相机模型与标定（十三）--鱼眼相机标定

最新推荐文章于 2025-04-02 20:52:49 发布

Eason.wxd

最新推荐文章于 2025-04-02 20:52:49 发布

阅读量1.5w

点赞数

分类专栏：相机标定

相机标定专栏收录该内容

14 篇文章

订阅专栏

原文：

http://blog.csdn.net/u010784534/article/details/50474371

先看一张鱼眼相机拍摄出来的结果：

这里写图片描述

从图中可以看出很明显的畸变。对鱼眼相机标定，有时候也可以用普通相机的标定方法对其进行标定，但是却不能保证去畸变后的效果是最好的。因此对于Gopro等鱼眼镜头拍摄出来的图像去畸变，最好的方法就是采用鱼眼相机标定方法进行标定。

鱼眼相机模型

鱼眼相机的内参模型依然可以表示为：

⎧ ⎩ ⎨ ⎪ ⎪ f x 00 0 f y 0 c x c y 1 ⎫ ⎭ ⎬ ⎪ ⎪

这与普通镜头的成像模型没有区别。两者之间的区别主要体现在畸变系数，鱼眼相机的畸变系数为{

k1,k2,k3,k4 }，畸变系数不同，就导致鱼眼相机的投影关系也发生了变化，主要变化发生在考虑畸变情况下的投影关系转化：
设（X,Y,Z）为空间中一个三维点，它在成像平面内的成像坐标为（u,v），在考虑畸变的情况下，

⎧ ⎩ ⎨ ⎪ ⎪ x c y c z c ⎫ ⎭ ⎬ ⎪ ⎪ = R * ⎧ ⎩ ⎨ ⎪ ⎪ X Y Z ⎫ ⎭ ⎬ ⎪ ⎪ + T

a=xc/zc,b=yc/zc

r2=a2+b2

θ=atan(r)

θ′=θ(1+k1θ2+k2θ4+k3θ6+k4θ8)

x′=(θ′/r)xc

y′=(θ′/r)yc

u=fxx′+cx

v=fyy′+cy

OpenCV实现鱼眼相机标定

利用opencv实现鱼眼相机的标定和普通相机标定的标定流程基本一致，具体流程如下：

检测角点
cv::findChessboardCorners(InputArray image, Size patternSize, OutputArray corners, int flags=CALIB_CB_ADAPTIVE_THRESH+CALIB_CB_NORMALIZE_IMAGE}
获得棋盘标定板的角点位置，使用
cornerSubPix(InputArray image, InputOutputArray corners, Size winSize, Size zeroZone, TermCriteria criteria)获取角点更精细的检测结果
初始化标定板上角点的三维坐标
开始标定
double fisheye::calibrate(InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, const Size& image_size, InputOutputArray K, InputOutputArray D, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs, int flags=0, TermCriteria criteria=TermCriteria(TermCriteria::COUNT + TermCriteria:: EPS, 100, DBL_EPSILON))
注意：K，D 分别表示内参矩阵和畸变系数向量，在定义时要定义为double型，这里推荐使用Matx33d和Vec4d数据类型，更为方便简单。objectPoints，imagePoints可以是float型，也可以是double型，但是再stereorectify中需要时double型。flags的可选项有很多，其中需要注意的是必须要指定CALIB_FIX_SKEW，代表求解时假设内参中 fx=fy .
4.评定误差(可选项)

