opencv4.6.0版本进行四路相机标定，并将四路相机的的标定结果内参、外参、畸变系数分别保存为yaml文件的C++代码

以下是使用OpenCV 4.6.0进行四路相机标定并保存标定结果的C++代码，其中假设四路相机的图片已经存储在了vector<vector<Mat>> images中，images[i][j]表示第i路相机的第j张图片：

#include <iostream>
#include <opencv2/opencv.hpp>

using namespace std;
using namespace cv;

int main()
{
    // 相机参数
    const int num_camera = 4;
    const int num_board_cols = 7;
    const int num_board_rows = 10;
    const int num_board_corners = num_board_cols * num_board_rows;
    const Size board_size = Size(num_board_cols, num_board_rows);
    const float square_size = 0.02f; // 单位：米

    // 读取图片
    vector<vector<Mat>> images(num_camera);
    for (int i = 0; i < num_camera; i++)
    {
        for (int j = 0; j < num_board_corners; j++)
        {
            string filename = "camera" + to_string(i) + "/img" + to_string(j) + ".png";
            Mat img = imread(filename);
            images[i].push_back(img);
        }
    }

    // 检测角点
    vector<vector<Point2f>> corners(num_camera);
    for (int i = 0; i < num_camera; i++)
    {
        for (int j = 0; j < num_board_corners; j++)
        {
            Mat gray;
            cvtColor(images[i][j], gray, COLOR_BGR2GRAY);

            vector<Point2f> corner;
            bool found = findChessboardCorners(gray, board_size, corner);
            if (found)
            {
                cornerSubPix(gray, corner, Size(11, 11), Size(-1, -1), TermCriteria(TermCriteria::EPS + TermCriteria::COUNT, 30, 0.1));
                corners[i].push_back(corner);
            }
            else
            {
                cout << "Cannot find chessboard corners in image " << j << " of camera " << i << endl;
            }
        }
    }

    // 检查角点是否符合要求
    bool is_corners_valid = true;
    for (int i = 0; i < num_camera; i++)
    {
        if (corners[i].size() != num_board_corners)
        {
            is_corners_valid = false;
            break;
        }
    }
    if (!is_corners_valid)
    {
        cout << "Error: the number of corners detected is not correct." << endl;
        return -1;
    }

    // 标定
    vector<vector<Point3f>> object_points(1, vector<Point3f>(num_board_corners));
    for (int i = 0; i < num_board_rows; i++)
    {
        for (int j = 0; j < num_board_cols; j++)
        {
            object_points[0][i * num_board_cols + j] = Point3f(j * square_size, i * square_size, 0);
        }
    }

    vector<Mat> camera_matrices(num_camera);
    vector<Mat> dist_coeffs(num_camera);
    vector<vector<Mat>> rvecs(num_camera), tvecs(num_camera);
    double rms = calibrateCamera(object_points, corners, images[0][0].size(), camera_matrices, dist_coeffs, rvecs, tvecs);

    cout << "RMS error: " << rms << endl;

    // 保存相机矩阵、畸变系数、旋转矩阵、平移向量
    for (int i = 0; i < num_camera; i++)
    {
        string filename = "camera" + to_string(i) + ".yaml";
        FileStorage fs(filename, FileStorage::WRITE);

        fs << "camera_matrix" << camera_matrices[i];
        fs << "dist_coeffs" << dist_coeffs[i];
        fs << "rvecs" << rvecs[i];
        fs << "tvecs" << tvecs[i];

        fs.release();
    }

    return 0;
}

在上面的代码中，我们首先读取四路相机的图片，并使用`findChessboardCorners`函数检测角点。如果检测到了角点，则使用`cornerSubPix`函数对角点进行亚像素级别的精细化处理。然后，我们使用`calibrateCamera`函数进行标定，并将结果保存到每个相机的yaml文件中。