运行OpenCV3.4.1的双目相机标定例程注释,OpenCV3.4.1+VS2017安装配置

新版本总是在迭代升级,目前opencv最新版本为3.4.1VS已经到了2017版本,由于opencv每个版本对应的VS都需要有不同的配置,这里也是鼓捣了好几天才完成配置,并且跑完这个双目标定的例程,期间踩了不少坑,下面贴出来分享给小伙伴,同时还有程序注释


opencv安装:https://opencv.org/releases.html


VS2017安装:https://www.visualstudio.com/zh-hans/thank-you-downloading-visual-studio/?sku=Professional&rel=15


这里记得安装一些必要组件!!建议全程在线,因为后边可能需要获取一些计算机权限,360等软件会禁止权限,导致最后各种配置不成功!

接下来就是配置opencv了:

  1.  配置环境变量:D:\work_soft\opencv\build\x64\vc15\bin
                这里VC15对应VS2017,

    2.在VS2017配置属性管理器Debug x64:

     VC++目录: 
        (Include Directories)
        包含目录(头文件)D:\work_soft\opencv\build\include
                                      D:\work_soft\opencv\build\include\opencv

                                      D:\work_soft\opencv\build\include\opencv2



        (Library Directories)

        库目录(库文件)   D:\work_soft\opencv\build\x64\vc15\lib

这里下边需要继承响应的库文件,否则会提示缺少msvcprt.lib等库文件,很麻烦


    连接器-输入:
        (Linker-Input-Additional Dependencies)

        附加依赖项 opencv_world341d.lib



好了,到了这里就配置结束了,接下来尝试下运行第一个标定例程;

 D:\work_soft\opencv是我的opencv安装位置

  1. 找到官方的例程    D:\work_soft\opencv\sources\samples\cpp\stereocalib.cpp
  2. 找到官方的图片和.xml   D:\work_soft\opencv\sources\samples\data
  3. 新建工程,并且把第二部得到的图片和.xml放到新建的工程中
  4. 选择x64和debug环境

  5. 更改一行代码cv::CommandLineParser parser(argc, argv, "{w|9|}{h|6|}{s|1.0|}{nr||}{help||}{@input|stereo_calib.xml|}");

这里的@input|stereo_calib.xml|指的是你存入stereo_calib.xml的地址,之前是放在和stereocalib.cpp同目录下。


以下是代码注释

/* This is sample from the OpenCV book. The copyright notice is below */


/* *************** License:**************************
   Oct. 3, 2008
   Right to use this code in any way you want without warranty, support or any guarantee of it working.


   BOOK: It would be nice if you cited it:
   Learning OpenCV: Computer Vision with the OpenCV Library
     by Gary Bradski and Adrian Kaehler
     Published by O'Reilly Media, October 3, 2008


   AVAILABLE AT:
     http://www.amazon.com/Learning-OpenCV-Computer-Vision-Library/dp/0596516134
     Or: http://oreilly.com/catalog/9780596516130/
     ISBN-10: 0596516134 or: ISBN-13: 978-0596516130


   OPENCV WEBSITES:
     Homepage:      http://opencv.org
     Online docs:   http://docs.opencv.org
     Q&A forum:     http://answers.opencv.org
     Issue tracker: http://code.opencv.org
     GitHub:        https://github.com/opencv/opencv/
   ************************************************** */


#include "opencv2/calib3d.hpp"
#include "opencv2/imgcodecs.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/imgproc.hpp"


#include <vector>
#include <string>
#include <algorithm>
#include <iostream>
#include <iterator>
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>


using namespace cv; //包含相应的Opencv命名空间
using namespace std;


static int print_help()
{
    cout <<
            " Given a list of chessboard images, the number of corners (nx, ny)\n"
            " on the chessboards, and a flag: useCalibrated for \n"
            "   calibrated (0) or\n"
            "   uncalibrated \n"
            "     (1: use cvStereoCalibrate(), 2: compute fundamental\n"
            "         matrix separately) stereo. \n"
            " Calibrate the cameras and display the\n"
            " rectified results along with the computed disparity images.   \n" << endl;
    cout << "Usage:\n ./stereo_calib -w=<board_width default=9> -h=<board_height default=6> -s=<square_size default=1.0> <image list XML/YML file default=../data/stereo_calib.xml>\n" << endl;
    return 0;
}


