Camera Calibration using ArUco Boards Demo

Calibration using a ArUco Planar Grid board.

To capture a frame for calibration, press 'c', If input comes from video, press any key for next frame To finish capturing, press 'ESC' key and calibration starts.

Sources:

Contents

Parameters

% options
vidFile = '';              % Use video file instead of camera as input
markersX = 5;              % Number of markers in X direction
markersY = 7;              % Number of markers in Y direction
markerLength = 60;         % Marker side length (in meters)
markerSeparation = 15;     % Separation between two consecutive markers in the grid (in meters)
dictionaryId = '6x6_250';  % Dictionary id
refindStrategy = true;     % Apply refined strategy
calibrationFlags = {
    'UseIntrinsicGuess',false, ...
    'FixAspectRatio',false, ...     % Fix aspect ratio (fx/fy)
    'ZeroTangentDist',true, ...     % Assume zero tangential distortion
    'FixPrincipalPoint',false       % Fix the principal point at the center
};
aspectRatio = 1;                    % Fix aspect ratio (fx/fy) to this value

% FixAspectRatio in camera parameters
if calibrationFlags{4}
    camMatrix = eye(3);
    camMatrix(1,1) = aspectRatio;
    calibrationFlags = [calibrationFlags, 'CameraMatrix',camMatrix];
end

% marker detector parameters
detectorParams = struct();
if false
    %detectorParams.nMarkers = 1024;
    detectorParams.adaptiveThreshWinSizeMin = 3;
    detectorParams.adaptiveThreshWinSizeMax = 23;
    detectorParams.adaptiveThreshWinSizeStep = 10;
    detectorParams.adaptiveThreshConstant = 7;
    detectorParams.minMarkerPerimeterRate = 0.03;
    detectorParams.maxMarkerPerimeterRate = 4.0;
    detectorParams.polygonalApproxAccuracyRate = 0.05;
    detectorParams.minCornerDistanceRate = 0.05;
    detectorParams.minDistanceToBorder = 3;
    detectorParams.minMarkerDistanceRate = 0.05;
    detectorParams.cornerRefinementMethod = 'None';
    detectorParams.cornerRefinementWinSize = 5;
    detectorParams.cornerRefinementMaxIterations = 30;
    detectorParams.cornerRefinementMinAccuracy = 0.1;
    detectorParams.markerBorderBits = 1;
    detectorParams.perspectiveRemovePixelPerCell = 8;
    detectorParams.perspectiveRemoveIgnoredMarginPerCell = 0.13;
    detectorParams.maxErroneousBitsInBorderRate = 0.04;
    detectorParams.minOtsuStdDev = 5.0;
    detectorParams.errorCorrectionRate = 0.6;
end

% create board
dictionary = {'Predefined', dictionaryId};
board = {'GridBoard', ...
    markersX, markersY, markerLength, markerSeparation, dictionary};

Input source

if ~isempty(vidFile) && exist(vidFile, 'file') == 2
    vid = cv.VideoCapture(vidFile);
    waitTime = 1;     % 1 sec
else
    vid = createVideoCapture([], 'aruco');
    waitTime = 0.01;  % 10 msec
end
if ~vid.isOpened(), error('failed to initialize VideoCapture'); end

Collect

% collect frames for calibration
allCorners = {};
allIds = {};
imgSize = [];
hImg = []; hFig = [];
while true
    % grab frame
    img = vid.read();
    if isempty(img), break; end

    % detect markers
    [corners, ids, rejected] = cv.detectMarkers(img, dictionary, ...
        'DetectorParameters',detectorParams);

    % refined strategy to detect more markers
    if refindStrategy
        [corners, ids, rejected] = cv.refineDetectedMarkers(img, ...
            board, corners, ids, rejected);
    end

    % draw results
    out = img;
    if ~isempty(ids)
        out = cv.drawDetectedMarkers(out, corners, 'IDs',ids);
    end
    out = cv.putText(out, ['Press "c" to add current frame. ', ...
        '"ESC" to finish and calibrate'], [10 20], ...
        'FontScale',0.5, 'Color',[255 0 0], 'Thickness',2);

    if isempty(hImg)
        hImg = imshow(out);
        hFig = ancestor(hImg, 'figure');
        set(hFig, 'KeyPressFcn',@(o,e) setappdata(o, 'key',e.Key));
        setappdata(hFig, 'key','');
    elseif ishghandle(hImg)
        set(hImg, 'CData',out);
    else
        break;
    end
    drawnow; pause(waitTime);

    % collect frame
    switch getappdata(hFig, 'key')
        case {'space', 'return', 'c'}
            if ~isempty(ids)
                fprintf('Frame captured at %s\n', datestr(now()));
                allCorners{end+1} = corners;
                allIds{end+1} = ids;
                imgSize = size(img);
            else
                disp('frame skipped, no corners detected!');
            end
        case {'escape', 'q'}
            disp('Finished collecting frames.');
            break;
        case 'p'
            pause(5);
    end
    setappdata(hFig, 'key','');
end
vid.release();

Frame captured at 03-Dec-2017 20:02:08
Frame captured at 03-Dec-2017 20:02:09
Frame captured at 03-Dec-2017 20:02:10
Frame captured at 03-Dec-2017 20:02:12
Frame captured at 03-Dec-2017 20:02:13
Frame captured at 03-Dec-2017 20:02:15
Frame captured at 03-Dec-2017 20:02:16
Frame captured at 03-Dec-2017 20:02:18
Frame captured at 03-Dec-2017 20:02:19
Frame captured at 03-Dec-2017 20:02:21
Finished collecting frames.

Calibration

disp('Calibrating...')

% calibrate camera
if isempty([allIds{:}]), error('Not enough captures for calibration'); end
[camMatrix, distCoeffs, repError, rvecs, tvecs] = ...
    cv.calibrateCameraAruco([allCorners{:}], [allIds{:}], ...
        cellfun(@numel, allCorners), board, ...
        imgSize([2 1]), calibrationFlags{:});

% calibration results
fprintf('Calibration Time: %s\n', datestr(now()));
fprintf('Image Width: %d, Image Height: %d\n', imgSize(2), imgSize(1));
disp('Flags:'); cellfun(@disp,calibrationFlags);
if calibrationFlags{4}, fprintf('Aspect Ratio: %f\n', aspectRatio); end
disp('Camera Matrix:'); disp(camMatrix)
disp('Distortion Coefficients:'); disp(distCoeffs)
fprintf('Average Reprojection Error = %f\n', repError);
save camera_parameters.mat -mat camMatrix distCoeffs

Calibrating...
Calibration Time: 03-Dec-2017 20:02:23
Image Width: 512, Image Height: 512
Flags:
UseIntrinsicGuess
   0
FixAspectRatio
   0
ZeroTangentDist
   1
FixPrincipalPoint
   0
Camera Matrix:
  469.7718         0  259.0107
         0  469.5324  251.2255
         0         0    1.0000
Distortion Coefficients:
   -0.1007    0.2118         0         0   -0.6476
Average Reprojection Error = 0.856354