CASIA OpenIR  > 毕业生  > 硕士学位论文
鱼眼镜头标定及其定位技术研究
其他题名Fisheye Lens Calibration and Positioning
高舒
学位类型工程硕士
导师邹伟
2013-05-27
学位授予单位中国科学院大学
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词鱼眼镜头 标定 椭圆检测 P4p问题 多路标定位 Fisheye Lens Calibration Ellipse Detection P4p Problem Multiple Landmarks Positioning
摘要基于视觉的定位方式是机器人领域应用较为普遍的定位方式之一,目前已得到了国内外众多科研工作者的广泛重视,具有较为广阔的应用前景。基于普通视觉的定位方式视野狭窄,路标检测困难,从而导致定位可靠性难以时时得到保证。鱼眼视觉具有近180度的宽视野成像特性,能够较好地克服以上缺陷。本论文以国家自然科学基金项目“基于鱼眼视觉的移动机器人运动估计及其视觉伺服控制研究”为背景,对鱼眼视觉的标定问题和室内环境下的定位问题进行了研究。 本论文的主要工作包括鱼眼镜头标定板的设计和图像特征的提取、鱼眼镜头的标定和鱼眼图像的校正以及基于多个路标的鱼眼镜头定位三个部分。具体介绍如下: 第一、结合当前鱼眼镜头标定算法的发展现状,设计了一种适用于鱼眼镜头的标定板,并利用标定板的特征实现了特征点的提取,为鱼眼镜头的标定奠定了基础。首先,该鱼眼镜头标定板根据鱼眼镜头的成像特点设计,可以适应鱼眼图像径向畸变较大而切向畸变可以忽略的特点,在大广角和大畸变的情况下依旧能够保持特征点的特性不改变;其次,为了能够实现鱼眼镜头的实时准确定位,利用标定板的特点,实现了鱼眼镜头成像区域和特征椭圆的识别和拟合;最后,在椭圆拟合的基础上,实现了标定板椭圆的排序。 第二、在标定板的设计和特征识别的基础上,利用采集到的特征信息,实现鱼眼图像的标定和鱼眼图像校正。首先, 在对算法原理进行深入理解的基础上,对Kannala的优化迭代标定算法进行了实现,并结合鱼眼成像模型,推导出了该算法Levenberg-Marquardt优化所需的Jacobian矩阵;然后,分别利用Kannala方法和张正友方法对鱼眼镜头的参数进行了标定,并分别在所标定的两组摄像机参数的基础上,实现鱼眼图像的校正。最后,通过编写Matlab程序和C++程序得到了标定参数和校正结果,并对结果进行了比较。 第三、在上述工作的基础上,实现了基于P4P问题求解方法的多路标成像定位算法。为了充分利用鱼眼镜头大广角的优势,分别在特征圆之间距离较近和较远两种情况下进行定位实验,并对实验数据进行了分析和对比。实验结果表明,利用鱼眼镜头,可以在路标稀疏的情况下实现摄像机的准确定位。 最后对论文中的工作进行了总结,并讨论了可在其基础上进行的拓展工作。
其他摘要The positioning method based on vision is one of the common ways in engineering application. So far, it is emphasized by many researchers, and will have a broad prospect in the future. The positioning approach based on traditional cameras has a narrow view, and it is hard to detect landmarks. As a result, the reliability is unstable. One imaging feature of fisheye vision is 180-degree wide field. So it can overcome the defects discussed above. This paper studies the fisheye calibration and indoor positioning based on the project supported by National Natural Science Foundation of China which is titled “Mobile robot motion estimation and its visual servo control based on fish eye vision”. This thesis mainly studies the design of calibration board used for the fisheye camera, the extraction of image characteristics, the fisheye lens calibration, the image correction and the positioning of fisheye camera based on multiple landmarks. The main contents of this paper are listed as follows: Firstly, a calibration board used for fisheye camera is designed, and the extraction of feature points is realized using this board. It lays the foundations for the fisheye lens calibration. For one thing, this board can adapt to the large radial distortion and ignore the tangential distortion so that the characteristics of landmarks can be kept. For another, the fisheye lens imaging area and ellipses are fitted in order to implement positioning. Then, the sorting of ellipses on the board is achieved. Secondly, the calibration of fisheye lens and the correction of fisheye images are obtained based on collected information. For one thing, the fisheye lens is calibrated using Kannala calibration algorithm and the Jacobian Matrix which is needed in Levenberg-Marquardt algorithm is derived. For another, calibrate fisheye lens using Kannala algorithm and Zhengyou Zhang algorithm respectively. Then, the correction of the fisheye image is achieved based on the parameters which are obtained from calibration. In addition, Matlab and C++ program are written in order to implement the processes above and the results are compared. Thirdly, on the basis of the work above, the positioning algorithm based on multiple landmarks is implemented using P4P solving method. Positioning experiments are carried in the conditions of arranging characteristic circles closer and farther in order to make full use of the large wide-angle of the fisheye lens. Then, the data are analyzed and com...
馆藏号XWLW1933
其他标识符2010E8014669011
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/7666
专题毕业生_硕士学位论文
推荐引用方式
GB/T 7714
高舒. 鱼眼镜头标定及其定位技术研究[D]. 中国科学院自动化研究所. 中国科学院大学,2013.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_2010E801466901(2634KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[高舒]的文章
百度学术
百度学术中相似的文章
[高舒]的文章
必应学术
必应学术中相似的文章
[高舒]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。