D435 Uncalibrated? Color camera distortion coefficients are {0, 0, 0, 0, 0}

[idata]

Hi,

The D435 camera seems to be UNCALIBRATED?

Using the rs-sensor-control app, and selecting the color camera at:

4 : Color # 0 (Video Stream: RGB8 1920x1080@ 30Hz)

Principal Point : 956.491, 544.128

Focal Length : 1376.13, 1376.61

Distortion Model : Brown Conrady

Distortion Coefficients : [0,0,0,0,0]

First, is this correct?

Second, I'm expecting non-zero values for the distortion coefficients if this camera is to be considered calibrated. Any idea's what's going on here? The ZR300's would come factory calibrated, is that not the case for the D400 series cameras? Is there a calibration procedure?

Thank you!

PS, note that I did try other color camera configurations, notably "22 : Color # 0 (Video Stream: RGB8 1280x720@ 30Hz)", and still see {0}'s for distortion with the appropriate changes to FOV and PP reflected in the resolution change.

[MartyG · Accepted solution]

If you want to perform your own calibration, this link may be useful to you:

[idata]

OK, for completeness for future readers, I've just tested a D415 sensor, and the distortion coefficients are also {0}.

Thanks, Marty. I also want to note that the impact of the color camera's uncalibrated lens distortion coefficients is limited to the accuracy of the alignment of the color image when projected onto the depth image. It does not affect Z accuracy.

[MWils1]

I see this on the D415 as well, is intrinsic calibration of the color camera not included in the cost of the unit? How do they get a good extrinsic calibration without it?

[jb455]

Maybe the camera undistorts the image in hardware so by the time it gets to the client PC distortion is 0?

[MWils1]

That's unlikely, even so what's the harm in exposing them through the documented call. If the harm would be a user applying un-distortion over an already corrected image then why expose in the first place. Also if they did just that then why publish instruction on how to calibrate if you can never replace the hardware implementation. I am sure they are not that nieve to think their calibration is fool-proof because from experience it isn't especially when made in China.

[MartyG]

Technically, the very small print on the D415's casing says "Made in Thailand".

[jb455]

You can recalibrate using the calibration app Intel made, and the new calibration is stored on the device and used going forward (including distortion I would assume).

I guess maybe some streams the camera serves may be raw distorted images (or the ability may be added in the future) so the distortion parameters are in the struct and aren't always zero, but if the camera does undistortion for other streams it sets those values to zero because the distortion has already been dealt with.

[idata]

I do believe that you're correct that the color stream has been rectified and {0}'s make sense for the distortion coefficients as such. Without calibration, the principal point would be at the image center, which it is not.

It was nevertheless a startling change from the ZR300, and one that sent me looking for a what at first glance appeared to be a bug. So, it's good to note these things in the forum.

[NPurv]

Hi guys,

We are seeing the distortion coefficients on the depth stream on 415 and 435's all 0's, is this because the depth streams are already rectified in hardware or something else is up?

Thanks,

Neil

[idata]

Hello neil_scatter,

The distortion coefficients on the depth stream are suppose to be 0 on D400 camera.You can find more about this here: https://github.com/IntelRealSense/librealsense/issues/1430Thank you,Eliza