opencv读取使用MATLAB双目标定的结果进行双目校正

概述：

opencv内部也有双目标定的函数，但是结果有时候会飘，所以想先用MATLAB标定工具箱试试看。使用MATLAB先队两个相机进行单目标定，然后进行双目标定，将结果保存在xml文件中。xml文件时opencv支持的文件格式，更详细的解释可以去搜索相关资料。

1.工具准备：

MATLAB标定工具箱。

这一步我认为应该单独强调一下。我的MATLAB版本是2014a，应用程序里面自带一个标定工具箱但是只支持单目。后来在http://www.vision.caltech.edu/bouguetj/calib_doc/#examples了解到应该再下载一个双目标定的工具箱。下载链接

将其解压到Matlab的工作目录下，然后将其包含到matlab的工作路径下。运行calib_gui就可以打开matlab标定工具箱主界面：

2.标定过程

1.点击Image names会出现当前MATLAB路径下的所有文件，如上图，输入L表示对L开头的图片进行标定，然后输入后缀b表示图片格式是bmp格式的。



2.点击Extract grid corners，检测角点，这个过程比较麻烦，一路回车下来，会依次弹出各张图片，需要用鼠标顺时针或逆时针选定最外层的角点。选择完第一张的时候会要求输入棋盘格的宽度，我的棋盘格长宽都是15mm。





之后就是对所有L开头的图片选取四个最外层角点，这一步完成后，点击Calibration进行标定，点击Save保存结果Calib_Result.mat，将Calib_Results.mat修改为Calib_Results_left.mat方便进行双目标定。

右相机同理。

3.双目标定

在MATLAB控制台输入stereo_gui，打开双目标定工具箱，点击Load left and right calibration files，再点击stereo calibration进行标定。得到如下结果：

3.保存为xml文件

Distortion_Camera_Left.xml

<?xml version="1.0"?>
<opencv_storage>
<Distortion type_id="opencv-matrix">
<rows>5</rows>
<cols>1</cols>
<dt>d</dt>
<data>
-0.15761   4.24565   -0.00236   0.02423  0.00000
</data>
</Distortion>
</opencv_storage>

Distortion_Camera_Right.xml

<?xml version="1.0"?>
<opencv_storage>
<Distortion type_id="opencv-matrix">
<rows>5</rows>
<cols>1</cols>
<dt>d</dt>
<data>
-0.09281   1.60263   0.00408   -0.00825  0.00000
</data>
</Distortion>
</opencv_storage>

Intrinsics_Camera_Left.xml

<?xml version="1.0"?>
<opencv_storage>
<Intrinsics_Camera_Left type_id="opencv-matrix">
<rows>3</rows>
<cols>3</cols>
<dt>d</dt>
<data>
2752.23366  0.      732.90263
0.          2786.18468  529.17431
0.          0.         1.
</data>
</Intrinsics_Camera_Left>
</opencv_storage>

Intrinsics_Camera_Right.xml

<?xml version="1.0"?>
<opencv_storage>
<Intrinsics_Camera_Right type_id="opencv-matrix">
<rows>3</rows>
<cols>3</cols>
<dt>d</dt>
<data>
2667.95661   0.          584.91164
0.          2699.92379   539.37941
0.          0.          1.
</data>
</Intrinsics_Camera_Right>
</opencv_storage>

RotRodrigues.xml

<?xml version="1.0"?>
<opencv_storage>
<RotRodrigues type_id="opencv-matrix">
<rows>3</rows>
<cols>1</cols>
<dt>d</dt>
<data>
0.02812   0.54890  0.00618
</data>
</RotRodrigues>
</opencv_storage>

Translation.xml

<?xml version="1.0"?>
<opencv_storage>
<Translation type_id="opencv-matrix">
<rows>3</rows>
<cols>1</cols>
<dt>d</dt>
<data>
-160.89829   -1.36460  25.91248
</data>
</Translation>
</opencv_storage>

