时间:2022-11-09 09:12:25 | 栏目:C代码 | 点击:次
bool findChessboardCorners( InputArray image, Size patternSize, OutputArray corners, int flags = CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE );
第一个参数是输入的棋盘格图像(可以是8位单通道或三通道图像);
第二个参数是棋盘格内部的角点的行列数(注意:不是棋盘格的行列数,如棋盘格的行列数分别为4、8,而内部角点的行列数分别是3、7,因此这里应该指定为cv::Size(3, 7));
第三个参数是检测到的棋盘格角点,类型为std::vectorcv::Point2f。
第四个参数flag,用于指定在检测棋盘格角点的过程中所应用的一种或多种过滤方法,可以使用下面的一种或多种,如果都是用则使用OR:
drawChessboardCorners( InputOutputArray image, Size patternSize, InputArray corners, bool patternWasFound );
find4QuadCornerSubpix( InputArray img, InputOutputArray corners, Size region_size );
void cornerSubPix( InputArray image, InputOutputArray corners, Size winSize, Size zeroZone, TermCriteria criteria );
double calibrateCamera( InputArrayOfArrays objectPoints, InputArrayOfArrays imagePoints, Size imageSize, InputOutputArray cameraMatrix, InputOutputArray distCoeffs, OutputArrayOfArrays rvecs, OutputArrayOfArrays tvecs, int flags = 0, TermCriteria criteria = TermCriteria(TermCriteria::COUNT + TermCriteria::EPS, 30, DBL_EPSILON) );
objectPoints,世界坐标,用vector<vector>,输入x,y坐标,z坐标为0
imagePoints,图像坐标,vector<vector>
imageSize,图像的大小用于初始化标定摄像机的image的size
cameraMatrix,内参数矩阵
distCoeffs,畸变矩阵
rvecs,位移向量
tvecs,旋转向量
flags,可以组合:
CV_CALIB_USE_INTRINSIC_GUESS:使用该参数时,将包含有效的fx,fy,cx,cy的估计值的内参矩阵cameraMatrix,作为初始值输入,然后函数对其做进一步优化。如果不使用这个参数,用图像的中心点初始化光轴点坐标(cx, cy),使用最小二乘估算出fx,fy(这种求法好像和张正友的论文不一样,不知道为何要这样处理)。注意,如果已知内部参数(内参矩阵和畸变系数),就不需要使用这个函数来估计外参,可以使用solvepnp()函数计算外参数矩阵。
CV_CALIB_FIX_PRINCIPAL_POINT:在进行优化时会固定光轴点,光轴点将保持为图像的中心点。当CV_CALIB_USE_INTRINSIC_GUESS参数被设置,保持为输入的值。
CV_CALIB_FIX_ASPECT_RATIO:固定fx/fy的比值,只将fy作为可变量,进行优化计算。当
CV_CALIB_USE_INTRINSIC_GUESS没有被设置,fx和fy的实际输入值将会被忽略,只有fx/fy的比值被计算和使用。
CV_CALIB_ZERO_TANGENT_DIST:切向畸变系数(P1,P2)被设置为零并保持为零。
CV_CALIB_FIX_K1,…,CV_CALIB_FIX_K6:对应的径向畸变系数在优化中保持不变。如果设置了CV_CALIB_USE_INTRINSIC_GUESS参数,就从提供的畸变系数矩阵中得到。否则,设置为0。
CV_CALIB_RATIONAL_MODEL(理想模型):启用畸变k4,k5,k6三个畸变参数。使标定函数使用有理模型,返回8个系数。如果没有设置,则只计算其它5个畸变参数。
CALIB_THIN_PRISM_MODEL (薄棱镜畸变模型):启用畸变系数S1、S2、S3和S4。使标定函数使用薄棱柱模型并返回12个系数。如果不设置标志,则函数计算并返回只有5个失真系数。
CALIB_FIX_S1_S2_S3_S4 :优化过程中不改变薄棱镜畸变系数S1、S2、S3、S4。如果cv_calib_use_intrinsic_guess设置,使用提供的畸变系数矩阵中的值。否则,设置为0。
CALIB_TILTED_MODEL (倾斜模型):启用畸变系数tauX and tauY。标定函数使用倾斜传感器模型并返回14个系数。如果不设置标志,则函数计算并返回只有5个失真系数。
CALIB_FIX_TAUX_TAUY :在优化过程中,倾斜传感器模型的系数不被改变。如果cv_calib_use_intrinsic_guess设置,从提供的畸变系数矩阵中得到。否则,设置为0。
void initUndistortRectifyMap(InputArray cameraMatrix, InputArray distCoeffs, InputArray R, InputArray newCameraMatrix, Size size, int m1type, OutputArray map1, OutputArray map2)
这里自己画一个棋盘格用作标定,长度为1280像素,宽490像素,横向10方格,纵向7方格
std_cb = Vision::makeCheckerboard(1280, 490, 10, 7, 0, (char *)"../blizzard/res/calibration/std_cb.png");
效果如图
Vision是我个人创建的视觉类,可以用来绘制标准的棋盘格。