/************相机标定程序***************/
static void
StereoCalib(const vector<string>& imagelist, Size boardSize, float squareSize, bool displayCorners = false, bool useCalibrated=true, bool showRectified=true)
{
    if( imagelist.size() % 2 != 0 ) //判断图片是否成对
    {
        cout << "Error: the image list contains odd (non-even) number of elements\n";
        return;
    }


    const int maxScale = 2; //设定寻找角点的图像尺寸,若scale未找到,则将图像放大寻找角点
    // ARRAY AND VECTOR STORAGE:


    vector<vector<Point2f> > imagePoints[2]; //分别为左右图像的角点
    vector<vector<Point3f> > objectPoints; //物体
    Size imageSize; //??


    int i, j, k, nimages = (int)imagelist.size()/2; //nimages为左或右图像的个数


    imagePoints[0].resize(nimages);
    imagePoints[1].resize(nimages);
    vector<string> goodImageList;


    for( i = j = 0; i < nimages; i++ ) //依次寻找13对图片
    {
        for( k = 0; k < 2; k++ ) //依次寻找左右图片
        {
            const string& filename = imagelist[i*2+k];
            Mat img = imread(filename, 0); //载入灰度图 0代表灰度图
            if(img.empty())
                break;
            if( imageSize == Size() ) //判断图像尺寸是否达到预先设置的要求
                imageSize = img.size();
            else if( img.size() != imageSize )
            {
                cout << "The image " << filename << " has the size different from the first image size. Skipping the pair\n";
                break;
            }
            bool found = false;
//设置图像矩阵的引用(指针),此时指向左右视图的矩阵首地址 
            vector<Point2f>& corners = imagePoints[k][j]; //赋值角点个数及位置
            for( int scale = 1; scale <= maxScale; scale++ )
            {
                Mat timg;
                if( scale == 1 )
                    timg = img;
                else
                    resize(img, timg, Size(), scale, scale, INTER_LINEAR_EXACT);
                found = findChessboardCorners(timg, boardSize, corners, //找角点函数,得到内角点坐标
                    CALIB_CB_ADAPTIVE_THRESH | CALIB_CB_NORMALIZE_IMAGE);


// drawChessboardCorners(timg, boardSize, corners, found); //画出角点,自己调试时使用,但会降低精度,正式运行时应该注释掉


                if( found )
                {
                    if( scale > 1 )
                    {
                        Mat cornersMat(corners);
                        cornersMat *= 1./scale;
                    }
                    break;
                }
            }
            if( displayCorners )
            {
                cout << filename << endl;
                Mat cimg, cimg1;
                cvtColor(img, cimg, COLOR_GRAY2BGR);
                drawChessboardCorners(cimg, boardSize, corners, found);
                double sf = 640./MAX(img.rows, img.cols);
                resize(cimg, cimg1, Size(), sf, sf, INTER_LINEAR_EXACT);
                imshow("corners", cimg1);
                char c = (char)waitKey(500);
                if( c == 27 || c == 'q' || c == 'Q' ) //Allow ESC to quit
                    exit(-1);
            }
            else
                putchar('.');
            if( !found )
                break;
            cornerSubPix(img, corners, Size(11,11), Size(-1,-1),
                         TermCriteria(TermCriteria::COUNT+TermCriteria::EPS,
                                      30, 0.01));
        }
        if( k == 2 )
        {
            goodImageList.push_back(imagelist[i*2]);
            goodImageList.push_back(imagelist[i*2+1]);
            j++;
        }
    }
    cout << j << " pairs have been successfully detected.\n"; //命令行打印检测到的图片
    nimages = j; //nimages为左右图像个数
    if( nimages < 2 )
    {
        cout << "Error: too little pairs to run the calibration\n";
        return;
    }


    imagePoints[0].resize(nimages); //左相机 角点位置
    imagePoints[1].resize(nimages); //右相机 角点位置
    objectPoints.resize(nimages);


    for( i = 0; i < nimages; i++ )
    {
        for( j = 0; j < boardSize.height; j++ )
            for( k = 0; k < boardSize.width; k++ )
                objectPoints[i].push_back(Point3f(k*squareSize, j*squareSize, 0));
    }


    cout << "Running stereo calibration ...\n";


