Distance from camera using sparse 3D reconstruction

So I am doing sparse 3D reconstruction using stereo camera system. So I stereo calibrated the left and right cameras using Caltech's toolbox. Then stereo rectification. Then feature detection and matching. So far it is similar to this example in MATLAB: http://www.mathworks.com/help/vision/examples/sparse-3-d-reconstruction-from-two-views.html

But since I do not have a checkerboard in the scene, I cannot find the extrinsic camera calibration matrix directly. So I had to estimate fundamental matrix, then essential matrix, then camera matrices and finally the 3D point cloud using triangulation.The whole process is similar to this: http://vgl-ait.org/cvwiki/doku.php?id=matlab:tutorial:3d_reconstruction_with_calibrated_image_sequences

Since I only used intrinsic camera calibration matrix, my point cloud is not in metric. It is in pixels.

Now my question is how can I convert this point cloud from pixels to metric?

Answer 1

[Dima Lisin]

Hi Luca,

I would highly recommend you try the Stereo Camera Calibrator app in the Computer Vision System Toolbox. After you calibrate your cameras, you can use the triangulate function to do the sparse reconstruction.