• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 1560 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.

IWR6843: Object detection: Range-Doppler vs Range-Angle

Tosin Babarinde
Intellectual 435 points
Part Number: IWR6843

Hi,

After several articles, guides and materials, I have come to the realization that TI has two approaches to Object detection/point cloud formation prior group tracking:

1. range processing(1Dwindowing and 1D FFT)-->Angle processing(Clutter removal and Capon beam-forming)-->Objection detection(CFAR)-->Doppler Processing(2D or 3D FFT?)-->Group tracking

2. range processing(1Dwindowing and 1D FFT)-->Doppler Processing(clutter removal and 2D FFT)-->Objection detection(CFAR and peak grouping)-->Angle processing(3D FFT)-->Group tracking

Method (1) was used in Area Scanner demo and reference design while Method (2) was used in People counting demo and reference design.

My questions are as follows:

1. What are the (dis)advantages of say method (1) over method (2) or method (2) over method (1)?

2. For a customer interested in group tracking, which method is more recommended?

-Tosin

  • +1
    TI__Mastermind 37780 points

    Hi Tosin,

    Your assessment is accurate. 

    Method 2 is better at doppler compensation in scenarios where there could be objects moving very quickly and very slowly. When objects move very quickly, the phase of their returns changes because they move within their range bin. However, the phase of the returns is also needed for angle estimation. When we detect velocity then angle, we can use the estimated velocity to back-subtract off the phase generated by the angle of arrival, improving angle estimation. Method 2 is also less computationally intensive because it uses the Bartlett Beamformer for angle estimation.

    Method 1 is more computationally intensive because it uses a Capon Beamformer for angle estimation (also known as MVDR). However, we do find improved localization results on people when using method 1, which is why it's in our people counting demo.  While method 1 is not able to do doppler compensation, we don't find this to be limiting in people tracking applications, as people do not move very quickly with respect to the radar timing. For the group tracker, I would recommend method 1.

    Best,

    Nate

  • 0 in reply to Nathan Block
    Intellectual 435 points

    thanks Nathan for your detailed response.

    Seems to me that you got the order of the methods the order way round(?). As per documentation, Method 1 uses the Capon Beamforming in angle estimation while Method 2 uses the Bartlett Beamforming (or the Multi Object Beamforming)

  • 0 in reply to Tosin Babarinde
    TI__Mastermind 37780 points

    Thank you yes that's correct. I've resolved the above accordingly