Strange LiDAR intrinsics

[blackccpie]

Hi,

I have a strange behavior with my iPad Pro when inspecting the intrinsics of LiDAR frames I receive through USB streaming: the fx, fy, tx & ty parameters are not constant over time, they change depending on the scene... which I find quite strange. Is that a normal behavior? Also, the frame resolution is 192x256, and the tx/ty values are around 366 and 476 respectively, which is also strange because the should not exceed frame resolution. Am I missing something?

In comparison, with FaceID camera, intrinsincs are consistent compared to frame resolution, and constant over time.

Thanks for your help,

Regards

Albert

PS: I'm using C++ Record3DStream class.

[marek-simonik]

Hi Albert,

thank you for reporting this bug, it has been fixed now — please update to Record3D 1.6 or newer and download the newest version of the record3d library.

In the newest update of Record3D and the record3d library, the provided intrinsic parameters are the intrinsic parameters of the RGB image. In case you would stream with the "Higher-quality LiDAR recording" option enabled (turn on/off in the Settings tab), you will need to either:

• scale-up the corresponding depth map to match the RGB image resolution (because the RGB image has higher resolution than the depth map) and use the provided intrinsic parameters
• or scale the RGB intrinsic parameters to arrive at the intrinsic parameters of the depth image (depthIntrParams = (depthMap.width / rgbImg.width) * rgbIntrParams)

Let me know in case it would not work for you.

Best regards, Marek

[blackccpie]

Hi Marek,

Thanks for the new release, everything is working fine now. As discussed with you, the non-constant intrinsics values may be related to camera autofocus. Do you think autofocus enable/disable could be an option included in the record3d app?

Regards

Albert

[marek-simonik]

Hi Albert,

yes, in a future update, I will add a switch into the Settings section, which will allow you to toggle autofocus.

Best regards, Marek

[saurabheights]

@marek-simonik As per your comment above, shouldnt documentation for get_intrinsic_mat be "Returns the intrinsic matrix of current rgb frame." instead of "Returns the intrinsic matrix of current Depth frame."

See https://github.com/marek-simonik/record3d/blob/master/python-bindings/src/PythonBindings.cpp#L46