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: What is the fastest possible framerate if lowing the resolution simplify the algorithm (like tracking)

Part Number: IWR6843ISK-ODS

I have tried the 3D people counting, area scanner, and out-of-the-box demos. 

For 3D people counting and area scanner, the frame rate seems to be only 10 Hz, and by tuning the parameters in the out-of-the-box demo, the highest frame rate is 30

For my current application, a much higher frame rate is favorable, so I wonder if it is possible to increase the frame rate to, say 100 fps for maybe even 1000 fps

A reduced resolution and maximum range are acceptable, also I could also remove the tracking algorithm in the 3D people counting and the area scanner demo

I guess my question is that what is the bottleneck for the frame rate of the current demos? 

If it is due to the chirping sequence, and cannot be accelerated, I guess I will need a workaround for my current application.

But if this is due to the heavy calculation for the tracking, I wonder if a higher fps can be achieved by only gathering the low-level data with a reduced resolution


And if there are ways to accelerate the process, please let me know where to start.

I guess this is anther wired question, any suggestions are very welcome!

  • Hi Toby,

    You should be able to achieve about more than 10 FPS on our people counting and area scanner demos. The AOP_6m_default.cfg has a frame rate of 18 Hz.

    The fundamental bottle-neck for data collection will be the time needed in between frames to measure velocity and movement, as well as the time needed for the HWA to run its operations. If you want to do all processing on a separate device, you may be able to speed up your collection, but I find it hard to believe you'll be able to achieve speeds of 1000 FPS with our device. 1000 may be possible, but you would need to reduce the number of transmitters and receivers to 1 and 1, chirp as fast as possible, and as few times as possible.

    Best,

    Nate

  • Thank you, Nathan!

    As far as I know, the module needs several chirps to obtain the velocity.

    If I do not need the velocity information, can I reduce this process (using multiple chirps to get the speed) and speed up the fps?

    If so, is there any example of this? or any suggestion about how I could start?

    Thanks!

  • Hi Toby,

    You could set the number of chirps per frame to 1, but this is not usually recommended. Do you mind sharing what you want the radar to do? Most people detection/tracking requires velocity information.

    See our sensing estimator tool here to see how to create the config file that sets the number of chirps per frame to 1.

    https://dev.ti.com/gallery/view/mmwave/mmWaveSensingEstimator/ver/2.0.0/

    Best,

    Nate

  • Thank you Nate!

    The issue is that we are going to install the sensor onto a moving platform, which could move very fast (in position and orientation)

    If we keep the speed information, we will need to find a way to compensate for its speed for all the points (we do have the moving information of the platform). We believe that this is very hard, but maybe you have some insight into this?

    Or, is there any better way to implement people detection for our scenario (sensor installed on a moving platform)?

    We are digging into the vital-sigh approach (to check whether there is a person even when the platform is moving), yet it doesn't seem to work very well Disappointed


    Best,

    Toby

  • Hello Toby.

    Nate is currently out of office and will be able to provide an answer next week.

    Sincerely,

    Santosh

  • sure, thank you, Santosh!

  • Hi Toby,

    Have you looked at our small-object detection lab? This might be closer to what you're looking for.

    Best,

    Nate

  • Have looked at the guide, yet need to prepare for the ROS platform to really try it.

    Thanks a lot though!