4.opencv读取xml文件，并进行双目校正

//读取内部参数
CvMat *Intrinsics_Camera_Left = (CvMat *)cvLoad("Intrinsics_Camera_Left.xml");
CvMat *Intrinsics_Camera_Right = (CvMat *)cvLoad("Intrinsics_Camera_Right.xml");
CvMat *Distortion_Camera_Left = (CvMat *)cvLoad("Distortion_Camera_Left.xml");
CvMat *Distortion_Camera_Right = (CvMat *)cvLoad("Distortion_Camera_Right.xml");
CvMat *Translation_matlab = (CvMat *)cvLoad("Translation.xml");
CvMat *RotRodrigues_matlab = (CvMat *)cvLoad("RotRodrigues.xml");
CvMat *R_opencv = cvCreateMat(3, 3, CV_64F);
cvRodrigues2(RotRodrigues_matlab, R_opencv);
//创建映射矩阵
IplImage *Left_Mapx = cvCreateImage(cvSize(1280, 960), IPL_DEPTH_32F, 1);
IplImage *Left_Mapy = cvCreateImage(cvSize(1280, 960), IPL_DEPTH_32F, 1);
IplImage *Right_Mapx = cvCreateImage(cvSize(1280, 960), IPL_DEPTH_32F, 1);
IplImage *Right_Mapy = cvCreateImage(cvSize(1280, 960), IPL_DEPTH_32F, 1);
CvMat *Rl = cvCreateMat(3, 3, CV_64F);
CvMat *Rr = cvCreateMat(3, 3, CV_64F);
CvMat *Pl = cvCreateMat(3, 4, CV_64F);
CvMat *Pr = cvCreateMat(3, 4, CV_64F);
//双目校正
cvStereoRectify(Intrinsics_Camera_Left, Intrinsics_Camera_Right,
Distortion_Camera_Left, Distortion_Camera_Right,
cvSize(1280, 960), R_opencv, Translation_matlab,
Rl, Rr, Pl, Pr, 0, 2048, 0);//增加图像缩放，去除死区
//畸变矫正
cvInitUndistortRectifyMap(Intrinsics_Camera_Left, Distortion_Camera_Left, Rl, Pl,
Left_Mapx, Left_Mapy);
cvInitUndistortRectifyMap(Intrinsics_Camera_Right, Distortion_Camera_Right, Rr, Pr,
Right_Mapx, Right_Mapy);

cvRemap(IplimgLeft, img_left_Change, Left_Mapx, Left_Mapy);
cvRemap(IplimgRight, img_right_Change, Right_Mapx, Right_Mapy);

cvLine(img_left_Change, cvPoint(0, 96 * 1), cvPoint(1280 - 1, 96), cvScalar(255, 0, 0));

cvLine(img_left_Change, cvPoint(0, 96 * 2), cvPoint(1280 - 1, 96 * 2), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 3), cvPoint(1280 - 1, 96 * 3), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 4), cvPoint(1280 - 1, 96 * 4), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 5), cvPoint(1280 - 1, 96 * 5), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 6), cvPoint(1280 - 1, 96 * 6), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 7), cvPoint(1280 - 1, 96 * 7), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 8), cvPoint(1280 - 1, 96 * 8), cvScalar(255, 0, 0));
cvLine(img_left_Change, cvPoint(0, 96 * 9), cvPoint(1280 - 1, 96 * 9), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 1), cvPoint(1280 - 1, 96), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 2), cvPoint(1280 - 1, 96 * 2), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 3), cvPoint(1280 - 1, 96 * 3), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 4), cvPoint(1280 - 1, 96 * 4), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 5), cvPoint(1280 - 1, 96 * 5), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 6), cvPoint(1280 - 1, 96 * 6), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 7), cvPoint(1280 - 1, 96 * 7), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 8), cvPoint(1280 - 1, 96 * 8), cvScalar(255, 0, 0));
cvLine(img_right_Change, cvPoint(0, 96 * 9), cvPoint(1280 - 1, 96 * 9), cvScalar(255, 0, 0));

6.效果

如有任何问题，邮件gao2014#nwsuaf.edu.cn

欢迎交流！