Visual Odometry
Overview
Visual odometry estimates how a camera moves through a scene over time, providing motion information for robots and vehicles when GPS or other positioning systems are unavailable. Geometric methods track points, lines, or image intensities and refine poses with tools such as optical flow, epipolar geometry, filtering, and bundle adjustment; learned methods predict motion or improve matching, often combining both approaches. Work spans monocular, stereo, RGB-D, event-camera, and multi-sensor systems, with growing attention to long sequences, dynamic scenes, sparse observations, and real-time deployment. A central challenge is balancing accuracy and low drift with robustness to changing environments, uncertain scale or calibration, and limited computation, as evaluated on benchmarks such as KITTI, EuRoC, TUM-RGBD, and TartanAir.
Latest papers
55 papers6 papers in the last four weeks, down 14% on the four weeks before. 0.0% of all new papers.