I refer to your method of using
success, rotation_vector, translation_vector, inliers = cv2.solvePnPRansac(points3D,points2D,camera_matrix,dist_coeffs,flags=cv2.SOLVEPNP_ITERATIVE
I know that this method was originally for the correction between camera image frames
But when clicking 3d(velodyne coordinate) and 2d(pixel coordinate) points, the four points are all on the same plane
So this method may be ok
In this paper : https://arxiv.org/abs/2003.04260 in Fig.4
They have 100 points(velodyne coordinate) and use RANSAC to get inlier points
And they say they use IPPE method to find extrinsic between lidar and camera
I pick 3D points inlier and corresponding 2d point
Why I use success, rotation_vector, translation_vector, inliers = cv2.solvePnPRansac(points3D,points2D,camera_matrix,dist_coeffs,flags=cv2.SOLVEPNP_IPPE return wrong so far
can you give mesome advice how they do that ?
Thanks!
I refer to your method of using
success, rotation_vector, translation_vector, inliers = cv2.solvePnPRansac(points3D,points2D,camera_matrix,dist_coeffs,flags=cv2.SOLVEPNP_ITERATIVE
I know that this method was originally for the correction between camera image frames But when clicking 3d(velodyne coordinate) and 2d(pixel coordinate) points, the four points are all on the same plane So this method may be ok
In this paper : https://arxiv.org/abs/2003.04260 in Fig.4 They have 100 points(velodyne coordinate) and use RANSAC to get inlier points And they say they use IPPE method to find extrinsic between lidar and camera I pick 3D points inlier and corresponding 2d point Why I use
success, rotation_vector, translation_vector, inliers = cv2.solvePnPRansac(points3D,points2D,camera_matrix,dist_coeffs,flags=cv2.SOLVEPNP_IPPE
return wrong so farcan you give mesome advice how they do that ? Thanks!