    Mat cameraMatrix[2], distCoeffs[2]; //相机参数  畸变矩阵
    cameraMatrix[0] = initCameraMatrix2D(objectPoints,imagePoints[0],imageSize,0);
    cameraMatrix[1] = initCameraMatrix2D(objectPoints,imagePoints[1],imageSize,0);
    Mat R, T, E, F; //R旋转矩阵 T平移矩阵 E本征矩阵 F输出基本矩阵
//最关键的地方,求解校正后的相机参数
    double rms = stereoCalibrate(objectPoints, imagePoints[0], imagePoints[1],
                    cameraMatrix[0], distCoeffs[0],
                    cameraMatrix[1], distCoeffs[1],
                    imageSize, R, T, E, F,
                    CALIB_FIX_ASPECT_RATIO +
                    CALIB_ZERO_TANGENT_DIST +
                    CALIB_USE_INTRINSIC_GUESS +
                    CALIB_SAME_FOCAL_LENGTH +
                    CALIB_RATIONAL_MODEL +
                    CALIB_FIX_K3 + CALIB_FIX_K4 + CALIB_FIX_K5,
                    TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 100, 1e-5) );
    cout << "done with RMS error=" << rms << endl;


// CALIBRATION QUALITY CHECK
// because the output fundamental matrix implicitly
// includes all the output information,
// we can check the quality of calibration using the
// epipolar geometry constraint: m2^t*F*m1=0
    double err = 0; //计算投影标定误差
    int npoints = 0;
    vector<Vec3f> lines[2]; //极线
    for( i = 0; i < nimages; i++ ) //水平校正
    {
        int npt = (int)imagePoints[0][i].size();
        Mat imgpt[2];
        for( k = 0; k < 2; k++ )
        {
            imgpt[k] = Mat(imagePoints[k][i]);
            undistortPoints(imgpt[k], imgpt[k], cameraMatrix[k], distCoeffs[k], Mat(), cameraMatrix[k]);
            computeCorrespondEpilines(imgpt[k], k+1, F, lines[k]);
        }
        for( j = 0; j < npt; j++ )
        {
            double errij = fabs(imagePoints[0][i][j].x*lines[1][j][0] +
                                imagePoints[0][i][j].y*lines[1][j][1] + lines[1][j][2]) +
                           fabs(imagePoints[1][i][j].x*lines[0][j][0] +
                                imagePoints[1][i][j].y*lines[0][j][1] + lines[0][j][2]);
            err += errij;
        }
        npoints += npt;
    }
    cout << "average epipolar err = " <<  err/npoints << endl;


    // save intrinsic parameters
    FileStorage fs("intrinsics.yml", FileStorage::WRITE); //存储内参
    if( fs.isOpened() )
    {
        fs << "M1" << cameraMatrix[0] << "D1" << distCoeffs[0] <<
            "M2" << cameraMatrix[1] << "D2" << distCoeffs[1];
        fs.release();
    }
    else
        cout << "Error: can not save the intrinsic parameters\n";


//外参数  
// R--右相机相对左相机的旋转矩阵  
// T--右相机相对左相机的平移矩阵  
// R1,R2--左右相机校准变换(旋转)矩阵  3×3  
// P1,P2--左右相机在校准后坐标系中的投影矩阵 3×4  
// Q--视差-深度映射矩阵,我利用它来计算单个目标点的三维坐标


    Mat R1, R2, P1, P2, Q;
    Rect validRoi[2]; //图像校正之后,会对图像进行裁剪,这里的validROI就是指裁剪之后的区域
//进行立体校正
    stereoRectify(cameraMatrix[0], distCoeffs[0],
                  cameraMatrix[1], distCoeffs[1],
                  imageSize, R, T, R1, R2, P1, P2, Q,
                  CALIB_ZERO_DISPARITY, 1, imageSize, &validRoi[0], &validRoi[1]);