以上就是鱼眼相机标定的基本流程，部分代码片段如下：

<code class="hljs cpp has-numbering">    <span class="hljs-keyword">for</span> (<span class="hljs-keyword">int</span> i = <span class="hljs-number">0</span>; i != image_count; i++)
    {
        <span class="hljs-built_in">cout</span> << <span class="hljs-string">"Frame #"</span> << i + <span class="hljs-number">1</span> << <span class="hljs-string">"..."</span> << endl;
        <span class="hljs-built_in">string</span> image_Name;
        <span class="hljs-built_in">stringstream</span> stream;
        stream << (i + startNum);
        stream >> image_Name;
        image_Name = path_ChessboardImage + image_Name + <span class="hljs-string">".jpg"</span>;
        cv::Mat image = imread(image_Name);
        Mat image_gray;
        cvtColor(image, image_gray, CV_RGB2GRAY);
        <span class="hljs-stl_container"><span class="hljs-built_in">vector</span><Point2f></span> corners;                   
        <span class="hljs-keyword">bool</span> patternFound = findChessboardCorners(image_gray, board_size, corners,
            CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE + CALIB_CB_FAST_CHECK);
        <span class="hljs-keyword">if</span> (!patternFound || fullcornersNum != corners.size())
        {
            <span class="hljs-built_in">cout</span> << <span class="hljs-string">"can not find chessboard corners!\n"</span>;
            <span class="hljs-keyword">continue</span>;
        }
        <span class="hljs-keyword">else</span>
        {
            cornerSubPix(image_gray, corners, Size(<span class="hljs-number">11</span>, <span class="hljs-number">11</span>), Size(-<span class="hljs-number">1</span>, -<span class="hljs-number">1</span>), TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, <span class="hljs-number">30</span>, <span class="hljs-number">0.1</span>));
            count = count + corners.size();
            corners_Seq.push_back(corners);
            successImageNum = successImageNum + <span class="hljs-number">1</span>;
            image_Seq.push_back(image);
        }
    }
    <span class="hljs-comment">/************************************************************************
    摄像机定标
    *************************************************************************/</span>
    <span class="hljs-stl_container"><span class="hljs-built_in">vector</span><<span class="hljs-stl_container"><span class="hljs-built_in">vector</span><Point3f></span>></span>  object_Points;        <span class="hljs-comment">/****  保存定标板上角点的三维坐标   ****/</span>

    Mat image_points = Mat(<span class="hljs-number">1</span>, count, CV_32FC2, Scalar::all(<span class="hljs-number">0</span>));  <span class="hljs-comment">/*****   保存提取的所有角点   *****/</span>
    <span class="hljs-stl_container"><span class="hljs-built_in">vector</span><<span class="hljs-keyword">int</span>></span>  point_counts;                                                         
    <span class="hljs-comment">/* 初始化定标板上角点的三维坐标 */</span>
    <span class="hljs-keyword">for</span> (<span class="hljs-keyword">int</span> t = <span class="hljs-number">0</span>; t<successImageNum; t++)
    {
        <span class="hljs-stl_container"><span class="hljs-built_in">vector</span><Point3f></span> tempPointSet;
        <span class="hljs-keyword">for</span> (<span class="hljs-keyword">int</span> i = <span class="hljs-number">0</span>; i<board_size.height; i++)
        {
            <span class="hljs-keyword">for</span> (<span class="hljs-keyword">int</span> j = <span class="hljs-number">0</span>; j<board_size.width; j++)
            {
                <span class="hljs-comment">/* 假设定标板放在世界坐标系中z=0的平面上 */</span>
                Point3f tempPoint;
                tempPoint.x = i*square_size.width;
                tempPoint.y = j*square_size.height;
                tempPoint.z = <span class="hljs-number">0</span>;
                tempPointSet.push_back(tempPoint);
            }
        }
        object_Points.push_back(tempPointSet);
    }
    <span class="hljs-keyword">for</span> (<span class="hljs-keyword">int</span> i = <span class="hljs-number">0</span>; i< successImageNum; i++)
    {
        point_counts.push_back(board_size.width*board_size.height);
    }
    <span class="hljs-comment">/* 开始定标 */</span>
    Size image_size = image_Seq[<span class="hljs-number">0</span>].size();
    cv::Matx33d intrinsic_matrix;    <span class="hljs-comment">/*****    摄像机内参数矩阵    ****/</span>
    cv::Vec4d distortion_coeffs;     <span class="hljs-comment">/* 摄像机的4个畸变系数：k1,k2,k3,k4*/</span>
    <span class="hljs-built_in">std</span>::<span class="hljs-stl_container"><span class="hljs-built_in">vector</span><cv::Vec3d></span> rotation_vectors;                           <span class="hljs-comment">/* 每幅图像的旋转向量 */</span>
    <span class="hljs-built_in">std</span>::<span class="hljs-stl_container"><span class="hljs-built_in">vector</span><cv::Vec3d></span> translation_vectors;                        <span class="hljs-comment">/* 每幅图像的平移向量 */</span>
    <span class="hljs-keyword">int</span> flags = <span class="hljs-number">0</span>;
    flags |= cv::fisheye::CALIB_RECOMPUTE_EXTRINSIC;
    flags |= cv::fisheye::CALIB_CHECK_COND;
    flags |= cv::fisheye::CALIB_FIX_SKEW;
    fisheye::calibrate(object_Points, corners_Seq, image_size, intrinsic_matrix, distortion_coeffs, rotation_vectors, translation_vectors, flags, cv::TermCriteria(<span class="hljs-number">3</span>, <span class="hljs-number">20</span>, <span class="hljs-number">1e-6</span>));</code>

标定结果：
这里写图片描述