Regarding opencv calibration of corner of the world coordinates of confusion-CodePudding

Corner of the world coordinates of opencv calibration method of confusion, now found that there are three kinds of calculation methods:

1 \ opencv \ sources \ samples \ CPP in calibration.
the CPP religionFor (int I=0; i For (int j=0; J & lt; BoardSize. Width; J++)
Corners. The push_back (Point3f (float (j * squareSize), float (I * squareSize), 0));

2. In the book "learning OpenCV" chapter 11 cases of 11-1
CV_MAT_ELEM (* object_points, float, I, 0)=j/board_w;
CV_MAT_ELEM (* object_points, float, I, 1)=j % board_w;
CV_MAT_ELEM (* object_points, float, I, 2)=0.0 f;

3. The blog of http://blog.csdn.net/zhouyelihua/article/details/38421377#comments
CV_MAT_ELEM (* cam_object_points, float, I, 0)=(j/cam_board_w) * cam_Dx;
CV_MAT_ELEM (* cam_object_points, float, I, 1)=(j % cam_board_w) * cam_Dy;
CV_MAT_ELEM (* cam_object_points, float, I, 2)=0.0 f;

Question 1: what kind of method is right?
Question 2: why different image corresponding to the corner of the world coordinates are the same?
Question 3: method 1 squareSize refers to the checkerboard width/height of each small grid?
Question 4: method 2 coordinate method of true not understand, can help explain?

CodePudding user response:

1. Three kinds of calculation methods are right, it is the same,
2. Don't know
3. Is dripping
4. The coordinate output look to understand,

CodePudding user response:

Question 2: why different image corresponding to the corner of the world coordinates are the same?

Actual camera calibration is motionless, the calibration plate in the move, but at the time of calibration calculation, you don't know the world coordinates of the corresponding relationship between calibration board,
So can, in turn, assume that the calibration plate is motionless, the camera on the mobile (movement is relative), so that the world coordinates are the same, image coordinates are know anyway. So we can solve the calibration results
Seek out after you can assume a fixed position of the camera, the other of the coordinate system of the camera calibration plate to change camera coordinate system

Can look at the calibration of the matlab toolbox, the comparison of image, and finally the calibration result can Switch to world - centered view or Switch to camera - centered view
http://www.vision.caltech.edu/bouguetj/calib_doc/htmls/example.html

CodePudding user response:

How to photograph by solving the parameters of the pixels of the projection to the world coordinate

CodePudding user response:

This problem can you solve it, I now also is very confused

CodePudding user response:

Said the very image of the 2 floor

CodePudding user response:

refer to the second floor SSBQRM response:

question 2: why different image corresponding to the corner of the world coordinates are the same?

Actual camera calibration is motionless, the calibration plate in the move, but at the time of calibration calculation, you don't know the world coordinates of the corresponding relationship between calibration board,
So can, in turn, assume that the calibration plate is motionless, the camera on the mobile (movement is relative), so that the world coordinates are the same, image coordinates are know anyway. So we can solve the calibration results
Seek out after you can assume a fixed position of the camera, the other of the coordinate system of the camera calibration plate to change camera coordinate system

Can look at the calibration of the matlab toolbox, the comparison of image, and finally the calibration result can Switch to world - centered view or Switch to camera - centered view
http://www.vision.caltech.edu/bouguetj/calib_doc/htmls/example.html