头文件vision.h
// // Created by czh on 18-10-16. // #ifndef OPENGL_PRO_VISION_H #define OPENGL_PRO_VISION_H #include "opencv2/opencv.hpp" #include <opencv2/core/core.hpp> #include <opencv2/highgui/highgui.hpp> #include <opencv2/imgcodecs/imgcodecs.hpp> #include "iostream" class Vision { public: static cv::Mat read(std::string file_path, int flags = cv::IMREAD_ANYCOLOR | cv::IMREAD_ANYDEPTH); static cv::Mat write(std::string file_path, int flags = cv::IMREAD_ANYCOLOR | cv::IMREAD_ANYDEPTH); static void dispConfig(cv::Mat img); static cv::Mat makeCheckerboard(int bkgWidth, int bkgHeight, int sqXnum, int sqYnum = 0, int borderThickness = 0, char *savePath = NULL); private: }; #endif //OPENGL_PRO_VISION_H
源文件vision.cpp
// // Created by czh on 18-10-16. // #include "vision.h" #include "string.h" using namespace std; using namespace cv; const char *findName(const char *ch) { const char *name = strrchr(ch, '/'); return ++name; } cv::Mat Vision::read(std::string file_path, int flags) { printf("#Vision read\n"); cv::Mat img; img = cv::imread(file_path, flags); if (img.data == NULL) { printf("\tError:vision read\n"); } else { dispConfig(img); } return img; } void Vision::dispConfig(cv::Mat img) { printf("\tpixel:%d*%d, channels:%d\n", img.size().width, img.size().height, img.channels()); } cv::Mat Vision::makeCheckerboard(int bkgWidth, int bkgHeight, int sqXnum, int sqYnum, int thickNum, char *savePath) { if(sqYnum == 0){ sqYnum = sqXnum; } if(savePath == NULL){ char *defaultPath = (char *)"../res/calibration/maths.png"; savePath = defaultPath; } int checkboardX = 0;//棋盘x坐标 int checkboardY = 0;//棋盘y坐标 int xLen = bkgWidth / sqXnum;//x方格长度 int yLen = bkgHeight / sqYnum;//y方格长度 cv::Mat img(bkgHeight + thickNum * 2, bkgWidth + thickNum * 2, CV_8UC4, cv::Scalar(0, 255, 255, 255)); for (int i = 0; i < img.rows; i++) { for (int j = 0; j < img.cols; j++) { if (i < thickNum || i >= thickNum + bkgHeight || j < thickNum || j >= thickNum + bkgWidth) { img.at<Vec<uchar, 4>>(i, j) = cv::Scalar(0, 0, 0, 255); continue; } checkboardX = j - thickNum; checkboardY = i - thickNum; if (checkboardY / yLen % 2 == 0) { if ((checkboardX) / xLen % 2 == 0) { img.at<Vec<uchar, 4>>(i, j) = cv::Scalar(255, 255, 255, 255); } else { img.at<Vec<uchar, 4>>(i, j) = cv::Scalar(0, 0, 0, 255); } } else{ if ((checkboardX) / xLen % 2 != 0) { img.at<Vec<uchar, 4>>(i, j) = cv::Scalar(255, 255, 255, 255); } else { img.at<Vec<uchar, 4>>(i, j) = cv::Scalar(0, 0, 0, 255); } } } } imwrite(savePath, img); //保存生成的图片 printf("#makeCheckerboard %d*%d\n", bkgWidth + thickNum, bkgHeight + thickNum); return img; }
用A4纸打印棋盘格,相机拍摄照片。
我偷懒,拿了别人的标定照片
下面是相机标定代码
