以下 3 个 opencv Mat 实例有啥区别？

Posted 2023-02-22

【中文标题】以下 3 个 opencv Mat 实例有啥区别？

【英文标题】：What is the difference between the following 3 opencv Mat instances?以下 3 个 opencv Mat 实例有什么区别？

【发布时间】：2016-07-04 09:31:12

【问题描述】：

在我的程序中，我需要一个 3*3 矩阵 K（相机校准矩阵）和一个 3*3 矩阵 F（基本矩阵）来计算 E（基本矩阵）：E = K.t()FK，但是当我使用

cv::Matx33d K (518.0, 0,325.5, 0,519.0,253.5, 0 , 0, 1);

它会产生错误：行中的“operator*”不匹配：

E = K.t() * F * K;

但是当我使用

cv::Mat K(3,3,CV_64FC1);      //cannot be 32 (float)
K.at<double>(0,0) = 518.0;
K.at<double>(1,1) = 519.0;
K.at<double>(0,2) = 325.5;
K.at<double>(1,2) = 253.5;
K.at<double>(2,2) = 1;

它有效，但是当我使用时

cv::Mat_<double> K = ( cv::Mat_<double>(3, 3) <<
         518.0,    0,325.5,
         0    ,519.0,253.5,
         0    ,    0,    1);

它也可以，但与上面的结果完全不同

谁能告诉我这 3 个 opencv Mat 实例之间有什么区别？谢谢。

这是所有代码

#include <iostream>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/features2d/features2d.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/calib3d/calib3d.hpp>

using namespace std;
using namespace cv;

int main()

    cout<<"Hello SLAM!"<<endl;
    cv::Mat img1,img2,img;


                                                              // first method :can run ,but yeild fault result
    cv::Mat K(3,3,CV_64FC1);      //cannot be 32 (float)
    K.at<double>(0,0) = 518.0;
    K.at<double>(1,1) = 519.0;
    K.at<double>(0,2) = 325.5;
    K.at<double>(1,2) = 253.5;
    K.at<double>(2,2) = 1;


                                                              // second method :cannont build
//    cv::Matx33d K (518.0,    0,325.5,
//                   0    ,519.0,253.5,
//                   0    ,    0,    1);//calibration matrix

//        cv::Matx33d K=cv::Matx33d(518.0,    0,325.5,
//                       0    ,519.0,253.5,
//                       0    ,    0,    1);//calibration matrix

                                                              // third method :can run ,yeild correcy result
    cv::Mat_<double> K = ( cv::Mat_<double>(3, 3) <<
             518.0,    0,325.5,
             0    ,519.0,253.5,
             0    ,    0,    1);


    img1= cv::imread("/home/kylefan/program/cmake_demo/image/rgb1.png",IMREAD_COLOR);
    img2= cv::imread("/home/kylefan/program/cmake_demo/image/rgb2.png",IMREAD_COLOR);

    vector<cv::Point2f> point1;
    vector<cv::Point2f> point2;
    cv::ORB orb;
    vector<cv::KeyPoint>kp1,kp2;
    cv::Mat desp1,desp2;

    orb(img1,cv::Mat(),kp1,desp1,false);
    orb(img2,cv::Mat(),kp2,desp2,false);

    cv::Ptr<cv::DescriptorMatcher> matcher = cv::DescriptorMatcher::create("BruteForce-Hamming");

    double knn_match_ratio = 0.3;
    vector< vector<cv::DMatch>> matches_knn;

    matcher->knnMatch(desp1,desp2,matches_knn,2);

    vector<cv::DMatch> goodmatches;

    for(int i=0;i<matches_knn.size();i++)
    
        if(matches_knn[i][0].distance < knn_match_ratio * matches_knn[i][1].distance)goodmatches.push_back(matches_knn[i][0]);
    

    if(goodmatches.size()<20)
    
        cout<<"too less goodmatches "<<endl;
    

    for(auto m:goodmatches)
    
        point1.push_back(kp1[m.queryIdx].pt);
        point2.push_back(kp2[m.queryIdx].pt);
    
//    for(int j=0;j<goodmatches.size();j++)
//    
//        //cout<<j<<":"<<point1[j].x<<","<<point1[j].y<<","<<point2[j].x<<","<<point2[j].y<<","<<endl;
//    

    cv::Mat F = cv::findFundamentalMat(point1,point2,FM_RANSAC,3,0.99);//in calib3d

    cv::Mat_<double> E = K.t() *F*  K;

    SVD svd(E);

    cv::Matx33d W(0,-1,0,
                  1,0,0,
                  0,0,1);

    cv::Mat_<double> R = svd.u * Mat(W) * svd.vt;
    cv::Mat_<double> t = svd.u.col(2);

    cv::Matx34d P1(R(0,0),R(0,1),R(0,2),t(0),
                   R(1,0),R(1,1),R(1,2),t(1),
                   R(2,0),R(2,1),R(2,2),t(2));

    cout<<goodmatches.size()<<" goodmatches "<<endl;

    cout<<"F="<<F<<endl;
    cout<<"K="<<K<<endl;
    cout<<"E="<<E<<endl;
    cout<<"R="<<R<<endl;
    cout<<"t="<<t<<endl;
    cout<<"P1="<<P1<<endl;

    if(fabsf(determinant(R))-1.0 > 1e-07)
    
        cerr << "det(R) != +-1.0, this is not a rotation matrix" << endl;
    
    else
    
        cout<<"rotation matrix is right"<<endl;
    
    return 0;

【参考方案1】：

在您的第二个版本中，您没有初始化矩阵元素，也没有设置它们。

cv::Mat K(3,3,CV_64FC1);      //cannot be 32 (float)
K.at<double>(0,0) = 518.0;
K.at<double>(1,1) = 519.0;
K.at<double>(0,2) = 325.5;
K.at<double>(1,2) = 253.5;
K.at<double>(2,2) = 1;

这意味着索引 (0,1) 中可以有任何值； (1,0); (2,0) 和 (2,1)。

在创建时使用 0 进行初始化尝试：cv::Mat::zeros()

cv::Mat K = cv::Mat::zeros(3,3,CV_64FC1);      //cannot be 32 (float)
K.at<double>(0,0) = 518.0;
K.at<double>(1,1) = 519.0;
K.at<double>(0,2) = 325.5;
K.at<double>(1,2) = 253.5;
K.at<double>(2,2) = 1;

【讨论】：

非常感谢，确实是这个原因，但是我的第一个版本有什么问题？ cv::Matx33d K (518.0, 0,325.5, 0 ,519.0,253.5, 0 , 0, 1);谢谢！

是F 还是Matx33d？

哦，我明白了，虽然 F 是 Mat ,K 而 K.t() 是 Matx33d，所以我写成 cv::Mat_ E = cv::Mat(K.t()) F cv::Mat(K);哪个是对的！非常感谢！