• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 30 subscribers
  • Views 684 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: Some static objects are detected as persistently-moving at closer proximity from the mmWave board

Victoria Rudakova
Intellectual 770 points
Part Number: IWR6843ISK-ODS

Dear TI team,

I'm trying to understand a physics of why some closely-located static objects could get detected as moving even though I turn on the static clutter removal.

For the context, here is what I have:

  • IWR6843ISK-ODS, installed overhead, the height ranges 2-4 meters, the environment is similar to an office.
  • I use 3D point cloud lab (FFT-based firmware)
  • I suppress clutter by using "clutterRemoval -1 1"
  • I use a sensitive chirp configs with velocity resolutions ranging 3-9 cm/sec, similar to the profile from the 3D people count lab.
  • I use the following values for CFAR settings: 

cfarCfg -1 0 2 8 4 3 0 15.0 0
cfarCfg -1 1 0 4 2 3 0 15.0 0
multiObjBeamForming -1 1 0.75

In certain circumstances I noticed that some static object (for example, a light fixture or an object on my desk) which are located within a closer distance, like 0.8-1.5 meters, could start getting detected when a person enters a room or in close proximity to those objects. At first I though it may be a reflection, but sometimes those detections persist even after the person leaves the room.

I'm just trying to understand why this would happen and whether you may have some insight into this problem?
Could it be those subject vibrate slightly because of the person's steps and those vibrations are not so noticeable to the human eye?
Or is this something to do with the CFAR thresholding?
Or could there be another reason because of the closer proximity of those objects to the sensor? - if so, could you please explain the intuition behind it, if you have any ideas on that.

Also, in your opinion, would those "spots" be a good subject for classification work (in theory, like it is done in the HVAC lab)? It would be nice to be able to auto-detect them and then ignore the point cloud from that area all together.

Thank you.
Best regards,
Victoria

  • 0
    TI__Genius 14415 points

    Hello Victoria,

    Just to make sure I know exactly what you are loading, can you please confirm which version of the SDK and toolbox you are using? The latest is 3.5 and 4.5.1.  Also are you using the binary included at the following link, or is it a different one? 

    "C:\ti\mmwave_industrial_toolbox_4_5_1\labs\people_counting\overhead_3d_people_counting\prebuilt_binaries\overhead_3d_people_count_demo.bin"

    When you say 'some static object...could start getting detected", do you mean there are points in the pointcloud or is this adding a target track? 

    Regards,

    Jackson

  • 0 in reply to Jackson Thomas
    Intellectual 770 points

    Hello Jackson,

    Here are some clarifications:

    • I use the 03.05 SDK and I use the binary from the Industrial toolbox 4.5.0.
    • I use 3D point cloud FW, not the people count FW, so there is no tracking information - only point cloud.
    • Thus, when I say "static objects are detected", I mean they appear as point cloud data even though I remove static noise.

    I hope that makes more sense.

    Best regards,

    Victoria

  • 0 in reply to Victoria Rudakova
    TI__Genius 14415 points

    Can you please specify which binary file and which visualizer you are using to display the point?

    If the lamp is shiny and metallic, there could be a strong reflection. This should be filtered out by the clutter removal. Do you have staticProcEnabled turned off in the staticRangeAngleCfg parameter? 

  • 0 in reply to Jackson Thomas
    Intellectual 770 points

    Hello Jackson,

    I used the binary from mmwave_industrial_toolbox_4_5_0/labs/out_of_box_demo/68xx_ods_point_cloud_demo/prebuilt_binaries/xwr64xxODS_mmw_demo.bin - It is FFT-based firmware, and not people count (Capon FW).

    The above firmware does not have options that you mentioned (staticRangeAngleCfg?).

    I use my own visualization tool that parses serial port and displays the point cloud, similarly to how it's done using the TI web tool.

    I hope that makes it more clear!

    Regards,

    Victoria

  • 0 in reply to Victoria Rudakova
    TI__Genius 14415 points

    Hi Victoria,

    Thank you for the info. Have you tried adjusting the threshold values in the cfarcfg and multiobjbeamforming parameters? This could help differentiate between the objects. There is information on these in the SDK user guide.

    Regards,

    Jackson