• DepthAI
  • Increase depth capability with OAK setup

Hello,
I have first used my OAK-D with own custom trained Yolov7 with success and I would like now to have few hints to go beyond:

1) I would like to get distances also for objects which are far away (e.g. up to 50-100m). In order to increase the depth capability to larger distances by triangulation, (how) can be possible to use 2 OAK cameras in a setup where the distance among them is fixed, known and much larger (e.g. 2m)? I imagine that as the distance will be larger than the 2 cameras embedded in the OAK-D one can get better precision, right?

2) Using a Raspberry to control the cameras, what are the pro/contra to use a OAK-POE vs simple OAK camera? Which device / HW shall I rather use to control 2 different OAK (even just for inference w/o depth)? better use 2 Raspberry and 1 OAK each and let them communicate or 1 Raspberry with 2 OAKs?

Thank you in advance,
Marco

  • erik replied to this.

    Hi MZOAK !

    1. For 50-100m, you would need wider baseline distance and/or narrower FOV. We also started working on Long-Range OAK camera which would be perfect suited for this. For now, you can use the modular OAK cams for this.
    2. For PoE vs USB OAK; PoE has much lower bandwidth (so streaming of unencoded high-res frames isn't possible), and it can run in standalone mode. USB needs host computer (RPI). And you can easily run 2x OAK using 1 RPI, docs here.
      Thoughts?
      Thanks, Erik

      erik Thanks for the fast answer! I will think about it, Marco