Depth range enhancement

gmaranta

Hello community,

So i was playing around with the oak-d and oak-d pro. I figure the claim to measure depth up to 35 meter a bit special. I know the environment can have a huge impact on the depth measurements, but even if i provide good conditions, there is no way i get up to 35 meters i figure, it does not even measure up to my upper threshold of 15 meters. So for my application i need cameras which can track people around 4-15 meters.

From what i saw, if using a stereodepth setup, the only way to have reliable tracking, in desired range, would be to have a wider baseline. Lets say if in theory i can measure up to 100 meters, would that according to the disparity graph, give me a high probabilty to work within the ranges i want? Would the measuring error probably scale as well down for that range?

Warm regards,
G.

erik

Hi gmaranta ,
Please see the depth accuracy evaluation here. The value 35m for OAK-D is theoretical, and even theoretical error at such range is very high (33% iirc). From our testing 15m was achievable below 10% error. Regarding long range - we are working on OAK-D-LR (LR = long range) which will have about 15cm baseline distance and option to narrow M12 lenses to achieve up to 100m (really depends on lens FOV). The image below was taken with our Modular Camera series and iirc it was able to perceive depth at 200m+ (very wide baseline and narrow FOV lenses). Thoughts?
Thanks, Erik

gmaranta

Ok that looks quite promising 🙂 what error would you expect this system to have in a depth range of 10-15 meters? Would you think the wider baseline is actually improving the near field error or am i forgetting other circumstances which will have more impact and thus hold the error equally high or even worsen it?

Cheers

erik

Hi gmaranta ,
Depends on the system. Feel free to copy&change some values in stereo calculator spreadsheet. Note that this is full pixel estimation, it is better when enabling subpixel mode.
Thanks, Erik