1

我正在使用 OpenCV 来校准和校正立体声系统。我有一个眼睛会聚的立体相机，实际上我正在按以下顺序运行这些功能：

for(int j=0; j < ChessBoard.numSquares; j++)
    obj.push_back(Point3f((j/ChessBoard.numCornersHor)*ChessBoard.squareDim, (j%ChessBoard.numCornersHor)*ChessBoard.squareDim, 0.0f));
[...]

然后我循环这个以获得我想要获取的图像的数量

found_L = findChessboardCorners(image_L, ChessBoard.board_sz, corners_L, CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE + CV_CALIB_CB_FILTER_QUADS + CALIB_CB_FAST_CHECK);
found_R= findChessboardCorners(image_R, ChessBoard.board_sz, corners_R, CALIB_CB_ADAPTIVE_THRESH + CALIB_CB_NORMALIZE_IMAGE + CV_CALIB_CB_FILTER_QUADS + CALIB_CB_FAST_CHECK);
found = found_L && found_R;
if(found)
  { 
    cornerSubPix(image_L, corners_L, Size(11, 11), Size(-1, -1), TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));
    cornerSubPix(image_R, corners_R, Size(11, 11), Size(-1, -1), TermCriteria(CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, 30, 0.1));
    drawChessboardCorners(image_L, ChessBoard.board_sz, corners_L, found);
    drawChessboardCorners(image_R, ChessBoard.board_sz, corners_R, found);

    image_points[0].push_back(corners_L);
    image_points[1].push_back(corners_R);
    object_points.push_back(obj);
    printf("Right: coordinates stored\n");
    printf("Left: coordinates stored\n");
  }

在这个块之后，我称之为两个：

cameraMatrix[0] = Mat::eye(3, 3, CV_64F);
cameraMatrix[1] = Mat::eye(3, 3, CV_64F);

calibrateCamera(object_points, image_points[0], imageSize, cameraMatrix[0], distCoeffs[0], rvecs_L, tvecs_L);

calibrateCamera(object_points, image_points[1], imageSize, cameraMatrix[1], distCoeffs[1], rvecs_R, tvecs_R);

进而：

rms = stereoCalibrate(object_points, image_points[0], image_points[1],
                    cameraMatrix[0], distCoeffs[0],
                    cameraMatrix[1], distCoeffs[1],
                    imageSize, R, T, E, F,
                    TermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 100, 1e-5),
                    CV_CALIB_FIX_ASPECT_RATIO+CV_CALIB_FIX_INTRINSIC);

最后：

stereoRectify(cameraMatrix[0], distCoeffs[0],
                  cameraMatrix[1], distCoeffs[1],
                  imageSize, R, T, R1, R2, P1, P2, Q,
                  CALIB_ZERO_DISPARITY, -1, imageSize, &roi1, &roi2 );

initUndistortRectifyMap(cameraMatrix[0], distCoeffs[0], R1, P1, imageSize, CV_16SC2, map11, map12);
initUndistortRectifyMap(cameraMatrix[1], distCoeffs[1], R2, P2, imageSize, CV_16SC2, map21, map22);
remap(imgL, imgL, map11, map12, INTER_LINEAR,BORDER_CONSTANT, Scalar());
remap(imgR, imgR, map21, map22, INTER_LINEAR,BORDER_CONSTANT, Scalar());

这基本上就是我正在做的，但结果非常糟糕，因为图像有一个非常大的黑色区域。这是一个例子：

离开 OpenCV

这是我必须获得的校正图像，直接从相机拍摄：

离开相机

如您所见，图像似乎在右侧平移并剪切，右侧相同，但在左侧平移，结果几乎相同。

那么如何才能获得类似于上一个更好的结果呢？哪里有问题？作为附加数据，我注意到 rms 不太好，大约为 0.4，重投影误差约为 0.2，我知道它们一定要低一些，但我已经尝试了很多次不同的图案、照明和依此类推，在校准中，但我总是采取相同的结果甚至是最差的结果。

4

1 回答 1

1

尝试这样调用stereoRectify：

stereoRectify(cameraMatrix[0], distCoeffs[0],
              cameraMatrix[1], distCoeffs[1],
              imageSize, R, T, R1, R2, P1, P2, Q,
              0, -1, imageSize, &roi1, &roi2 );

即使用0而不是标志CALIB_ZERO_DISPARITY。

此外，为了提高通过 获得的 RMS stereoCalibrate，请尝试使用 flag CV_CALIB_USE_INTRINSIC_GUESS（请参阅此相关答案）：

rms = stereoCalibrate(object_points, image_points[0], image_points[1],
                cameraMatrix[0], distCoeffs[0],
                cameraMatrix[1], distCoeffs[1],
                imageSize, R, T, E, F,
                TermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 100, 1e-5),
                CV_CALIB_USE_INTRINSIC_GUESS+
                    CV_CALIB_FIX_ASPECT_RATIO+CV_CALIB_FIX_INTRINSIC);
于 2014-04-23T14:38:43.187 回答