关键词: 摄像机标定；双目立体测量；三维重建；非线性优化

Abstract:

A practical 3-step binocular camera calibration method was proposed herein.Firstly,assuming pin-hole model,the internal and external parameters’ initial values were estimated by using projective transformation principles.Secondly,
the parameters were optimized by LM nonlinear optimization algorithm taking last step’s results as initial values and targeting at minimizing re-projection errors.Thirdly and lastly,another LM optimization was conducted adding plate’s feature 3D
coordinates as optimization parameters,afterwards real metrics were obtained by scaling.
Compared to traditional 2-step calibration, instead of strictly assuming calibration object’s 3D precision,this method takes great advantages of nonlinear optimization
algorithm to get real features’3D coordinates,thus eliminating the infection of errors of 3D features on calibration.The validity and efficiency of our method can be demonstrated by the results of post-calibration 3D reconstruction experiments.

Key words: camera calibration;binocular stereo vision;3D reconstruction;nonlinear optimization