光学学报, 2018, 38 (3): 0315005, 网络出版: 2018-03-20
基于单应性矩阵的变焦双目视觉标定方法 下载: 1399次
Calibration Method of Binocular Vision System with Zoom Lens Based on Homography Matrix
机器视觉 变焦双目标定 单应性矩阵 变焦镜头 图像匹配 machine vision zoom lens pairs calibration homography matrix zoom lens image matching
摘要
针对现有变焦镜头标定方法难度大、动态精度低等问题, 提出一种基于单应性矩阵的动态变焦双目内外参数估计方法和平面快速重建方法。利用双目图像匹配点及变焦前后的匹配点进行两类单应性矩阵估计;基于变焦数学模型和单应性矩阵, 求解变焦后双目内外参数, 实现畸变后双目参数动态估计与优化;通过双目图像单应性进行平面快速匹配和重建。实验结果表明, 计算的内外参数与标定结果吻合较好;变焦后, 推导的单应性矩阵归一化误差小于0.01, 图像重投影误差小于1 pixel;重建精度小于0.1 mm。
Abstract
The existing calibration methods of zoom lens are mostly characterized by great difficulty and low dynamic accuracy. We propose a dynamical estimation method of internal and external parameters based on homography matrix for binocular camera with zoom lens and a fast plane reconstruction method. Firstly, we estimate two types of homography matrix using matching points of binocular images and matching points before and after zooming. Secondly, using mathematic model of zoom lens and the homography matrix, we calculate internal and external parameters of camera pair after zooming to realize the dynamic estimation and optimization of binocular parameters after zoom lens distortion. Finally, we achieve the fast matching and reconstruction of the planes by homography matrix of binocular images. The experiment results show that the calculated internal and external parameters are in good agreement with the calibration results. After zooming, the normalized error of computational homography matrix is less than 0.01, the error of image reprojection is less than 1 pixel, and the accuracy of reconstruction is less than 0.1 mm.
姜涛, 程筱胜, 崔海华, 贾华宇, 张逢骏. 基于单应性矩阵的变焦双目视觉标定方法[J]. 光学学报, 2018, 38(3): 0315005. Jiang Tao, Cheng Xiaosheng, Cui Haihua, Jia Huayu, Zhang Fengjun. Calibration Method of Binocular Vision System with Zoom Lens Based on Homography Matrix[J]. Acta Optica Sinica, 2018, 38(3): 0315005.