• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 31 subscribers
  • Views 517 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR6843ISK: Adjustment of Point Cloud Data to Match Demo Performance

Zach W
Prodigy 130 points
Part Number: IWR6843ISK
Other Parts Discussed in Thread: IWR1642,

Hello,

I'm working on the tuning of the Long Range People Count Demo (Lab 19) on the IWR6843ISK board. Previously I did some testing with an IWR1642 with the Traffic Monitoring Demo (Lab 13) and I'm trying to match the results I got. The main thing I notice that gives me falsing is the point clouds as you see here. Both demos were ran in their default state and the LRPC Demo had a generous amount of noisy point clouds coming in and out that would at times effect classification and even cause ghosts. The end goal is not to get the whole 50m, even though we'd like the most as possible, but to have clean points clouds and classification similar to how well the Traffic Monitoring Demo ran. Does anyone have any specific suggestions on which parameters would reduce the noisy point clouds that are caused on the LRPC demo? Screenshots of the comparison are below.

Thanks!

  • 0
    TI__Mastermind 28055 points

    Hi Zach,

    Are there any moving objects like shrubs in the location where the noisy points are?

    In general, you can reduce noisy points by changing the CFAR parameters:

    1108.chirpParams_TMdemo.xlsx

    Increasing the thresholds (i and j in the xls) will require detected points to be stronger.  Please read up on the CFAR algorithm to understand what the other parameters do.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    TI__Mastermind 28055 points

    Hi Zach,

    If possible, can you send pictures or video of the environment where these detections occurred? I am trying to determine if the detections should be happening, in which case we can work on filtering, or if there is nothing there and the detections should not occur.

    Finally, as well as increasing the CFAR thresholds, you can try lowering the TX output power - change this line in the config:

    profileCfg 0 61.00 35 6 43.0 9 0 8.241 1 125 3433 0 0 48

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 130 points
    Okay I will try to adjust my CFAR values. I will try 9 for TX output power back-off code. Is there a definition for what the value means as default is 0. Do you have an email that I can reach you for the environment location?

    Thanks,
    Zach
  • 0 in reply to Zach W
    TI__Mastermind 28055 points

    Hi Zach,

    Looking at the testing environment, I understand that there is nothing in the area that is creating the detection points you are seeing.  If the already mentioned procedure does not help, you can tune gtrack to better ignore the false points.

    You can increase the SNR threshold or the points threshold.  You can also decrease the maxDistance threshold, which will limit the size of cluster that the tracker is allowed to allocate.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 130 points
    Okay. Thanks Justin! I'll close the issue for now and let you know after changing these parameters what results I get.