如何从openCV获得与triangulatePoints()的良好协调

时间:2014-06-09 21:15:18

标签: c++ opencv

所以我有一个点,x:1168px,y:1140px在左图像上,x:1062px,y:1167px在右图像上。我用我的程序得到的真实坐标是:120.91,4.94059,-175.609(x,y和z,以mm为单位)。 X和Y似乎是正确的但不是Z,这个点距离我的相机只有1米。

现在我有了这段代码:

FileStorage fsI("intrinsics.yml", FileStorage::READ);
Mat matI1, matI2, matD1, matD2;
fsI["M1"] >> matI1;
fsI["M2"] >> matI2;
fsI["D1"] >> matD1;
fsI["D2"] >> matD2;

CvMat camInt1 = matI1, camInt2 = matI2, camDist1 = matD1, camDist2 = matD2;

FileStorage fs("extrinsics.yml", FileStorage::READ);
Mat mat1, mat2;
fs["P1"] >> mat1;
fs["P2"] >> mat2;

CvMat projMat1 = mat1, projMat2 = mat2;

double pointImg1_a[2] = { 1168, 1140};
Mat pointImg1 = Mat(2,1, CV_64FC1, pointImg1_a);
CvMat _pointImg1 = cvMat(1,1,CV_64FC2,pointImg1_a);
double pointImg2_a[2] = { 1062, 1167};
Mat pointImg2 = Mat(2,1, CV_64FC1, pointImg2_a);
CvMat _pointImg2 = cvMat(1,1,CV_64FC2,pointImg1_a);

cvUndistortPoints(&_pointImg1,&_pointImg1,&camInt1,&camDist1);
cvUndistortPoints(&_pointImg2,&_pointImg2,&camInt2,&camDist2);

Mat point4D = Mat(4,1, CV_64FC1);
cv::triangulatePoints(mat1, mat2, pointImg1, pointImg2, point4D);   

double w = point4D.at<double>(3,0);
double x = point4D.at<double>(0,0)/w;
double y = point4D.at<double>(1,0)/w;
double z = point4D.at<double>(2,0)/w;

cout << x << ", " << y << ", " << z << endl;

return 0;

我试过没有使用cvUndistortPoints()和其他坐标,结果有时是正确但不总是。 我真的不知道从哪里来的错误。 (我也是openCV的新手) 谢谢你的帮助。

1 个答案:

答案 0 :(得分:1)

你不要犯错误: &#34; CvMat _pointImg2 = cvMat(1,1,CV_64FC2,pointImg1_a);&#34;

正确应该是: &#34; CvMat _pointImg2 = cvMat(1,1,CV_64FC2, pointImg2_a );&#34;