topic Correct Distance for test Subpixel RMS error? in Items with no label

Correct Distance for test Subpixel RMS error?

vhanded — Mon, 17 Jun 2019 06:56:16 GMT

From https://realsense.intel.com/wp-content/uploads/sites/63/BKM-For-Tuning-D435-and-D415-Cameras-Webinar_Rev3.pdf, we expect to see Subpixel RMS < 0.2 while testing white flat wall; and < 0.1, while testing on textured surface.

The error shown on the Depth Quality Tool is showing 0.24+- for 20% ROI, 560mm away from wall. However, while I move the sensor away, to about 1 meter from wall, Subpixel dropped to < 0.2.

My question is, what is the right distance to test?
In the attached screenshot, there is a weird orange dot appear in depth stream (squared in green). Is it expected?

Re: Correct Distance for test Subpixel RMS error?

MartyG — Mon, 17 Jun 2019 13:35:07 GMT

As long as the distance to the wall is above the camera's default minimum distance (about 50 mm or more from the wall, then), and your maximum distance is less than 3 m (as RMS error becomes noticable past 3 m distance) then it should be a "good" distance. It is very subjective though, as there are a range of environmental factors - such as light-sources - that can affect a camera's results.

The quality of the camera image is partly dependent on how much 'texture' detail it can read in a scene. Flat plain surfaces such as walls and doors do not provide the camera with much texture. In these circumstances, the camera's IR Projector component can be used to project a semi-random pattern of dots onto the surface to increase its readable texture detail. Alternatively, you can point the camera at a print-out of the textured image supplied in the tuning guide.

I would speculate that as you zoom in close on a plain white wall, the amount of readable texture is being reduced, like focusing the camera in on a sheet of plain white paper and excluding everything outside of the plain white featureless paper from the camera's view.

In regard to the orange dot, it is hard to say what that is. It is in a 'bad' area of the image, where the far-left of the image breaks up into an "invalid depth band". Pages 58-59 of the current edition of the data sheet for the 400 Series cameras explains about this phenomenon.

https://www.intel.com/content/dam/support/us/en/documents/emerging-technologies/intel-realsense-technology/Intel-RealSense-D400-Series-Datasheet.pdf

Re: Correct Distance for test Subpixel RMS error?

JesusG_Intel — Tue, 18 Jun 2019 05:53:06 GMT

As of June 10, 2019, Intel® RealSense™ product support has moved to https://support.intelrealsense.com. All new community posts must be submitted at https://support.intelrealsense.com. If you have existing threads on this Forum that were opened before June 10, 2019 and are not resolved, please re-post your question at https://support.intelrealsense.com. This Forum will continue to be searchable after the cutover.

You will still be able to create questions on this forum but those questions will not receive support from Intel. If you desire Intel support, please go to https://support.intelrealsense.com.