双目立体视觉的光学标定技术

为了通过二维图像信息计算三维空间中的几何信息,对摄像机系统进行精确标定。在建立和分析双目立体视觉模型和现有摄像机标定方法的基础上,提出一种新的光学标定方法。该方法通过构建和分析双目摄像机理论模型,并改进现有的方法实现了双目立体视觉的光学标定。实验在双目摄像机平台上,采用黑方格模板和通过算法实现了光学参数的标定,使用Levenberg-Marquardt算法优化单应矩阵,并通过最大似然估计法进行参数优化,通过试验模型测定,结果显示其实际标定精度为0.050 9mm,满足双目立体视觉的测定需求。

An optical calibration technology of binocular stereo vision

In order to calculate the geometry information of three dimension space by two dimension image information, accurate calibration of the camera system is necessary. Based on the building and analysis of binocular stereo vision model and the existing calibration method, a new optical calibration method was proposed. In this method, the binocular stereo vision theoretic model was built and analuzed, and the existing calibration method was mended. In the experiment, the optical parameter was calibrated by a black box template and algorithm on the basis of a binocular camera platform. The homographic matrix was optimized by using Levenberg-Marquardt algorithm. Then, these parameters were optimized by the maximum likelihood estimation. The result shows that the accuracy meets the requirement.