A new method of 3D reconstruction baseing on images sequence using hand-hold camera

3D reconstruction is an important technique in the field of computer vision. The traditional 3D reconstruction technology involves the self-calibration method basing on dual absolute quadric. However, some drawbacks in this method cause this technology less robust. First, usually the solution of the dual quadric matrix can not satisfy the positive-definiteness condition caused by loud noise in data; second, if the motion of the camera is too restricted, the calibration degeneracy will turn up causing the reconstruction failed. To these questions, we design a new method for calibrating the camera using the camera pure rotating calibration and decomposing the essential matrix. So, the metric structure can be got directly avoiding the less robustness caused by self-calibration. Experiments show that, using this method the reconstruction has an efficient result and the procedure is convenient.