/*
* R T 左相机到右相机的旋转平移矩阵  R 3*3      T  3*1  T中第一个Tx为 基线长度
立体校正的时候需要两幅图像共面并且行对准 以使得立体匹配更加的可靠
使得两幅图像共面的方法就是把两个摄像头的图像投影到一个公共成像面上,这样每幅图像从本图像平面投影到公共图像平面都需要一个旋转矩阵R
stereoRectify 这个函数计算的就是从图像平面投影都公共成像平面的旋转矩阵Rl,Rr。 Rl,Rr即为左右相机平面行对准的校正旋转矩阵。
左相机经过Rl旋转,右相机经过Rr旋转之后,两幅图像就已经共面并且行对准了。
其中Pl,Pr为两个相机的投影矩阵,其作用是将3D点的坐标转换到图像的2D点的坐标:P*[X Y Z 1]' =[x y w]
Q矩阵为重投影矩阵,即矩阵Q可以把2维平面(图像平面)上的点投影到3维空间的点:Q*[x y d 1] = [X Y Z W]。其中d为左右两幅图像的视差
*/


    fs.open("extrinsics.yml", FileStorage::WRITE);
    if( fs.isOpened() )
    {
        fs << "R" << R << "T" << T << "R1" << R1 << "R2" << R2 << "P1" << P1 << "P2" << P2 << "Q" << Q;
        fs.release();
    }
    else
        cout << "Error: can not save the extrinsic parameters\n";
//opencv可以辨认相机的物理位置
    // OpenCV can handle left-right
    // or up-down camera arrangements
    bool isVerticalStereo = fabs(P2.at<double>(1, 3)) > fabs(P2.at<double>(0, 3));


// COMPUTE AND DISPLAY RECTIFICATION 计算并显示校正信息
    if( !showRectified )
        return;


    Mat rmap[2][2];
// IF BY CALIBRATED (BOUGUET'S METHOD)
    if( useCalibrated )
    {
        // we already computed everything
    }
// OR ELSE HARTLEY'S METHOD
    else
 // use intrinsic parameters of each camera, but
 // compute the rectification transformation directly
 // from the fundamental matrix
    {
        vector<Point2f> allimgpt[2];
        for( k = 0; k < 2; k++ )
        {
            for( i = 0; i < nimages; i++ )
                std::copy(imagePoints[k][i].begin(), imagePoints[k][i].end(), back_inserter(allimgpt[k]));
        }
        F = findFundamentalMat(Mat(allimgpt[0]), Mat(allimgpt[1]), FM_8POINT, 0, 0);
        Mat H1, H2;
        stereoRectifyUncalibrated(Mat(allimgpt[0]), Mat(allimgpt[1]), F, imageSize, H1, H2, 3);


        R1 = cameraMatrix[0].inv()*H1*cameraMatrix[0];
        R2 = cameraMatrix[1].inv()*H2*cameraMatrix[1];
        P1 = cameraMatrix[0];
        P2 = cameraMatrix[1];
    }


/*
根据stereoRectify 计算出来的R 和 P 来计算图像的映射表 mapx,mapy
mapx,mapy这两个映射表接下来可以给remap()函数调用,来校正图像,使得两幅图像共面并且行对准
ininUndistortRectifyMap()的参数newCameraMatrix就是校正后的摄像机矩阵。在openCV里面,校正后的计算机矩阵Mrect是跟投影矩阵P一起返回的。
所以我们在这里传入投影矩阵P,此函数可以从投影矩阵P中读出校正后的摄像机矩阵
*/


    //Precompute maps for cv::remap()
    initUndistortRectifyMap(cameraMatrix[0], distCoeffs[0], R1, P1, imageSize, CV_16SC2, rmap[0][0], rmap[0][1]);
    initUndistortRectifyMap(cameraMatrix[1], distCoeffs[1], R2, P2, imageSize, CV_16SC2, rmap[1][0], rmap[1][1]);


