鱼眼摄像头单目标定双目标定

准确地说是使用鱼眼模型对摄像头进行单目和双目标定
直接贴代码如下，如果是非鱼眼，把相关函数名和相关入参改下就可以了，代码中处于注释状态

2019-12-7 21:47:07
后续又思考了一下，OpenCV鱼眼模型应该参考的是《A Generic Camera Model and Calibration Method for Conventional, Wide-Angle, and Fish-Eye Lenses》。那既然是通用模型，对一般相机也能用，也就是说拿到一个相机，直接用鱼眼模型来标定。。。这个后面验证一下。


#include <opencv2/opencv.hpp>
#include <iostream>
#include <dirent.h>
#include <fstream>using namespace std;bool hasEnding (std::string const &fullString, std::string const &ending) {if (fullString.length() >= ending.length()) {return (0 == fullString.compare (fullString.length() - ending.length(), ending.length(), ending));} else {return false;}
}// 扩展名必须为 jpg或者 png
void getImgNames(const char* imgsFolderL, const char* imgsFolderR, std::vector<std::string> &imgNamesL,std::vector<std::string> &imgNamesR)
{DIR* dp;struct dirent* dirp;dp = opendir(imgsFolderL);if(dp == NULL){std::cout << "open directory error: " << imgsFolderL << std::endl;}while((dirp = readdir(dp)) != NULL){std::string filename = dirp->d_name;if(filename.at(0) == '.'){continue;} else if(!hasEnding(filename, ".jpg") || !hasEnding(filename, ".png")) {continue;}filename = "/" + filename;std::cout << (imgsFolderL + filename) << std::endl;std::cout << (imgsFolderR + filename) << std::endl;imgNamesL.push_back(imgsFolderL + filename);imgNamesR.push_back(imgsFolderR + filename);}closedir(dp);}static bool saveCameraParams(const std::string &filename, int &leftHeight, int &leftWidth, cv::Mat &leftD, cv::Mat &leftK, cv::Mat &leftR, cv::Mat leftP,int &rightHeight, int &rightWidth, cv::Mat &rightD, cv::Mat &rightK, cv::Mat &rightR, cv::Mat &rightP, cv::Mat &trans, cv::Mat& Trl)
{std::ofstream fs(filename.c_str());if(!fs.is_open()){return false;}/* std::time_t tt; *//* std::time(&tt); *//* struct tm *t2 = localtime(&tt); *//* char buf[1024]; *//* std::strftime(buf, sizeof(buf), "%c", t2); *//* std::string str = "1.0"; */fs << "%YAML:1.0" << std::endl;fs << std::endl;fs << "#-----------------------------------------" << std::endl;fs << "# Camera Parameters. Adjust them!" << std::endl;fs << "#-----------------------------------------" << std::endl;fs << std::endl;/* fs << "#calibration_time " << buf << std::endl; */fs << std::endl;fs << "Camera.fx: " << leftP.ptr<double>(0)[0] << std::endl;fs << "Camera.fy: " << leftP.ptr<double>(1)[1] << std::endl;fs << "Camera.cx: " << leftP.ptr<double>(0)[2] << std::endl;fs << "Camera.cy: " << leftP.ptr<double>(1)[2] << std::endl;fs << std::endl;fs << "Camera.k1: " << 0.0 << std::endl;fs << "Camera.k2: " << 0.0 << std::endl;fs << "Camera.p1: " << 0.0 << std::endl;fs << "Camera.p2: " << 0.0 << std::endl;fs << std::endl;fs << "Camera.width: " << leftWidth << std::endl;fs << "Camera.height: " << leftHeight << std::endl;fs << std::endl;fs << "#Camera frames per second" << std::endl;fs << "Camera.fps: " << 20 << std::endl;fs << std::endl;fs << "#stereo baseline × Camera.fx" << std::endl;float bf = std::fabs(trans.ptr<double>(0)[0] * leftP.ptr<double>(0)[0]);fs << "Camera.bf: " << bf << std::endl; fs << std::endl;fs << "# Color order of image (0 BGR, 1 RGB, It is ignored if images are grayscale)" << std::endl;fs << "Camera.RGB: " << 1 << std::endl;fs << std::endl;fs << "# Close/Far threshold. Baseline times." << std::endl;fs << "ThDepth: " << 35 << std::endl;fs << std::endl;fs << "#-------------------------------------------" << std::endl;fs << "#Stereo Rectification. Camera.fx, Camera.fy must be the same as in LEFT.P" << std::endl;fs << "#-------------------------------------------" << std::endl;fs << std::endl;fs << "LEFT.height: " << leftHeight << std::endl;fs << "LEFT.width: " << leftWidth << std::endl;fs << "LEFT.D: !!opencv-matrix" << std::endl;fs << "    rows: " << 1 << std::endl;fs << "    cols: " << 5 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << leftD << std::endl;cv::Mat rowLeftK = leftK.reshape(0, 1);fs << "LEFT.K: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 3 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowLeftK << std::endl;cv::Mat rowLeftR = leftR.reshape(0, 1);fs << "LEFT.R: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 3 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowLeftR << std::endl;cv::Mat rowLeftP = leftP.reshape(0, 1);fs << "LEFT.P: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 4 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowLeftP << std::endl;fs << std::endl;fs << "RIGHT.height: " << rightHeight << std::endl;fs << "RIGHT.width: " << rightWidth << std::endl;fs << "RIGHT.D: !!opencv-matrix" << std::endl;fs << "    rows: " << 1 << std::endl;fs << "    cols: " << 5 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rightD << std::endl;cv::Mat rowRightK = rightK.reshape(0, 1);fs << "RIGHT.K: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 3 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowRightK << std::endl;cv::Mat rowRightR = rightR.reshape(0, 1);fs << "RIGHT.R: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 3 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowRightR << std::endl;cv::Mat rowRightP = rightP.reshape(0, 1);fs << "RIGHT.P: !!opencv-matrix" << std::endl;fs << "    rows: " << 3 << std::endl;fs << "    cols: " << 4 << std::endl;fs << "    dt: " << 'd' << std::endl;fs << "    data: " << rowRightP << std::endl;fs << std::endl;fs << "#-------------------------------------------" << std::endl;fs << "#ORB Parameters" << std::endl;fs << "#-------------------------------------------" << std::endl;fs << std::endl;fs << "ORBextractor.nFeatures: " << 1200 << std::endl;fs << "ORBextractor.scaleFactor: " << 1.2 << std::endl;fs << "ORBextractor.nLevels: " << 8 << std::endl;fs << "ORBextractor.iniThFAST: " << 20 << std::endl;fs << "ORBextractor.minThFAST: " << 7 << std::endl;fs << std::endl;fs << "#-------------------------------------------" << std::endl;fs << "#Viewer Parameters" << std::endl;fs << "#-------------------------------------------" << std::endl;fs << std::endl;fs << "bUseViewer: " << 0 << std::endl;fs << "Viewer.KeyFrameSize: " << 0.05 << std::endl;fs << "Viewer.KeyFrameLineWidth: " << 1 << std::endl;fs << "Viewer.GraphLineWidth: " << 0.9 << std::endl;fs << "Viewer.PointSize: " << 2 << std::endl;fs << "Viewer.CameraSize: " << 0.08 << std::endl;fs << "Viewer.CameraLineWidth: " << 3 << std::endl;fs << "Viewer.ViewpointX: " << 0 << std::endl;fs << "Viewer.ViewpointY: " << -0.7 << std::endl;fs << "Viewer.ViewpointZ: " << -1.8 << std::endl;fs << "Viewer.ViewpointF: " << 500 << std::endl;cv::Mat rowTrl = Trl.reshape(0, 1);fs << "############" << std::endl;fs << "#the extrinsic params from left camera to right camera !" << std::endl;fs << "#Or the extrinsic params from left to the main camera in VIO project!" << std::endl;fs << "LEFT.RIGHT: !!opencv-matrix" << std::endl;fs << "    rows: " << 4 << std::endl;fs << "    cols: " << 4 << std::endl;fs << "    dt: " << 'f' << std::endl;fs << "    data: " << rowTrl << std::endl;fs << std::endl;std::cout << "saved the parameters in " << filename << std::endl;return true;
}cv::Mat RPY2mat(double roll, double pitch, double yaw)
{cv::Mat m(3, 3, CV_64FC1);double cr = std::cos(roll);double sr = std::sin(roll);double cp = std::cos(pitch);double sp = std::sin(pitch);double cy = std::cos(yaw);double sy = std::sin(yaw);m.ptr<double>(0)[0] = cy * cp;m.ptr<double>(0)[1] = cy * sp * sr - sy * cr;m.ptr<double>(0)[2] = cy * sp * cr + sy * sr;m.ptr<double>(1)[0] = sy * cp;m.ptr<double>(1)[1] = sy * sp * sr + cy * cr;m.ptr<double>(1)[2] = sy * sp * cr - cy * sr;m.ptr<double>(2)[0] = - sp;m.ptr<double>(2)[1] = cp * sr;m.ptr<double>(2)[2] = cp * cr;return m.clone();
