3-D reconstruction and stereo self calibration for augmented reality

More about the book

This work focuses on developing new methods for self-calibration of a rigid stereo camera system, with broader implications for robot hand-eye calibration across various applications. Stereo self-calibration computes intrinsic and extrinsic parameters of a stereo rig without prior knowledge of the rig's movement or the scene's geometry. The resulting stereo parameters—rotation and translation between the left and right cameras—enable the creation of depth maps, essential for accurately rendering virtual objects in real scenes (Augmented Reality). The methods were evaluated against real and synthetic data and compared with existing algorithms. Additionally, an optical tracking system with a camera mounted on an endoscope was calibrated without a pattern using the proposed extended hand-eye calibration algorithm. The self-calibration methods are practical, relying solely on temporal feature tracking, which simplifies the process compared to left-to-right tracking with unknown parameters. They compute intrinsic and extrinsic camera parameters during self-calibration, eliminating the need for calibration patterns. The approach also extends to hand-eye calibration using structure-from-motion. A key challenge in hand-eye calibration is the necessity for general camera movements to compute rigid transformations; insufficient motion results in incomplete parameters for depth map computation. This work addresses data selection methods