变焦相机的标定技术研究

发布时间：2018-04-08 23:35

本文选题：变焦镜头　切入点：相机标定　出处：《郑州大学》2017年硕士论文

【摘要】：变焦镜头的成像机理是通过在一定范围内连续调整光学成像系统的焦距,来调整视场角宽窄、图像分辨率和场景范围,实现自动聚焦、变焦等功能。目前,变焦成像技术被广泛用于摄影、摄像、深度重建、安防监控、视觉跟踪、工业检测等诸多领域。本文的主要研究工作:(1)本文阐述了变焦相机标定的原理及成像机理,通过实验分析了两种畸变模型(乘法模型和除法模型)对标定结果的影响;总结了相机标定的评价准则。(2)分别研究了固定、变内参数变焦镜头的标定方法。提出了基于二消失点的相机自标定方法和绕光心旋转的主动式相机自标定方法;研究了变焦图像的精确特征匹配,并计算了变焦相机的内参数、外参数、缩放中心和光心位移。(3)开展了基于相机标定技术的图像校正、拼接和深度重建的实验,实验结果表明本文提出的标定方法在保证实时性的前提下,提高了精度和运算速度,提升了系统的实用性。本文的创新点:(1)提出了一种基于Harris的亚像素角点检测方法,对传统Harris角点检测算法进行优化改进,实验结果表明该方法提高了亚像素的角点检测的精确性。(2)提出了一种基于二消失点的相机自标定方法,只需被拍摄场景中包含矩形图案,仅从不同视点拍摄四幅图像就能线性求解出相机的内参数。通过与张正友和基于机器视觉软件Halcon的标定结果进行比较,实验结果表明该算法简单高效,具有较高精度和鲁棒性,提高系统的实用性。(3)提出了一种绕相机光心旋转的主动式相机自标定方法,适用于变焦相机和定焦相机。实验结果表明该方法不需标定物,对场景无要求,是一个完全自动的标定过程。(4)提出了一种精确匹配变焦图像的匹配算法,SIFT特征初匹配后,利用变焦图像的对极几何约束、RANSAC算法和缩放中心的几何性质对初匹配进一步优化;通过建立变焦镜头的厚透镜模型,根据已计算出的变焦相机外参数计算光心位移。
[Abstract]:The imaging mechanism of zoom lens is to adjust the width of field of view, image resolution and scene range by continuously adjusting the focal length of the optical imaging system within a certain range, so as to realize the functions of auto-focusing, zoom and so on.At present, zoom imaging technology is widely used in photography, camera, depth reconstruction, security monitoring, visual tracking, industrial detection and many other fields.In this paper, the principle and imaging mechanism of zoom camera calibration are described. The effects of two kinds of distortion models (multiplication model and division model) on calibration results are analyzed experimentally.The evaluation criterion of camera calibration is summarized. (2) the calibration methods of fixed and variable internal parameter zoom lens are studied respectively.The self-calibration method of camera based on two vanishing points and the self-calibration method of active camera with rotation around the center of light are proposed, the accurate feature matching of zoom image is studied, and the internal and external parameters of zoom camera are calculated.The experiments of image correction, splicing and depth reconstruction based on camera calibration technology are carried out. The experimental results show that the calibration method proposed in this paper improves the accuracy and operation speed under the premise of ensuring real-time performance.The utility of the system is improved.This paper proposes a sub-pixel corner detection method based on Harris, which optimizes and improves the traditional Harris corner detection algorithm.Experimental results show that this method improves the accuracy of sub-pixel corner detection. (2) A camera self-calibration method based on two vanishing points is proposed, which only needs to include rectangular patterns in the scene.The internal parameters of the camera can be obtained linearly by only taking four images from different viewpoints.Compared with the calibration results of Zhang Zhengyou and Halcon based on machine vision software, the experimental results show that the algorithm is simple and efficient, and has high accuracy and robustness.To improve the practicability of the system, a self-calibration method for active camera is proposed, which is suitable for zoom and fixed-focus cameras.The experimental results show that the method does not require calibration objects and is a fully automatic calibration process. (4) an accurate matching algorithm for zoom images is proposed after initial matching of sift features.The initial matching is further optimized by using the polar geometric constraints of zoom images and the geometric properties of the zoom center, and the displacement of the optical center is calculated according to the calculated external parameters of the zoom camera by establishing the thick lens model of the zoom lens.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】