}void mat2RPY(const cv::Mat &m, double& roll, double& pitch, double& yaw)
{roll = std::atan2(m.ptr<double>(2)[1], m.ptr<double>(2)[2]);pitch = std::atan2(-m.ptr<double>(2)[0], std::sqrt(m.ptr<double>(2)[1] * m.ptr<double>(2)[1] + m.ptr<double>(2)[2] * m.ptr<double>(2)[2]));yaw = std::atan2(m.ptr<double>(1)[0], m.ptr<double>(0)[0]);
}int main(int argc, char *argv[])
{//argv[1]: 左相机图片所在的文件夹的路径//argv[2]: 右相机图片所在的文件夹的路径//要求左右两张图片的名字是一样的std::vector<std::string> imgNamesL, imgNamesR;getImgNames(argv[1], argv[2], imgNamesL, imgNamesR);const int board_w = 7;const int board_h = 5;const int NPoints = board_w * board_h;const float boardSize = 0.034;cv::Size chessBoardSize = cv::Size(board_w, board_h);std::vector<cv::Point3f> object;for( int j = 0; j < NPoints; j++ ){object.push_back(cv::Point3f((j % board_w) * boardSize, (j / board_w) * boardSize, 0));}std::vector<std::vector<cv::Point3f> > objectv;std::vector<std::vector<cv::Point2f> > imagevL, imagevR;int height, width;cv::Mat imageL, grayImageL, imageR, grayImageR;std::vector<cv::Mat> imgsL, imgsR;for( int i = 0; i < (int)imgNamesL.size(); i++ ){imageL = cv::imread(imgNamesL[i]);imageR = cv::imread(imgNamesR[i]);imgsL.push_back(imageL);imgsR.push_back(imageR);if(imageL.empty()){std::cout << imgNamesL[i] << " is empty!" << std::endl;break;}if(imageR.empty()){std::cout << imgNamesR[i] << " is empty!" << std::endl;}height = imageL.rows;width = imageL.cols;if(imageL.channels() == 3){cv::cvtColor(imageL, grayImageL, CV_BGR2GRAY);}if(imageR.channels() == 3){cv::cvtColor(imageR, grayImageR, CV_BGR2GRAY);}IplImage tempgrayL = grayImageL;IplImage tempgrayR = grayImageR;bool findchessboardL = cvCheckChessboard(&tempgrayL, chessBoardSize);bool findchessboardR = cvCheckChessboard(&tempgrayR, chessBoardSize);std::vector<cv::Point2f> tempCornersL, tempCornersR;if(findchessboardL && findchessboardR){bool find_corners_resultL = cv::findChessboardCorners(grayImageL, chessBoardSize, tempCornersL, 3);bool find_corners_resultR = cv::findChessboardCorners(grayImageR, chessBoardSize, tempCornersR, 3);if(find_corners_resultL && find_corners_resultR){std::cout << "Capture " << imgNamesL[i] << std::endl;cv::cornerSubPix(grayImageL, tempCornersL, cv::Size(5, 5), cv::Size(-1, -1), cv::TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));cv::cornerSubPix(grayImageR, tempCornersR, cv::Size(5, 5), cv::Size(-1, -1), cv::TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));cv::drawChessboardCorners(imageL, chessBoardSize, tempCornersL, find_corners_resultL);cv::drawChessboardCorners(imageR, chessBoardSize, tempCornersR, find_corners_resultR);cv::imshow("L", imageL);cv::imshow("R", imageR);cv::waitKey();objectv.push_back(object);imagevL.push_back(tempCornersL);imagevR.push_back(tempCornersR);}else{std::cout << "find_corners_resultL: " << find_corners_resultL << " find_corners_resultR: " << find_corners_resultR << std::endl;cv::imshow("L", imageL);cv::imshow("R", imageR);cv::waitKey();}}else{std::cout << "findchessboardL: " << findchessboardL << " findchessboardR: " << findchessboardR << std::endl;cv::imshow("L", imageL);cv::imshow("R", imageR);cv::waitKey();}tempCornersL.clear();tempCornersR.clear();}//标定左右两个相机的内参std::vector<cv::Mat> rvecsL, tvecsL, rvecsR, tvecsR;cv::Mat intrinsicL, distortionL, intrinsicR, distortionR;//double errL = cv::calibrateCamera(objectv, imagevL, cv::Size(width, height), intrinsicL, distortionL, rvecsL, tvecsL);int flag = 0; flag |= cv::fisheye::CALIB_RECOMPUTE_EXTRINSIC; //flag |= cv::fisheye::CALIB_CHECK_COND; flag |= cv::fisheye::CALIB_FIX_SKEW; //flag |= cv::fisheye::CALIB_USE_INTRINSIC_GUESS;double errL =  cv::fisheye::calibrate(objectv, imagevL, cv::Size(width, height), intrinsicL, distortionL, cv::noArray(), cv::noArray(), flag, cv::TermCriteria(3, 20, 1e-6)); //  cv::Mat mapx, mapy;// cv::Mat corrected;// cv::Size corrected_size(width, height);//  cv::fisheye::initUndistortRectifyMap(intrinsicL, distortionL, cv::Matx33d::eye(), intrinsicL, //  corrected_size, CV_16SC2, mapx, mapy); std::cout << "左相机内参-------------------------" << std::endl;std::cout << "errL: " << errL << std::endl;std::cout << "intrinsicL: " << std::endl;std::cout << intrinsicL << std::endl;std::cout << "distortionL: " << std::endl;std::cout << distortionL << std::endl;std::cout << "右相机内参-------------------------" << std::endl;//double errR = cv::calibrateCamera(objectv, imagevR, cv::Size(width, height), intrinsicR, distortionR, rvecsR, tvecsR);double errR = cv::fisheye::calibrate(objectv, imagevR, cv::Size(width, height), intrinsicR, distortionR, cv::noArray(), cv::noArray(), flag, cv::TermCriteria(3, 20, 1e-6)); std::cout << "errR: " << errR << std::endl;std::cout << "intrinsicR: " << std::endl;std::cout << intrinsicR << std::endl;std::cout << "distortionR: " << std::endl;std::cout << distortionR << std::endl;//标定左右两个相机之间的外参cv::Mat E, F, R, t;cv::Size imgSize(width, height);//double errLR = cv::stereoCalibrate(objectv, imagevL, imagevR, intrinsicL, distortionL, intrinsicR, distortionR, //imgSize, R, t, E, F);double errLR = cv::fisheye::stereoCalibrate(objectv, imagevL, imagevR, intrinsicL, distortionL, intrinsicR, distortionR, imgSize,R, t);std::cout << "外参--------------------------------" << std::endl;std::cout << "errLR: " << errLR << std::endl;std::cout << "R: " << std::endl;std::cout << R << std::endl;std::cout << "t: " << std::endl;std::cout << t << std::endl;//立体校正cv::Mat Rl, Rr, Pl, Pr, Q;//cv::stereoRectify(intrinsicL, distortionL, intrinsicR, distortionR, imgSize, R, t, Rl, Rr, Pl, Pr, Q);cv::fisheye::stereoRectify(intrinsicL, distortionL, intrinsicR, distortionR, imgSize,R, t, Rl, Rr, Pl, Pr, Q,cv::CALIB_ZERO_DISPARITY,imgSize);cv::Mat Trl = cv::Mat::eye(4, 4, CV_64FC1);R.copyTo(Trl.rowRange(0, 3).colRange(0, 3));t.copyTo(Trl.rowRange(0, 3).col(3));std::string fileName = "stereoRectify.yaml";saveCameraParams(fileName, height, width, distortionL, intrinsicL, Rl, Pl, height, width, distortionR, intrinsicR, Rr, Pr, t, Trl);cv::Mat lmapx, lmapy, rmapx, rmapy;//cv::initUndistortRectifyMap(intrinsicL, distortionL, Rl, Pl, imgSize, CV_32F, lmapx, lmapy);//cv::initUndistortRectifyMap(intrinsicR, distortionR, Rr, Pr, imgSize, CV_32F, rmapx, rmapy);cv::fisheye::initUndistortRectifyMap(intrinsicL, distortionL, Rl, Pl, imgSize, CV_32F, lmapx, lmapy);cv::fisheye::initUndistortRectifyMap(intrinsicR, distortionR, Rr, Pr, imgSize, CV_32F, rmapx, rmapy);int num = 0;cv::Mat imLeftRect, imRightRect;while(num < (int)imgsL.size()){cv::Mat imgL = imgsL[num];cv::Mat imgR = imgsR[num];if(imgL.empty() || imgR.empty()){break;}cv::remap(imgL, imLeftRect, lmapx, lmapy, cv::INTER_LINEAR);cv::remap(imgR, imRightRect, rmapx, rmapy, cv::INTER_LINEAR);cv::Mat imgRectShow;imgRectShow.create(height, width * 2, CV_8UC3);imLeftRect.copyTo(imgRectShow.rowRange(0, height).colRange(0, width));imRightRect.copyTo(imgRectShow.rowRange(0, height).colRange(width, 2*width));int allLineNum = 20;for( int lineNum = 0; lineNum < allLineNum; lineNum++ ){cv::Point2f pt1(0, lineNum * 20);cv::Point2f pt2(imgRectShow.cols-1,lineNum * 20);cv::line(imgRectShow, pt1, pt2, cv::Scalar(0, 0, 255));}cv::imshow("imgRectShow", imgRectShow);cv::waitKey(0);num++;}return 0;
}

