search

ankitdhall / lidar_camera_calibration

ROS package to find a rigid-body transformation between a LiDAR and a camera for "LiDAR-Camera Calibration using 3D-3D Point correspondences"
http://arxiv.org/abs/1705.09785
GNU General Public License v3.0
1.48k stars 459 forks source link

Calibration result is not good #75

Open zheng-rong opened 5 years ago

zheng-rong commented 5 years ago

Hi all, I am trying to calibrate a VLP-16 and monocular camera (equipped with a lens with distortion). I followed the steps in the wiki, but the result seems not reasonable. The following picture shows my sensor configuration: _20180524160503

and the result is: plot The position of lidar points looks right. I checked the distance of the closest two points in the center area.

result

By far I haven't figure out the reason for this bad calibration result.

I am wondering if this is caused by the camera intrinsic matrix. I am using a monocular camera with distortion, and I got the K(intrinsic) and D(distortion) by Kalibr, and undistort the images by openCV function. IN the aruco_mapping config file, I filled in the K and D as required, and feed it with mono/image_raw(distorted). But in lidar_camera_calibration, only K is specified. So I use getOptimalCameraMatrix in openCV to get a new_K (I think this new_K is the intrinsic for undistorted camera image), and feed the new_K and mono/image_rect to the lidar_camera_calibration. Is this correct?

Any help will be appreciated. The config files I am using are attached.

Thanks in advance, Zheng

vi_0003.txt config_file.txt lidar_camera_calibration.txt marker_coordinates.txt data bag file: https://drive.google.com/open?id=1MQ9GarLO00LBQpODbdgiaedcEEZmeyzp

zheng-rong commented 5 years ago

Emmm... I got the wrong marker size in the config file. After revising that, the calibration result looks good, from the visualization of lidar points overlaid on the image. The RMSE is 0.0327. But compared with manually measured translation, the estimated translation has a large error. In my configuration, the manual measurement is: [-0.013,-0.072,-0.12] and the estimated result is: [-0.0230762,0.0149305,-0.185994]

What may be the reason for this result? Thanks.

hunkyu commented 5 years ago

Can you send your modified configuration file?

zheng-rong commented 5 years ago

Please see the attached file. Thanks. marker_coordinates_1.txt

hunkyu commented 5 years ago

screenshot from 2019-02-19 14-06-43 I don't know why it is projected?

hunkyu commented 5 years ago

@zheng-rong Can you leave a mailbox and follow up? My email 513459395@qq.com

zlbing commented 5 years ago

@zheng-rong hi, could you tell me how to make the RMSE small, i used you dataset, and can't get right result

chenwinki commented 5 years ago

@hunkyu Could you give me the pdf or picture of markers? Thank you!

Elkulas commented 5 years ago

@hunkyu how to project points to image?

SinghiDivyanshu commented 4 years ago

hi @Elkulas @hunkyu @zheng-rong can you able to project points to image.....i'm stuck after finding rigid body transformation