topic Re: Wrong data on RGB 3D on the borders in Items with no label

Wrong data on RGB 3D on the borders

ECosn1 — Fri, 05 Oct 2018 12:11:27 GMT

Hello,

I am using Intel RealSense D435, and when I take picture I always get wrong signal on the border of the image, like the image I attached you to this case, in the part I surrounded in blue.

The camera thinks that these parts are very deep, and the resulting point cloud is pretty bad, and I always need to clean manually these parts of the resulting ply.

I there a way to avoid this?

For me, as I know that what I want to capture is nearer than 1m, a solution would be to create a ply file with only the signal considered as closer to 1m, that is do say specify that the max depth is 1m. Is it possible?

Thank you,

Emmanuel

Re: Wrong data on RGB 3D on the borders

MartyG — Fri, 05 Oct 2018 12:30:00 GMT

Clamp MaxZ may suit your needs if you only want depth values under a certain MaxZ value to be considered.

https://github.com/IntelRealSense/librealsense/issues/2199 Set depth clamp min/max from C++ and python code · Issue # 2199 · IntelRealSense/librealsense · GitHub

Re: Wrong data on RGB 3D on the borders

ECosn1 — Fri, 05 Oct 2018 12:32:38 GMT

Ok thank you, that should fix the problem.

But do you know if there is another way to avoid this phenomenon?

Re: Wrong data on RGB 3D on the borders

MartyG — Fri, 05 Oct 2018 12:52:44 GMT

If the problem is with the area behind the running man (which looks like a dragon using a computer), this may be related to the D435's 'RMS error' factor. The greater the distance of the depth being analyzed, the greater the depth noise, or RMS error, is. The D435 has more noise over distance than the D415 has, due to the hardware design of the D435.

You could try changing the Disparity Shift setting. As Disparity Shift is increased, the MaxZ reduces. So this may be a good way for you to set a maximum distance. Left-click on the images below to see them in full size.