checkerboard_cam_calib.cpp

Go to the documentation of this file.

/* +------------------------------------------------------------------------+
   |                     Mobile Robot Programming Toolkit (MRPT)            |
   |                          http://www.mrpt.org/                          |
   |                                                                        |
   | Copyright (c) 2005-2018, Individual contributors, see AUTHORS file     |
   | See: http://www.mrpt.org/Authors - All rights reserved.                |
   | Released under BSD License. See details in http://www.mrpt.org/License |
   +------------------------------------------------------------------------+ */

#include "vision-precomp.h"  // Precompiled headers

#include <mrpt/system/filesystem.h>
#include <mrpt/config/CConfigFileMemory.h>

#include <mrpt/vision/chessboard_find_corners.h>
#include <mrpt/vision/pinhole.h>
#include <mrpt/vision/chessboard_camera_calib.h>

// Universal include for all versions of OpenCV
#include <mrpt/otherlibs/do_opencv_includes.h>

using namespace mrpt;
using namespace mrpt::vision;
using namespace mrpt::img;
using namespace mrpt::config;
using namespace mrpt::math;
using namespace mrpt::poses;
using namespace std;

/* -------------------------------------------------------
                checkerBoardCameraCalibration
   ------------------------------------------------------- */
bool mrpt::vision::checkerBoardCameraCalibration(
    TCalibrationImageList& images, unsigned int check_size_x,
    unsigned int check_size_y, double check_squares_length_X_meters,
    double check_squares_length_Y_meters, CMatrixDouble33& intrinsicParams,
    std::vector<double>& distortionParams, bool normalize_image,
    double* out_MSE, bool skipDrawDetectedImgs,
    bool useScaramuzzaAlternativeDetector)
{
    // Just a wrapper for the newer version of the function which uses TCamera:
    TCamera cam;
    bool ret = checkerBoardCameraCalibration(
        images, check_size_x, check_size_y, check_squares_length_X_meters,
        check_squares_length_Y_meters, cam, normalize_image, out_MSE,
        skipDrawDetectedImgs, useScaramuzzaAlternativeDetector);

    intrinsicParams = cam.intrinsicParams;
    distortionParams = cam.getDistortionParamsAsVector();
    return ret;
}

#if MRPT_HAS_OPENCV
// JL says:  This was copied here since it seems OpenCV 2.3 had a broken
// <opencv2/core/eigen.hpp> header.
//  It should be removed in the future when 2.3 becomes too old to support.
namespace cv
{
template <
    typename _Tp, int _rows, int _cols, int _options, int _maxRows,
    int _maxCols>
void my_cv2eigen(
    const Mat& src,
    Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst)
{
    CV_DbgAssert(src.rows == _rows && src.cols == _cols);
    if (!(dst.Flags & Eigen::RowMajorBit))
    {
        Mat _dst(
            src.cols, src.rows, DataType<_Tp>::type, dst.data(),
            (size_t)(dst.stride() * sizeof(_Tp)));
        if (src.type() == _dst.type())
            transpose(src, _dst);
        else if (src.cols == src.rows)
        {
            src.convertTo(_dst, _dst.type());
            transpose(_dst, _dst);
        }
        else
            Mat(src.t()).convertTo(_dst, _dst.type());
        CV_DbgAssert(_dst.data == (uchar*)dst.data());
    }
    else
    {
        Mat _dst(
            src.rows, src.cols, DataType<_Tp>::type, dst.data(),
            (size_t)(dst.stride() * sizeof(_Tp)));
        src.convertTo(_dst, _dst.type());
        CV_DbgAssert(_dst.data == (uchar*)dst.data());
    }
}
}  // namespace cv
#endif

/* -------------------------------------------------------
                checkerBoardCameraCalibration
   ------------------------------------------------------- */
bool mrpt::vision::checkerBoardCameraCalibration(
    TCalibrationImageList& images, unsigned int check_size_x,
    unsigned int check_size_y, double check_squares_length_X_meters,
    double check_squares_length_Y_meters, mrpt::img::TCamera& out_camera_params,
    bool normalize_image, double* out_MSE, bool skipDrawDetectedImgs,
    bool useScaramuzzaAlternativeDetector)
{
    MRPT_UNUSED_PARAM(skipDrawDetectedImgs);
#if MRPT_HAS_OPENCV
    try
    {
        ASSERT_(check_size_x > 2);
        ASSERT_(check_size_y > 2);
        ASSERT_(check_squares_length_X_meters > 0);
        ASSERT_(check_squares_length_Y_meters > 0);

        if (images.size() < 1)
        {
            std::cout << "ERROR: No input images." << std::endl;
            return false;
        }

        const unsigned CORNERS_COUNT = check_size_x * check_size_y;
        const CvSize check_size = cvSize(check_size_x, check_size_y);