cv::imwrite("../blizzard/res/calibration/cb_source.png", cb_source); printf("#Start scan corner\n"); cv::Mat img; std::vector<cv::Point2f> image_points; std::vector<std::vector<cv::Point2f>> image_points_seq; /* 保存检测到的所有角点 */ if (cv::findChessboardCorners(cb_source, cv::Size(aqXnum, aqYnum), image_points, 0) == 0) { printf("#Error: Corners not find "); return 0; } else { cvtColor(cb_source, img, CV_RGBA2GRAY); cv::imwrite("../blizzard/res/calibration/cb_gray.png", img); //find4QuadCornerSubpix(img, image_points, cv::Size(5, 5)); cv::cornerSubPix(img, image_points, cv::Size(11, 11), cv::Size(-1, -1), cv::TermCriteria(CV_TERMCRIT_ITER + CV_TERMCRIT_EPS, 30, 0.01)); image_points_seq.push_back(image_points); cv::Mat cb_corner; cb_corner = cb_source.clone(); drawChessboardCorners(cb_corner, cv::Size(aqXnum, aqYnum), image_points, true); cv::imwrite("../blizzard/res/calibration/cb_corner.png", cb_corner); } printf("#Start calibrate\n"); cv::Size square_size = cv::Size(14.2222, 12); std::vector<std::vector<cv::Point3f>> object_points; /* 保存标定板上角点的三维坐标 */ cv::Mat cameraMatrix = cv::Mat(3, 3, CV_32FC1, cv::Scalar::all(0)); /* 摄像机内参数矩阵 */ cv::Mat distCoeffs = cv::Mat(1, 5, CV_32FC1, cv::Scalar::all(0)); /* 摄像机的5个畸变系数:k1,k2,p1,p2,k3 */ std::vector<cv::Mat> tvecsMat; /* 每幅图像的旋转向量 */ std::vector<cv::Mat> rvecsMat; /* 每幅图像的平移向量 */ std::vector<cv::Point3f> realPoint; for (int i = 0; i < aqYnum; i++) { for (int j = 0; j < aqXnum; j++) { cv::Point3f tempPoint; /* 假设标定板放在世界坐标系中z=0的平面上 */ tempPoint.x = i * square_size.width; tempPoint.y = j * square_size.height; tempPoint.z = 0; realPoint.push_back(tempPoint); } } object_points.push_back(realPoint); printf("#objectPoints: %ld\n", sizeof(object_points[0])); std::cout << object_points[0] << std::endl; printf("#image_points: %ld\n", sizeof(image_points_seq[0])); std::cout << image_points << std::endl; printf("#image size\n"); std::cout << SCREEN_WIDTH << "*" << SCREEN_HEIGHT << std::endl; cv::calibrateCamera(object_points, image_points_seq, cb_source.size(), cameraMatrix, distCoeffs, rvecsMat, tvecsMat, CV_CALIB_FIX_K3); std::cout << "tvecsMat:\n" << tvecsMat[0] << std::endl; std::cout << "rvecsMat:\n" << rvecsMat[0] << std::endl; std::cout << "#cameraMatrix:\n" << cameraMatrix << std::endl; std::cout << "#distCoeffs:\n" << distCoeffs << std::endl;
cv::Mat cb_final; cv::Mat mapx = cv::Mat(cb_source.size(), CV_32FC1); cv::Mat mapy = cv::Mat(cb_source.size(), CV_32FC1); cv::Mat R = cv::Mat::eye(3, 3, CV_32F); //initUndistortRectifyMap(cameraMatrix, distCoeffs, R, cv::Mat(), cb_source.size(), CV_32FC1, // mapx, mapy); //cv::remap(cb_source, cb_final, mapx, mapy, cv::INTER_LINEAR); undistort(cb_source, cb_final, cameraMatrix, distCoeffs); cv::imwrite("../blizzard/res/calibration/cb_final.png", cb_final);
1.校正前的图片
2.校正后的图片