Camera Pose Estimation

Camera pose estimation determines where a camera is and how it is oriented, enabling localization, 3D reconstruction, visual odometry, and robotic navigation. Classical approaches use point, line, or region correspondences with PnP and relative-pose solvers, then refine poses through bundle adjustment or pose-graph optimization; visual odometry and SLAM extend these methods across sequences. Learned approaches range from pose-regression networks to feed-forward 3D models, often combining geometric constraints, depth priors, or camera data from inertial, event, thermal, and other sensors. Work spans settings such as indoor and outdoor localization, driving, medical imaging, and wide-angle or dynamic scenes, while balancing accuracy and efficiency against drift, ambiguous geometry, calibration errors, and poor generalization to unfamiliar conditions.