projecting points using calibration file

[amaanda]

Hello! I need some help in finding in the reprojection error.

I'm using a charuco board with opencv to calibrate my zed camera. On the same script I am calibrating the camera and generating the .conf file, I can easily get the reprojection error, since it is returned from cv2.calibratecamera function. I am saving the .conf on disk, and then reloading it on a different script to validate the parameters I got.

To do so, I am trying to project points from an image using the calibration file from ZED, and then calculate the reprojection error. Is this possible, or am I only able to calculate in the moment I am calibrating the camera?

To reproject the points, I am trying to use:

for k in range(len(left_all_points)):
    reprojected_points, jac = cv2.projectPoints(left_all_points[k], rvec=rvecs2, tvec=tvecs2,
cameraMatrix=camera_matrix, distCoeffs=dist_coeffs)
    error = cv2.norm(left_all_corners[k], reprojected_points, cv2.NORM_L2) / len(reprojected_points)
    mean_error_left_ += error

To get the correct reprojected_point, I need to use the rotation vector and translation vector for each camera. Are these information available in the .conf file? How can I calculate them?

Any help will be appreciated.

[KaiWU17TUM]

Hi, did you managed to solve this problem? I have used the calibration parameters (intrinsic matrix) to project the point cloud of ZED camera on the left RGB image, however I got quite large reprojection error. But I think since the point cloud is generated by triangulation of two views, the back projection should be very accurate. Could the rectification process of the SDK have effect on this?

[amaanda]

@KaiWU17TUM I don't use the point cloud provided by ZED SDK so I'm not really sure about what value of reprojection error is acceptable.

[github-actions[bot]]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment otherwise it will be automatically closed in 5 days