ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Revision history [back]

  1. Not sure about this one.

  2. The origin is likely at a center of focus somewhere behind the lens of one of the two cameras. I'm not very familiar with stereo operation but I believe one camera is primary and the other has an offset relative to it (the other possibility is that the origin is midway between the two cameras).

    You have to experimentally figure out where the origin is relative to the physical camera bodies- maybe using a fixture with a calibration target. Check out ROS Industrial extrinsic calibration http://wiki.ros.org/industrial_extrinsic_cal .

  3. The reported distance ought to be consistent with the units provided about the calibration target. If you told calibration meters, you should be measuring in meters.

  4. Try measuring a calibration target (like a chessbaord/checkerboard or circle grid). You can get the distance to the target from a single calibrated camera (either using opencv or maybe there is a chessboard finding ros node available), and compare that to the stereo generated points distance. Both those processes use the same camera calibrations so ought to be consistent with each other even if not consistent with independently measured distances, which would indicate problems elsewhere.

  1. Not sure about this one.

  2. The origin is likely at a center of focus somewhere behind the lens of one of the two cameras. I'm not very familiar with stereo operation but I believe one camera is primary and the other has an offset relative to it (the other possibility is that the origin is midway between the two cameras).

    You have to experimentally figure out where the origin is relative to the physical camera bodies- maybe using a fixture with a calibration target. Check out ROS Industrial extrinsic calibration http://wiki.ros.org/industrial_extrinsic_cal .

  3. The reported distance ought to be consistent with the units provided about the calibration target. If you told calibration meters, you should be measuring in meters.

  4. Try measuring a calibration target (like a chessbaord/checkerboard chessboard/checkerboard or circle grid). You can get the distance to the target from a single calibrated camera (either using opencv or maybe there is a chessboard finding ros node available), and compare that to the stereo generated points distance. Both those processes use the same camera calibrations so ought to be consistent with each other even if not consistent with independently measured distances, which would indicate problems elsewhere.