference to cluster lane marking features and to verify the lane boundary detection results. The proposed method is evaluated on the KITTI dataset and the detection performance is comparable to state-of-the-art lane detection methods.