摘要在计算机视觉当中，摄像机的标定过程不仅是整个系统的前提，也影响整个成像效果的精度与好坏。三维重建就需要相机标定后得到的参数再加上立体匹配后的结果。实时并精确标定结果对于各类侦测导航民用用途甚至是在军事用途，都有着很重要的意义。所以去研究更高效，更精确的相机标定有着很重要的社会实际价值。79483

     本文对于相机标定技术这一主题，进行了各方面的研究与分析，主要内容如下几点：

     介绍了相机标定现如今的研究背景及意义以及标定的重要性，各种与相机有关的坐标系，总结了一下各像点位置在各个坐标中的转换。介绍了Harris角点检测的原理与发展，并且提出了针对Harris角点检测的改进方法SUSAN算法。这种方法计算出模板核心点处的灰度，并与周围相近灰度值的像素比较，通过对区域大小的判定，选出像素点作为角点的候选点，在对这些点进行Harris检测，这样极大地提高了精度。

    综合介绍，分析了国内外比较有影响力的相机标定法，分别主要介绍了Tsai两步法，传统法中的基于DLT模型下的摄像机标定方法，张正友平面标定法。详细给出上面三个方法的原理算法模型与计算公式，还有其各个步骤。改进部分：基于线性模型给出了畸变模型；对于两步法进行最小二乘法优化，使得数值迅速收敛，让初始值更接近真值，从而提高参数精度。

    拍摄了棋盘标定格的图片，使用Matlab设计了仿真实验，利用了Camera Calibration toolbox 进行图片标定。给出了详细的过程，利用算法求出相机的内外参数，再考虑畸变系数，给出了误差估计，并保存消除畸变后的图片。

    对全文进行了总结，分析本文研究的成果，也给出不足。

毕业论文关键词   摄像机标定  畸变模型  角点检测   非线性优化

毕业设计说明书外文摘要

Title    Research on the Technique of Camera Calibration             

Abstract In domain of the computer vision,the process of camera calibration is not only the premise of the whole system,but also affect the overall accuracy and quality of imaging。It is need the results of stereo matching and the parameter of camera to complete three-dimensional reconstruction。Real-time and accurate results of the calibration is important significance for  the detection of various types of navigation even in civilian use for military purposes 。 So to study more efficient and more accurate calibration of the camera has a very important social real value。 This article on the subject for the technique of camera calibration , carried out research and analysis of all aspects of the main contents of the following:

I suppose to introduces the importance of research background and significance of camera calibration and  a variety of camera-related coordinate system ，summed up the position of each image point conversion in the respective coordinates。And i also want to introduces the principle and development of Harris corner detection, and SUSAN algorithm is proposed for an improved method of Harris corner detection。This method calculates the gray template core point, and relatively close to the surrounding pixel gray value, determined by the size of the region, selected pixels as candidate corner points in the detection of these points Harris, this  method  greatly improves the accuracy。 

And I give a comprehensive introduction, analyzed more influential abroad camera calibration method, respectively Tsai introduced the two-step method, the traditional method of camera calibration method based on DLT model, Zhang Zhengyou plane calibration method。Principle model and formulas detailed above three methods are given algorithm, as well as its inpidual steps。 Fixes: based on the linear model gives distortion model; two-step method for optimizing the least squares method, so that the value converges quickly, so that the initial value is closer to the true value, thereby improving the accuracy of the parameters。