3

Can someone please tell me how the projection Matrices look like for triangulatePoints? Its not that easy, I found several possible projection Matrices but couldn't figure out by now which one is the right one.

If I have KeyPoints from Robust Matcher/or other matchingmethods then I got rotation and translation from E. This would cause matrices like P0=[I|T] and P1=[R|T] where R and T are the extracted, not changed values. Am I right? When I have a stereo Rig where using calibrateCamera() (explicite not stereoCalibrate!!) or using SolvePnP I get R and T from camera1 and camera2. Does those matrices look like this?: P0=[R0|t0] and P1=[R1|t1] or P0=[R0|R0*t0] and P1=[R1|R1*t1] or is it something else?

I found something about "rectify" do you know if I have too rectify bevore getting the KeyPoints or is undistortPoints the only needed function which is needed for tiangulatePoints()?

Thanks for help

user1651460
  • 450
  • 1
  • 5
  • 16

1 Answers1

0

we have these in 3D reconstruction: 1-image coordinates 2-camera cordinates 3-world coordinates

we suppose one of them for example world coordinates and put other things like camera in that coordinates.

we have 2 important matrix 1-view matrix or projection matrix that describe the projection 2-model matrix that describe the position an rotation of the object

by multiplying two mentioned matrix we have modelview matrix that describe the scene opengl is a good library and open examples you can get help of this link: http://www.songho.ca/opengl/gl_transform.html

ABSSH
  • 110
  • 5
  • Hi AbSSH thanks for your answer but its more about how the projection matrix is build for the openCV function triangulatePoints(). I build it like mentioned but it didn't worked. Whats why i wanted to know how it has to be created. I know wow generaly a projectionmatrix is build but it doesn't fit into this function. – user1651460 Jan 29 '13 at 10:16