• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 30 subscribers
  • Views 621 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.

AWR1642BOOST: Accuracy for person tracking

Martin Waitz
Prodigy 70 points
Part Number: AWR1642BOOST

I have found your great sensor product. In our application, we would like to measure the distance of a person in the range of 0.1 – 15m.

Between 0.1m – 6m we need a accuracy below +-2cm. Is such an accuracy possible? Or what is possible "out of the box"?

The measurement should be done indoor and outdoor. Does the accuracy change within a building?

Can the measurement be done easily on the Evaluation board, or does the sensor need to be specially tuned?

Can a measuring rate of 20 Hz be achieved?

Thank you for your help.

  • Former Member
    0 Former Member

    Hello Martin,

    Have you seen our people counting and tracking demo example 

    "Out of the box"

    The range of 0.1-15m is achievable.

    However, the accuracy will be difficult - can you clarify on this requirement? When we are detecting people in this demo we see a spread of points to represent a person. The arm or leg can be swinging in front while the chest area is also detected. The "location" of the person is computed as the centroid of the detected points. The accuracy would be determined by what you consider the "true location" of the person. Is that the location of the chest? The closest part of them? We have not extensively tested this localization accuracy and I point out these questions for you to consider as you evaluate the performance.

    In an indoor environment there are more surfaces that can cause multipath or erroneous reflections. However, please refer to the tuning guide that illustrates parameter changes that can be used to improve performance indoors. In terms of the exact definition of localization accuracy there should not be a difference.

    This TI Reference Design illustrates how the 1642 evaluation board can be used for people counting and tracking: 

    Yes, the 20 Hz frame rate is used in the reference design linked above.

    If you have any other questions please reply. Otherwise if this answers your question please mark the question resolved.

    Best,

    Amanda

     

  • 0 in reply to Former Member
    Prodigy 70 points

    Dear Amanda,

    Thanks for the detailed answer.
    I did not know the link and papers for  "people counting" yet.

    In my application, I am interested in the distance and position to the sensor. Whereby only the closest point of a person is the important information.

    If a person is always in the same position, is it possible to measure the distance from the closest point to the sensor at a distance of up to 6 or 10m with an accuracy of less than 3cm? Or what is theoretically possible?
    And would this measurement be accurate as long as the person is always the same place (the person doesn't move and nothing is changed in the enviroment)?

    Can this distance and position be tranfered on the UART or SPI interface?

    Many Thanks.

  • Former Member
    0 Former Member in reply to Martin Waitz
    Hello Martin,

    With the people counting demo it is not possible to know if the returned reflection points are the "closest" part of the person or another point of the person. In the demo, the centroid of the point cloud that determines the person is considered the "location" of the person. Detecting people of a range of 6-10m is definitely possible However the accuracy of exact localization is dependent on the SNR which is determined by the object, scene conditions as well as device/EVM.

    In the TI Design and example video, you can see in the "box demonstration" that we can detect when a person steps in and out of a box. The tape that defines the box is about 3cm so in this case we can accurately determine within 3cm whether the person is on one side of the line or not. training.ti.com/people-counting-demonstration-using-ti-mmwave-sensors (~28sec)

    The demo ships via UART the position of the tracked object which can be used to calculate distance as well as other details. Please refer to the UG for the UART stream details. dev.ti.com/.../



    Best,
    Amanda