• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 2 answers
  • Subscribers 30 subscribers
  • Views 462 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-ODS: people counting

Ib Agard
Prodigy 10 points
Part Number: IWR6843ISK-ODS
Other Parts Discussed in Thread: IWR6843, IWR6843ISK

I am developing a product that uses the IWR6843, and for prototyping I am using the IWR6843ISK-ODS development board.  I have flashed both the basic demo software (versions 03.05.00.03 and 03.05.00.04) and the people-counting demo software (version 03.05.00.04) and have successfully communicated with these demos.

The basic demo software supported the command "measureRangeBiasAndRxChanPhase" that allowed one to obtain calibration coefficients that can be passed to the compRangeBiasAndRxChanPhase command.  The People-counting demo software does not seem to support this command, yet still supports the compRangeBiasAndRxChanPhase command.  How should I do calibration with the people-counting demo firmware?

While I have found all the firmware source code for the basic demo, and have successfully complied it, I have not been able to find all the source code for the people counting demo.  I've only found a portion of the people-counting source code, a portion that allowed me to see some of the commands supported by the people counting demo.

I want to perform a calibration with the people-counting demo firmware, to see if that will reduce the significant ghosting problems I am seeing when trying to count the number of people in a room.

Any help or recommendations would be appreciated.

  • 0
    TI__Mastermind 37805 points

    Hi,

    Calibration is the same regardless of demo. So, you should be able to obtain your calibration coefficients from the out-of-box code, then use those for people counting.

    That being said, ghosting is not really a symptom of bad calibration. See the people counting documents here for what to do about ghosting.

    https://dev.ti.com/tirex/explore/node?a=VLyFKFf__4.12.1&node=A__AD7Cm.UWpaYqqCsnR6Gl-A__com.ti.mmwave_industrial_toolbox__VLyFKFf__4.12.1

    Best,

    Nate

  • 0 in reply to Nathan Block
    Prodigy 10 points

    I'll read through the documentation again.  I have been playing with some of the following configuration parameters:

    - Increasing the SNR allocation parameter
    - Increasing the min-points allocation parameter
    - Increasing the value of det2active
    - decreasing the value of det2free, static2free, active2free, and exit2free

    Object tracking reliably gets confused.  I was hoping better calibration might help, but you're suggesting it won't.  I am scanning a space, and walk in a straight line through the space (from left to right, on a line (parallel to the X axis) about 1.5 meters from the sensor).  When entering the space I am identified as an object, and tracked partway through, but then one or two ghost objects appear (usually ahead of me), and the initial object disappears.  Shortly after the initial object disappearing, one of the ghosts shifts position to where I actually am and more or less tracks me as I exit the space.  Any ghosts that are still around when I exit the space also disappear as I exit the space.  From a people-counting perspective, rather on a count of "1", I end up with a count of between 2 and 4 people.

    Where can I find the complete source code for the people-counting demo image?  I'm thinking of adding the ability to generate calibration files to the people-counting image.  Also having periodic temperature messages, and info on how much unused processing time is avalable would also be nice (like the basic demo image provides).

    Can I achieve a people-count accuracy of 20% or better with this product?  The Youtube videos suggest that's possible.  My experiments with your demo boards say otherwise, so far.

  • 0 in reply to Ib Agard
    TI__Mastermind 37805 points

    Hi,

    Have you tried decreasing the transmit power or increasing the SNR detection threshold? All the parameters you're referring to here come from the tracking layer, which assumes that you've already optimized the detection layer. The Detection Layer tuning guide will tell you how to adjust these parameters. You may consider creating and sharing a video showing a screen capture video of the radar output, as well as a camera capture of the space. This will help us diagnose the problem. 

    The source code for the people counting can be found in the industrial toolbox. You can follow the instructions for modifying it here.

    https://dev.ti.com/tirex/explore/node?a=VLyFKFf__4.12.1&node=A__AMJCnZM2ZQMOjvSAopWC8g__com.ti.mmwave_industrial_toolbox__VLyFKFf__4.12.1

    I don't know what you mean by a 20% accuracy, but I can say that our IWR6843ISK board, running the people counting software with an appropriately tuned cfg file can track about 10 people with about 85% accuracy +/- 1 person (85% of the time it detects between 9 and 11 people).

    Best,

    Nate