Linear Six-Point Algorithm of Camera Self-Calibration and 3D Reconstruction from Single-View or Multi-views

摘要

Camera self-calibration and 3D reconstruction is one of the most important tasks in computer vision system. In a kind of applications, the 3D shape model is a priori known. How to calibrate the used camera and how to determine the 3D pose of the viewed target from its single-view or multi-views in this case is studied. A linear six-point algorithm is developed. The computation in the suggested algorithm applies null subspace projection and some of QR decompositions. The affections of noise strength, point-number and frame-number on algorithm performance are experimentally demonstrated.