        // Fill the pattern of expected pattern points only once out of the
        // loop:
        vector<cv::Point3f> pattern_obj_points(CORNERS_COUNT);
        {
            unsigned int y, k;
            for (y = 0, k = 0; y < check_size_y; y++)
            {
                for (unsigned int x = 0; x < check_size_x; x++, k++)
                {
                    pattern_obj_points[k].x = -check_squares_length_X_meters *
                                              x;  // The "-" is for convenience,
                    // so the camera poses appear
                    // with Z>0
                    pattern_obj_points[k].y = check_squares_length_Y_meters * y;
                    pattern_obj_points[k].z = 0;
                }
            }
        }

        // First: Assure all images are loaded:
        // -------------------------------------------
        TCalibrationImageList::iterator it;
        for (it = images.begin(); it != images.end(); ++it)
        {
            TImageCalibData& dat = it->second;

            dat.projectedPoints_distorted.clear();  // Clear reprojected points.
            dat.projectedPoints_undistorted.clear();

            // Skip if images are marked as "externalStorage":
            if (!dat.img_original.isExternallyStored() &&
                !mrpt::system::extractFileExtension(it->first).empty())
            {
                if (!dat.img_original.loadFromFile(it->first))
                    THROW_EXCEPTION_FMT(
                        "Error reading image: %s", it->first.c_str());

                dat.img_checkboard = dat.img_original;
                dat.img_rectified = dat.img_original;
            }
        }

        // For each image, find checkerboard corners:
        // -----------------------------------------------
        vector<vector<cv::Point3f>>
            objectPoints;  // final container for detected stuff
        vector<vector<cv::Point2f>>
            imagePoints;  // final container for detected stuff

        unsigned int valid_detected_imgs = 0;
        vector<string> pointsIdx2imageFile;
        cv::Size imgSize(0, 0);

        unsigned int i;
        for (i = 0, it = images.begin(); it != images.end(); it++, i++)
        {
            TImageCalibData& dat = it->second;

            // Make grayscale version:
            const CImage img_gray(
                dat.img_original, FAST_REF_OR_CONVERT_TO_GRAY);

            if (!i)
            {
                imgSize = cv::Size(img_gray.getWidth(), img_gray.getHeight());
                out_camera_params.ncols = imgSize.width;
                out_camera_params.nrows = imgSize.height;
            }
            else
            {
                if (imgSize.height != (int)img_gray.getHeight() ||
                    imgSize.width != (int)img_gray.getWidth())
                {
                    std::cout << "ERROR: All the images must have the same size"
                              << std::endl;
                    return false;
                }
            }

            // Try with expanded versions of the image if it fails to detect the
            // checkerboard:
            unsigned corners_count;
            bool corners_found = false;

            corners_count = CORNERS_COUNT;

            vector<cv::Point2f> this_img_pts(
                CORNERS_COUNT);  // Temporary buffer for points, to be added if
            // the points pass the checks.

            dat.detected_corners.clear();

            // Do detection (this includes the "refine corners" with
            // cvFindCornerSubPix):
            vector<TPixelCoordf> detectedCoords;
            corners_found = mrpt::vision::findChessboardCorners(
                img_gray, detectedCoords, check_size_x, check_size_y,
                normalize_image,  // normalize_image
                useScaramuzzaAlternativeDetector);

            corners_count = detectedCoords.size();

