ORB-SLAM3 Testing

This is a testing implementation that runs ORB‑SLAM3 with ROS, on the KITTI dataset. Modifications are limited to ROS/package integration, dataset I/O, and logging; the core SLAM pipeline follows the original ORB‑SLAM3 design.

Algorithm features:

• ORB feature extraction: fast, rotation- and scale-invariant keypoints are detected and described.
• Visual tracking: features are matched frame-to-frame to estimate the camera pose relative to the map.
• Map management: selected frames become keyframes; 3D map points are created and refined by local bundle adjustment.
• Loop closing & relocalization: loop closures are detected, a pose‑graph optimization is run to remove drift, and full bundle adjustment refines the map.
• Multi-sensor support: ORB‑SLAM3 supports monocular/stereo/RGB‑D and optional IMU fusion to recover scale and improve robustness.