鱼眼摄像头单目标定双目标定相关推荐

OpenCV | 双目相机标定之OpenCV获取左右相机图像+MATLAB单目标定+双目标定
博主github:https://github.com/MichaelBeechan 博主CSDN:https://blog.csdn.net/u011344545 原本网上可以搜到很多关于双目相机标 ...
OpenCV3+VS2017+单目标定+双目标定+双目测距
理论知识请参考<学习OpenCV中文版>(公式,函数描述方面可能有错误注意一下,还有不要看<学习OpenCV3中文版>,可以看<Learning OpenCV3>英 ...
使用opencv实现单目标定相机标定(摘抄)
使用opencv实现单目标定相机标定的目的:获取摄像机的内参和外参矩阵(同时也会得到每一幅标定图像的选择和平移矩阵),内参和外参系数可以对之后相机拍摄的图像就进行矫正,得到畸变相对很小的图像. 相机 ...
单目标定---从原理到实现（c++）（利用张氏标定法）
文章转自:https://www.cnblogs.com/zyly/p/9366080.html 双目视觉是建立在几何数学的基础上,数学推导是枯燥乏味的.因此这里不去过多的介绍数学原理,只是简要的叙述 ...
相机标定——单目标定和双目标定
相机标定--单目标定和双目标定 1.标定目的在图像测量过程以及机器视觉应用中,为确定空间物体表面某点的三维几何位置与其在图像中对应点之间的相互关系,必须建立相机成像的几何模型,这些几何模型参数就是相 ...
双目标定（一）单目标定与矫正的基本介绍
1.单目相机标定首先,任何标定都是用基于小孔模型的数学模型去近似相机模型,我们需要用fx = f/dx, fy = f/dy,图像坐标系中的光心原点坐标(和可能的缩放因子ks)这5个相机内参数,切向 ...
双目标定（二）单目标定基本原理
主体思路,先处理纯二维平面的畸变问题(此处略过),矫正图片后,再来求解相机内外参数.基本思路是求得每个标定板对应的单应矩阵,再联合优化所有标定板数据得到相机内参矩阵,再得到每个标定板对应的外参. 1. ...
opencv 图像上画出目标运动的轨迹_基于opencv的单目和双目标定平台手眼标定
背景介绍基于机器视觉引导的智能机器人,在现代社会各个领域已经得到了广泛的应用,尤其是大型工厂的自动化生产线上,视觉机器人可以和基于示教器按照预定轨迹进行作业的机器人互为补充,来共同提高生产的自动 ...
结构光三维重建之单目标定的一种方法——建立“相位-像点-真实三维坐标”之间的关系
结构光三维重建之单目标定的一种方法--建立"相位-像点-真实三维坐标"之间的关系 1.目的为了让像我一样刚接触结构光三维扫描的朋友们能更快速地理解整个框架,我先介绍一下写作本文的 ...

鱼眼摄像头单目标定双目标定

鱼眼摄像头单目标定双目标定相关推荐

最新文章

热门文章

鱼眼摄像头 单目标定 双目标定

鱼眼摄像头 单目标定 双目标定相关推荐

最新文章

热门文章

鱼眼摄像头单目标定双目标定

鱼眼摄像头单目标定双目标定相关推荐