In the measurement field, objects contact, irregular curved surfaces'geometric features, subjective operations and empirical formula, etc, contribute to unqualification of traditional methods on the aspect of high precision.3-D measurement systems based on optics and computer vision can provide a non-contact, rapid and reliable solution. It is of important theoretical significance and practical value to research of computer vision measurement technology.First, feature points are extracted on sub-pixel level and the correlation pixels are matched up adopting manual interactive operation. Second, a non-linear camera model is established. Zhang's calibration method is studied and then a constructure reprojected calibration method using planar pattern is proposed. The result is optimized using non-linear least squares method and then be used to adjust the images. Several fundamental matrix solutions are compared and algorithms of shape-from-motion reconstruction is proposed using epipolar geometry constrain. At last, the effectiveness of the entire technological process is tested and verified in the experiments and geometric probabilty theory is utilized in uncertainty analysis of the whole system.The thesis makes research of some crucial technologies for measurement-oriented surface reconstruction with analysis of 3-D vision measurement and proposes a lieable solution and makes verification based on the study of current works and literatures theoretically and experimentally. |