/*
显示校正结果
把左右两幅图像显示到同一个画面上
这里只显示了最后一副图像的校正结果。并没有把所有的图像都显示出来
*/


    Mat canvas;
    double sf;
    int w, h;
    if( !isVerticalStereo )
    {
        sf = 600./MAX(imageSize.width, imageSize.height);
        w = cvRound(imageSize.width*sf);
        h = cvRound(imageSize.height*sf);
        canvas.create(h, w*2, CV_8UC3);
    }
    else
    {
        sf = 300./MAX(imageSize.width, imageSize.height);
        w = cvRound(imageSize.width*sf);
        h = cvRound(imageSize.height*sf);
        canvas.create(h*2, w, CV_8UC3);
    }


    for( i = 0; i < nimages; i++ )
    {
        for( k = 0; k < 2; k++ )
        {
            Mat img = imread(goodImageList[i*2+k], 0), rimg, cimg;
            remap(img, rimg, rmap[k][0], rmap[k][1], INTER_LINEAR); //对准左右图像,
            cvtColor(rimg, cimg, COLOR_GRAY2BGR);
            Mat canvasPart = !isVerticalStereo ? canvas(Rect(w*k, 0, w, h)) : canvas(Rect(0, h*k, w, h));
            resize(cimg, canvasPart, canvasPart.size(), 0, 0, INTER_AREA);
            if( useCalibrated )
            {
                Rect vroi(cvRound(validRoi[k].x*sf), cvRound(validRoi[k].y*sf),
                          cvRound(validRoi[k].width*sf), cvRound(validRoi[k].height*sf));
                rectangle(canvasPart, vroi, Scalar(0,0,255), 3, 8);
            }
        }


        if( !isVerticalStereo ) //画上对应的线条 
            for( j = 0; j < canvas.rows; j += 16 )
                line(canvas, Point(0, j), Point(canvas.cols, j), Scalar(0, 255, 0), 1, 8); //画平行线 
        else
            for( j = 0; j < canvas.cols; j += 16 )
                line(canvas, Point(j, 0), Point(j, canvas.rows), Scalar(0, 255, 0), 1, 8);
        imshow("rectified", canvas); //输出立体校正后的图片
        char c = (char)waitKey();
        if( c == 27 || c == 'q' || c == 'Q' )
            break;
    }
}




static bool readStringList( const string& filename, vector<string>& l )
{
    l.resize(0);
    FileStorage fs(filename, FileStorage::READ);
    if( !fs.isOpened() )
        return false;
    FileNode n = fs.getFirstTopLevelNode();
    if( n.type() != FileNode::SEQ )
        return false;
    FileNodeIterator it = n.begin(), it_end = n.end();
    for( ; it != it_end; ++it )
        l.push_back((string)*it);
    return true;
}


int main(int argc, char** argv)
{
    Size boardSize;
    string imagelistfn;
    bool showRectified;
    cv::CommandLineParser parser(argc, argv, "{w|9|}{h|6|}{s|1.0|}{nr||}{help||}{@input|stereo_calib.xml|}");
    if (parser.has("help"))
        return print_help();
    showRectified = !parser.has("nr");
    imagelistfn = parser.get<string>("@input");
    boardSize.width = parser.get<int>("w"); //获取图像的宽
    boardSize.height = parser.get<int>("h"); //获取图像的高
    float squareSize = parser.get<float>("s"); //正方形棋盘格子 边长
    if (!parser.check()) //标定板 格子尺寸参数错误 
    {
        parser.printErrors();
        return 1;
    }
    vector<string> imagelist; //定义个一个字符串容器 存放图像列表
    bool ok = readStringList(imagelistfn, imagelist);
    if(!ok || imagelist.empty())
    {
        cout << "can not open " << imagelistfn << " or the string list is empty" << endl;
        return print_help();
    }
//标定的主程序
    StereoCalib(imagelist, boardSize, squareSize, false, true, showRectified);
    return 0;
}



阅读更多
想对作者说点什么?

博主推荐

换一批

没有更多推荐了,返回首页