• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 2 answers
  • Subscribers 31 subscribers
  • Views 636 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: The maximum detection distance of people count FW in front-direction(y-direction).

Big Fields
Intellectual 475 points
Part Number: IWR6843ISK-ODS

Hi,

I am now trying to measure the maximum detection distance of IWR6843ISK-ODS, and did some experiments as below.

[Experiment 1]

- Device

IWR6843ISK-ODS

- FW

"C:\ti\mmwave_industrial_toolbox_3_6_1__win\mmwave_industrial_toolbox_3_6_1\labs\lab0020_pplcount_Overhead\Pre-compiled Binary\ODS\ods_pplcount_lab_68xx.bin"

- configuration file

"C:\ti\mmwave_industrial_toolbox_3_6_1__win\mmwave_industrial_toolbox_3_6_1\labs\lab0020_pplcount_Overhead\GUI_A_base_no_logic\mmw_pplcount_demo_default.cfg"

- radar mount position

(see below)

- How I measured

I ran matlab app in SideMount mode and measured d1 where points(not target tracks) started to appear.

matlab app: "C:\ti\mmwave_industrial_toolbox_3_6_1__win\mmwave_industrial_toolbox_3_6_1\labs\lab0020_pplcount_Overhead\GUI_A_base_no_logic\main_pplcount_viz.m"

- result

d1 = 3.5m

[Experiment 2]

- Device

Same as Experiment 1.

- FW

Same as Experiment 1.

- configuration file

Same as Experiment 1.

- radar mount position

(see below)

- How I measured

I ran the same matlab app in Overhead mode and measured d2 where points(not target tracks) started to appear.

- result

d2 = 5.5m -> (d3 would be calculated as 6~7m geometrically.)

[Question]

According to the Experiment 2 result, the detection distance of 6~7m is achieved even around the edge of FOV.

Therefore, I expected that the detection distance would be longer, say 8~9m, around 0 degree(y-direction) of FOV.

However, the Experiment 1 result doesn't satisfy this expectation.

Why is this so?

(Since it is regarded to mount the radar on ceiling at 3m high for people count demo, is there any software implementation that cuts off points beyond y = 3m maybe?)

Best regards,

Fields

  • 0
    TI__Mastermind 28055 points

    Hi Fields,

    I replicated your experiment 1 and was able to detect past 3.5 meters.  Are there any potential environmental obstructions effecting the ability of the radar to detect when it is mounted lower?

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Intellectual 475 points

    Hi Justin,

    > I replicated your experiment 1 and was able to detect past 3.5 meters.

    What was your result?

    > Are there any potential environmental obstructions effecting the ability of the radar to detect when it is mounted lower?

    I did the experiment 1 in an open area where nothing was around the radar, so it is difficult to think of any environmental obstructions effecting the radar.

    I also did the experiment 3 as below and checked if a person past 3.5m can be detected in range profile level.

    [Experiment 3]

    - Device

    Same as Experiment 1.

    - FW

    "C:\ti\mmwave_sdk_03_01_01_02\packages\ti\demo\xwr68xx\mmw\xwr68xx_mmw_demo.bin"

    - configuration file

    (see attached file)

    profile_rangeProfile_lab0020_default.cfg

    I set the parameters so that profileCfg and frameCfg will be same as those of people count demo.

    - radar mount position

    (see below)

    - How I measured.

    I ran mmWave Demo Visualizer 3.2.0 and captured range profile screen shot when I was standing at 8m away from radar.

    - result

    (see below)

    According to the experiment 3 result, it seems to be possible to detect a person at 8m away from radar in y-direction.

    What can you think of as a cause of this issue?

    What information do you need to investigate this issue?

    Best regards,

    Fields

  • 0 in reply to Big Fields
    TI__Mastermind 28055 points

    Hi Big Fields,

    Please recompile the lab - the src code has been updated to remove a filter on the Y axis, which was used to improve accuracy in the overhead detection use case. This filtered at Y = 2.7 meters. The pre-compiled binary has not been updated, that is why there is an issue here.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Intellectual 475 points

    Hi, Justin.

    I have some questions to your comment.

    Q1.

    > the src code has been updated to remove a filter on the Y axis, which was used to improve accuracy in the overhead detection use case.

    What do you mean by "accuracy" and how does it improve accuracy when filtering Y axis? I would like to know whether this filtering is actually needed or not from the view point of people count performance.

    Q2.

    > The pre-compiled binary has not been updated, that is why there is an issue here.

    Is there any other difference between pre-compiled binary and source code?

    Q3.

    Is the same issue applied to AoP and ISK versions as well?

    (Do pre-compiled binaries for AoP and ISK also filter Y-axis?)

    Best regards,

    Fields.

  • 0 in reply to Big Fields
    TI__Mastermind 28055 points

    Hi Big Fields,

    Q1: I meant accuracy of count. When mounted on the ceiling, there is a lot of power radiated back from the floor, and high chance of multi-path reflection. GTrack should filter these points below the floor, but it was useful to filter all of the underground points as in some cases these could still cause problems, notably if the Boundary box wasn't set exactly right. 

    Q2: For each track, the source code outputs:

    Position: X, Y, Z

    Velocity: X, Y, Z

    Dimension: X, Y, Z - where as previously it output Acceleration here. I believe this is the same between precompiled binary and source code.

    Q3: Precompiled ISK and AoP do not filter Y, they were added later.  

    Regards,

    Justin