RealSense D435 distance depends on aspect

[CCarl8]

Hi

I am testing the new RealSense D435 using the Intel Realsense Viewer app that comes with the 2.0 SDK.

I have put up an object in my room and measured the precise distance to it. 3.0m.

When I set the Stereo Module resolution to 848x480 it measures 3.04m. Pretty good!

But when I set the resolution to 640x480 and measure again i get 3.13m. About 10cm off!

At 424x240 it seems to be about 15cm off. But at 480x270 it's similar to that of 848x480.

What am I missing. Is this a bug? I also tried expanding the depth in Unity and I get the same odd result.

All the best

Carl Emil

[MartyG]

Intel has a documented procedure for testing the depth quality of the 400 Series cameras..

https://www.intel.com/content/dam/support/us/en/documents/emerging-technologies/intel-realsense-technology/RealSense_DepthQualityTesting.pdf https://www.intel.com/content/dam/support/us/en/documents/emerging-technologies/intel-realsense-technology/RealSense_Dep…

[CCarl8]

Thank you for the reply. I read the document and did a quick test.

The distance varies when I change resolution. The change is very noticeable when switching between 640x480 and 848x480, this time being most correct at 640x480.

Is this a common issue?

[MartyG]

As far as I am aware it is not a common issue, though I remember one similar case earlier this year. I would recommend using the 400 Series' calibrator tool to ensure that the camera is calibrated correctly, as there were a few cases where users believed that their D435 was uncalibrated when they first used it new out of the box.

[PSnip]

Hi,

When you decrease the resolution, the depth error increases. That's why you should aim to use the camera at optimal resolution. The official recommendation from Intel is not to use D435 below 848x480 and not to use D415 below 1280x720. Doing this will reduce the depth accuracy.

What is the purpose of operating camera at lower resolution? If you want to reduce the resolution or number of pixels, Intel recommends that you capture at max resolution and then do post processing to reduce resolution.

Please refer to following paper for more information

https://www.intel.com/content/dam/support/us/en/documents/emerging-technologies/intel-realsense-technology/BKMs_Tuning_RealSense_D4xx_Cam.pdf https://www.intel.com/content/dam/support/us/en/documents/emerging-technologies/intel-realsense-technology/BKMs_Tuning_R…

[CCarl8]

Thank you for pointing that out. Very useful. I was attempting to run at lower resolutions because I didn't need so many pixels and I was hoping that it would up the performance. But I can just downsample the depth into another texture.

[RTasa]

Think of as a difference algorithm which it is. When the dots are larger, as you lower the resolution, the accuracy less because the size of the pixel is larger

and therefore measurement is more coarse.

[jb455]

Just for information: one benefit of decreasing the resolution is that it also decreases the minimum range so the camera can get closer to its subject, which may be more important to some people than the increased accuracy.