• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 31 subscribers
  • Views 752 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.

AWR1642: RED frame on　AWR1642 Occupant Detection Demo

Eiji Tsuruta
Prodigy 55 points
Part Number: AWR1642

Hi,Team,

My customer is using TI's occupant detection demo by AWR1642  BOOST,

In customer's field demo, they can observe one target like the  attached right picture but not display 'RED frame'.

They want to display RED frame,

So Would you let me me know the below questions?

Q1;Would you inform  the condition of the  red frame on the attached file (in the right picture)?'

Q2: Can I adjust the threshold of  the condition of RED frame?,How do I  to realize it?

Q3:Can I add more RED frames?

Cheers

E.T

  • 0
    TI__Mastermind 21025 points
    Hi Eiji,

    > Q1;Would you inform the condition of the red frame on the attached file (in the right picture)?'

    The frames in the heatmap display are gray when the decision from the AWR1642 declares the zone unoccupied. Red means there is a moving object (a positive detection) in the zone. This is explained in the user guide in the docs folder of the demo. If you see a bright spot inside the zone, but the zone remains gray, I suspect one of the following:
    1) the "decision vector" flag in the guiMonitor profile command is turned off and the last command line argument is set to zero.
    2) some of the other profile commands such as coeffMatrixRow or oddemoParms have been modified such that positive decisions are difficult.
    3) your coefficients require tuning for your field environment.

    I would suggest trying with an unmodified profile found in the chirp_configs/ folder. If you use an unmodified profile, and have flashed unmodified MSS and DSS code, you should see positive detections when a bright spot is in the zone. If it doesn't behave optimally, you may need to tune the coefficients as described in the user guide, but it should detect some of the time.

    > Q2: Can I adjust the threshold of the condition of RED frame?,How do I to realize it?

    There is no threshold. You can re-tune the coefficients as described in the user guide, or tweak some of the other parameters in the profile.

    > Q3:Can I add more RED frames?

    The algorithms used in this version of the demo are somewhat tuned for two zones. You can find the algorithms here: www.ti.com/.../TIDEP-01001. You can create your own algorithms using the heatmap as a starting point. We are working on variations of this demo that create 4 or more zones, but they aren't complete at this time.

    -dave
  • 0 in reply to db_woodall
    TI__Expert 3305 points
    Hello Dave,

    In the steps you suggested in (1), if the object does not found (due to heatmap level is lower than expected detection level) with default profile/chirp configurations in OD demo .cfg, how to tune coeffs? My idea is put the coner post or stell can (should be found more easier) and take the detected pattern (both zones, and zone1 or zone2 only) and non-detected pattern as coeffs tuning tools in UG. Is this right way? If not, could you suggest recommended way?

    Regards,
    Takeo
  • 0 in reply to Takeo Sekiguchi
    TI__Mastermind 21025 points

    Hi,

    From the information provided in the emails on this topic, it appears "dolls" are being used.  This is somewhat problematic as a doll will not provide the same movement energy as a live person or animal.  If dolls must be used, I would suggest retraining the coefficients. This is explained in detail in the User Guide that accompanies the Occupant Detection demo source, GUI and executables.  Look for the section "Zone and Coefficient Tuning" at the bottom of the user guide.

      -dave

  • 0 in reply to db_woodall
    TI__Expert 3305 points
    Hi Dave,

    (1) What I wanted to know is if the object was always not find in both (or one of) Zone how to do the "Zone and Coefficient Tuning" procedure. Now using the defalut od_demo_car_0318_1p0.cfg but could not detect the objects in both zone. In this case, what is next step to tune the coeffs?
    At latest try results with default 1p0.cfg is, able to detect object on left side but not on right size. Please give advice.

    (2) Separated from (1), in case of no matlab license, ie. not able to build coeffs_data.exe, is there altenative way to tune the parameters?

    (3) Could you tell me the each parametes role - coeffMatrixRow, meanVector and stdVector?

    Regards,
    Takeo
  • 0 in reply to Takeo Sekiguchi
    TI__Mastermind 21025 points

    Hi Takeo,

    Sorry for not responding sooner.

    1) I am sending you a Box link with the compiled tools. As I mentioned before, the instructions are in the demo's user guide.

    2) Nope.

    3) These coefficients are generated by the tools by processing the collected training data. Since the training data is known (occupied vs not occupied) for all cases, the coefficients give the demo the knowledge of what the power will look like when the zone is occupied.  In the case of the dolls, I suspect that the power level is lower than when a real subject is present in the zone.

    An alternative to all of this is for the customer to replace the decision making code with other algorithms (such as image processing) for processing the heatmap. We have customers already doing this.

      -dave