Ask Your Question
0

How to perform a good calibration for the Kinect?

asked 2013-03-06 01:57:37 -0500

anto1389 gravatar image

updated 2016-10-24 09:03:30 -0500

ngrennan gravatar image

Hello,

I am doing an app to merge point clouds. I am using kinect but i need to improve accuracy because I think my program is working (using Iterative closest point and detecting features such as SURF) but I don't get good results because the RGB camera transformation is not good enough to get "perfect point clouds" (I mean that i can see parts of objects at places where they shouldn't be, like seeing the edge of a box in the wall) This is the problem that make my results bad.

I have performed several calibration with the camera_calibration package in Ros fuerte to solve the problem but i get very different results and the calibration is done in the same way from each calibration. Some results:

[532.486920156757, 0, 313.943582352682, 0, 532.067717610884, 259.017556441565, 0, 0, 1]

[512.878303, 0 , 310.799036, 0, 513.067828 256.711335, 0, 0, 1]

In my opinion this results makes me think that i am doing something wrong in the calibration process but i follow exactly the tutorial http://www.ros.org/wiki/openni_launch/Tutorials/IntrinsicCalibration?action=show&redirect=openni_camera%2Fcalibration so i don't know what i am doing wrong.

Besides, there is some problems with the openni_launch as you can see at this post: http://answers.ros.org/question/53706/registered-depth-image-is-black/ I solved the problem but now i am wondering if the variable depth_regitration should be on or off.

Does anyone know what this hapens to me? could you give me any advice?

Thanks you very much! Antonio.

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted
2

answered 2013-03-06 05:21:35 -0500

updated 2013-03-06 05:22:02 -0500

You can use the cameracheck.py utility to assess the quality of your calibration. Unfortunately, that utility isn't documented very well.

edit flag offensive delete link more
0

answered 2013-03-06 22:15:48 -0500

anto1389 gravatar image

updated 2013-03-06 22:16:20 -0500

Thank you very much!

I did what you suggest and i the highest values that i get are the followings:

Linearity RMS Error: 0.205 Pixels Reprojection RMS Error: 0.137 Pixels for RGB

Linearity RMS Error: 0.437 Pixels Reprojection RMS Error: 0.327 Pixels for IR

I think this values are good enough, right? And one more question, i did the calibration of the IR camera with the calibration file of the RGB camera loaded. Is it the right way to do it or i should calibrate the IR camera whithout loading the RGB calibration file?

Thanks again!

edit flag offensive delete link more

Comments

1

In the future, please edit your original question with any updates rather than posting them as a new answer. Yes, those error values look good. For calibrating the IR camera, it shouldn't matter if the RGB calibration is loaded.

Dan Lazewatsky gravatar image Dan Lazewatsky  ( 2013-03-07 02:35:52 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

3 followers

Stats

Asked: 2013-03-06 01:57:37 -0500

Seen: 2,393 times

Last updated: Mar 06 '13