            // Copy the data into the overall array of coords:
            ASSERT_(detectedCoords.size() <= CORNERS_COUNT);
            for (size_t p = 0; p < detectedCoords.size(); p++)
            {
                this_img_pts[p].x = detectedCoords[p].x;
                this_img_pts[p].y = detectedCoords[p].y;
            }

            if (corners_found && corners_count != CORNERS_COUNT)
                corners_found = false;

            cout << format(
                "Img %s: %s\n",
                mrpt::system::extractFileName(it->first).c_str(),
                corners_found ? "DETECTED" : "NOT DETECTED");

            if (corners_found)
            {
                // save the corners in the data structure:
                int x, y;
                unsigned int k;
                for (y = 0, k = 0; y < check_size.height; y++)
                    for (x = 0; x < check_size.width; x++, k++)
                        dat.detected_corners.push_back(mrpt::img::TPixelCoordf(
                            this_img_pts[k].x, this_img_pts[k].y));

                // Draw the checkerboard in the corresponding image:
                // ----------------------------------------------------
                if (!dat.img_original.isExternallyStored())
                {
                    const int r = 4;
                    CvPoint prev_pt = cvPoint(0, 0);
                    const int line_max = 8;
                    CvScalar line_colors[8];

                    line_colors[0] = CV_RGB(255, 0, 0);
                    line_colors[1] = CV_RGB(255, 128, 0);
                    line_colors[2] = CV_RGB(255, 128, 0);
                    line_colors[3] = CV_RGB(200, 200, 0);
                    line_colors[4] = CV_RGB(0, 255, 0);
                    line_colors[5] = CV_RGB(0, 200, 200);
                    line_colors[6] = CV_RGB(0, 0, 255);
                    line_colors[7] = CV_RGB(255, 0, 255);

                    // Checkboad as color image:
                    dat.img_original.colorImage(dat.img_checkboard);

                    void* rgb_img = dat.img_checkboard.getAs<IplImage>();

                    for (y = 0, k = 0; y < check_size.height; y++)
                    {
                        CvScalar color = line_colors[y % line_max];
                        for (x = 0; x < check_size.width; x++, k++)
                        {
                            CvPoint pt;
                            pt.x = cvRound(this_img_pts[k].x);
                            pt.y = cvRound(this_img_pts[k].y);

                            if (k != 0) cvLine(rgb_img, prev_pt, pt, color);

                            cvLine(
                                rgb_img, cvPoint(pt.x - r, pt.y - r),
                                cvPoint(pt.x + r, pt.y + r), color);
                            cvLine(
                                rgb_img, cvPoint(pt.x - r, pt.y + r),
                                cvPoint(pt.x + r, pt.y - r), color);
                            cvCircle(rgb_img, pt, r + 1, color);
                            prev_pt = pt;
                        }
                    }
                }

                // Accept this image as good:
                pointsIdx2imageFile.push_back(it->first);
                imagePoints.push_back(this_img_pts);
                objectPoints.push_back(pattern_obj_points);

                valid_detected_imgs++;
            }

        }  // end find corners

        std::cout << valid_detected_imgs << " valid images." << std::endl;
        if (!valid_detected_imgs)
        {
            std::cout << "ERROR: No valid images. Perhaps the checkerboard "
                         "size is incorrect?"
                      << std::endl;
            return false;
        }

        // ---------------------------------------------
        // Calculate the camera parameters
        // ---------------------------------------------
        // Calibrate camera
        cv::Mat cameraMatrix, distCoeffs(1, 5, CV_64F, cv::Scalar::all(0));
        vector<cv::Mat> rvecs, tvecs;

        const double cv_calib_err = cv::calibrateCamera(
            objectPoints, imagePoints, imgSize, cameraMatrix, distCoeffs, rvecs,
            tvecs, 0 /*flags*/);

        // Load matrix:
        out_camera_params.intrinsicParams =
            CMatrixDouble33(cameraMatrix.ptr<double>());

        out_camera_params.dist.fill(0);
        for (int k = 0; k < 5; k++)
            out_camera_params.dist[k] = distCoeffs.ptr<double>()[k];

        // Load camera poses:
        for (i = 0; i < valid_detected_imgs; i++)
        {
            CMatrixDouble44 HM;
            HM.zeros();
            HM(3, 3) = 1;

            {
                // Convert rotation vectors -> rot matrices:
                cv::Mat cv_rot;
                cv::Rodrigues(rvecs[i], cv_rot);

                Eigen::Matrix3d rot;
                cv::my_cv2eigen(cv_rot, rot);
                HM.block<3, 3>(0, 0) = rot;
            }

