Stereo Visual SLAM Based On Plane Features In Road Environments

The road surface, as a fundamental feature for robots in road environments, should be given due attention. This paper improves the accuracy and efficiency of visual SLAM(Simultaneous Localization and Mapping) by fully utilizing the geometric information provided by the road surface. We propose a stereo vision-based SLAM framework in road environments, which is capable of real-time extraction of the observed road surface by the robot, and tightly integrates geometric constraints provided by the road surfaces. The observed road surfaces are modeled as plane features. And then coplanar constraints between visual feature points and their corresponding road surface are constructed. In order to validate the performance of our proposed system, we conducted experiments on the KITTI odometry dataset and compared our proposed system with state-of-the-art stereo vision-based SLAM algorithms. The experimental results demonstrate that our system effectively improves the accuracy of stereo vision-based SLAM in road environments.