• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 619 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.

Original question:

IWR1642BOOST: How to improve the position accuracy of targets in people counting

IWR1642BOOST: How to improve the position accuracy of target tracks in people counting

michaelbcd
Prodigy 210 points
Part Number: IWR1642BOOST
Other Parts Discussed in Thread: IWR1642, IWR6843

Hi Guys,

I am using IWR1642Boost for people counting and localization, but the position accuracy of the target tracks is not very good. The scene I concern is just as the people count demo shows (refer to mmwave_industrial_toolbox_3_1_1\labs\lab0011-pplcount), and I run the demo in a room of  6x10 meters. I set up the demo environment exactly as the demo guide requires.

I've noticed that it lists several parameters which will impact the target tracks in the document of pplcount_customization_guide.pdf. But, even if I set the exact or correct parameters, it doesn't improve the track position accuracy. Moreover, according to my research on the algorithm process in IWR1642, I find the position data of the points cloud is calculated in the DSP, and the tracker algorithm in ARM just utilize that data and output the track info. While, the parameters in customization guide may just refine the target track result.

Q1. As it shows in document of People Tracking and Counting Reference Design Using mmWave Radar Sensor.pdf of IWR1642, the WAYPOINT of (1, 4) is detected at (0.87, 4.23), the error(difference) of which is very large and my demo result also gets the similar errors. So, my question is how to improve the position accuracy of the target in IWR1642? The way could be of either algorithm, software or hardware.

Q2. I've also noticed the position accuracy of IWR6843 is much better than IWR1642, the WAYPOINT of (0,4) being detected at (0.03, 4.0125).  What is the main difference between IWR6843 and IWR1642 caused the performance difference? The algorithm or the hardware? If being the algorithm, can the algorithm of IWR6843 be ported to IWR1642. or is there any algorithm update for IWR1642?

Thank you.

Regards,

Michael

  • 0
    TI__Guru 50240 points
    Hi Michael,

    I will need more information in order to assist you, is this discrepancy repeatable 100% of the time? Would you be able to post some pictures of your scene? Which chirp configuration are you using?

    It is worth noting that the algorithm on both devices are the same, there should not be any performance difference between either devices unless you have made some changes at code level.


    Cheers,
    Akash
  • 0 in reply to Akash Gondalia
    Prodigy 210 points

    Hi Akash,

    The scene is as shown in the following pictures:

    labsetup01

    In the above scenario, we tested the positions of (0, 1), (0, 2), (0, 3), (0, 4), (0, 5). I just show the test result of (0, 4) in the following (the result of other points as almost the same):

    The whole configuration is as following:

    dfeDataOutputMode 1
    channelCfg 15 3 0
    adcCfg 2 1
    adcbufCfg 0 1 1 1
    profileCfg 0 77 30 7 62 0 0 60 1 128 2500 0 0 30
    chirpCfg 0 0 0 0 0 0 0 1
    chirpCfg 1 1 0 0 0 0 0 2
    frameCfg 0 1 128 0 50 1 0
    lowPower 0 1
    guiMonitor 1 1 0 0
    cfarCfg 6 4 4 0 0 16 16 4 4 50 62 0
    doaCfg 600 1875 30 1
    SceneryParam -2.15 3.78 0.88 8.12
    GatingParam 4 3 2 0
    StateParam 5 5 10 100 5
    AllocationParam 130 0.01 8 1 3
    VariationParam 0.289 0.289 1.0
    trackingCfg 1 2 250 20 200 50 90
    sensorStart

    For Q2, I've noticed there are many difference of the algorithm code between IWR1642 and IWR6843, but I don't know whether this difference would impact the location performance.

    All the test are run with the original code, we didn't modify anything yet.

    Thanks.

    Regards,

    Michael

  • 0 in reply to michaelbcd
    TI__Guru 50240 points
    Hi Michael,

    The algorithms are actually quite similar, if there is a particular discrepancy you'd like to discuss feel free to post it.

    A discrepancy of range data is more likely a result of your test setup, such as the distance of the sensor to the floor, movement of sensors between trials, and even movement of people themselves.

    One document I'd recommend consulting is the Customization Guide, which is available in the People Counting 'docs' folder.


    Cheers,
    Akash
  • 0 in reply to Akash Gondalia
    Prodigy 210 points
    Hi Akash,
    OK, got it.
    thanks.