            {
                Eigen::Matrix<double, 3, 1> trans;
                cv::my_cv2eigen(tvecs[i], trans);
                HM.block<3, 1>(0, 3) = trans;
            }

            CPose3D p = CPose3D(0, 0, 0) - CPose3D(HM);

            images[pointsIdx2imageFile[i]].reconstructed_camera_pose = p;

            std::cout << "Img: "
                      << mrpt::system::extractFileName(pointsIdx2imageFile[i])
                      << ": " << p << std::endl;
        }

        {
            CConfigFileMemory cfg;
            out_camera_params.saveToConfigFile("CAMERA_PARAMS", cfg);
            std::cout << cfg.getContent() << std::endl;
        }

        // ----------------------------------------
        // Undistort images:
        // ----------------------------------------
        for (it = images.begin(); it != images.end(); ++it)
        {
            TImageCalibData& dat = it->second;
            if (!dat.img_original.isExternallyStored())
                dat.img_original.rectifyImage(
                    dat.img_rectified, out_camera_params);
        }  // end undistort

        // -----------------------------------------------
        // Reproject points to measure the fit sqr error
        // -----------------------------------------------
        double sqrErr = 0;

        for (i = 0; i < valid_detected_imgs; i++)
        {
            TImageCalibData& dat = images[pointsIdx2imageFile[i]];
            if (dat.detected_corners.size() != CORNERS_COUNT) continue;

            // Reproject all the points into pixel coordinates:
            // -----------------------------------------------------
            vector<TPoint3D> lstPatternPoints(
                CORNERS_COUNT);  // Points as seen from the camera:
            for (unsigned int p = 0; p < CORNERS_COUNT; p++)
                lstPatternPoints[p] = TPoint3D(
                    pattern_obj_points[p].x, pattern_obj_points[p].y,
                    pattern_obj_points[p].z);

            vector<TPixelCoordf>& projectedPoints =
                dat.projectedPoints_undistorted;
            vector<TPixelCoordf>& projectedPoints_distorted =
                dat.projectedPoints_distorted;

            vision::pinhole::projectPoints_no_distortion(
                lstPatternPoints,  // Input points
                dat.reconstructed_camera_pose,
                out_camera_params.intrinsicParams,  // calib matrix
                projectedPoints  // Output points in pixels
            );

            vision::pinhole::projectPoints_with_distortion(
                lstPatternPoints,  // Input points
                dat.reconstructed_camera_pose,
                out_camera_params.intrinsicParams,  // calib matrix
                out_camera_params.getDistortionParamsAsVector(),
                projectedPoints_distorted  // Output points in pixels
            );

            ASSERT_(projectedPoints.size() == CORNERS_COUNT);
            ASSERT_(projectedPoints_distorted.size() == CORNERS_COUNT);

            for (unsigned int p = 0; p < CORNERS_COUNT; p++)
            {
                const double px = projectedPoints[p].x;
                const double py = projectedPoints[p].y;

                const double px_d = projectedPoints_distorted[p].x;
                const double py_d = projectedPoints_distorted[p].y;

                // Only draw if the img is NOT external:
                if (!dat.img_original.isExternallyStored())
                {
                    if (px >= 0 && px < imgSize.width && py >= 0 &&
                        py < imgSize.height)
                        cvCircle(
                            dat.img_rectified.getAs<IplImage>(),
                            cvPoint(px, py), 4, CV_RGB(0, 0, 255));
                }

                // Accumulate error:
                sqrErr +=
                    square(px_d - dat.detected_corners[p].x) +
                    square(py_d - dat.detected_corners[p].y);  // Error relative
                // to the
                // original
                // (distorted)
                // image.
            }
        }

        if (valid_detected_imgs)
        {
            sqrErr /= CORNERS_COUNT * valid_detected_imgs;
            std::cout << "Average err. of reprojection: " << sqrt(sqrErr)
                      << " pixels (OpenCV error=" << cv_calib_err << ")\n";
        }
        if (out_MSE) *out_MSE = sqrt(sqrErr);

        return true;
    }
    catch (std::exception& e)
    {
        std::cout << e.what() << std::endl;
        return false;
    }
#else
    THROW_EXCEPTION("Function not available: MRPT was compiled without OpenCV");
#endif
}