nmagerko / pose Goto Github PK

make computeRT take an image, and try to estimate the parameters from the single image rather than trying multiple times to get a good one.

To compute the center of the camera,

take (-transpose(R))*T

Having finished the program for calculating the intrinsics (camera matrix) of the Kinect, we should now work on calculating the rotation and translation vectors for a scene every image.

This will require using solvePnP and, as with the intrinsics program, converting everything back over to cv2.

Using the libfreenect (OpenKinect) library of functions, get values of the depth map of the Kinect and convert them into 3D values. Since both the OpenCV programs and the libfreenect programs utilize numpy now, it should be relatively simple to use the results of this program with any work that needs to be done with OpenCV and vice versa.

Reference: http://graphics.stanford.edu/~mdfisher/Kinect.html