• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 30 subscribers
  • Views 458 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: can't detect object when stopped (static)

JINHYEONG LEE
Intellectual 740 points
Part Number: IWR6843ISK

Hi,

I tested OOB Demo in SDK 3.5

 

I detected the vehicle in front and observed the result using IWR6843ISK.

It was detected when the vehicle was moving, not when it was stopped.

 

The vehicle was detected from 40 m and was not detected after stopping.

The distance from the stopped vehicle is about 28 m, and when the vehicle is moving, it is detected again.

The configuration used are as follows.

% ***************************************************************
% Created for SDK ver:03.05
% Created using Visualizer ver:3.5.0.0
% Frequency:60
% Platform:xWR68xx
% Scene Classifier:best_range
% Azimuth Resolution(deg):15
% Range Resolution(m):0.195
% Maximum unambiguous Range(m):50
% Maximum Radial Velocity(m/s):15
% Radial velocity resolution(m/s):0.94
% Frame Duration(msec):100
% RF calibration data:None
% Range Detection Threshold (dB):15
% Doppler Detection Threshold (dB):15
% Range Peak Grouping:enabled
% Doppler Peak Grouping:enabled
% Static clutter removal:disabled
% Angle of Arrival FoV: Full FoV
% Range FoV: Full FoV
% Doppler FoV: Full FoV
% ***************************************************************
sensorStop
flushCfg
dfeDataOutputMode 1
channelCfg 15 5 0
adcCfg 2 1
adcbufCfg -1 0 1 1 1
profileCfg 0 60 8 7 33.64 0 0 30 1 320 12499 0 0 158
chirpCfg 0 0 0 0 0 0 0 1
chirpCfg 1 1 0 0 0 0 0 4
frameCfg 0 1 32 0 100 1 0
lowPower 0 0
guiMonitor -1 1 1 1 0 0 1
cfarCfg -1 0 2 8 4 3 0 15 1
cfarCfg -1 1 0 8 4 4 1 15 1
multiObjBeamForming -1 1 0.5
clutterRemoval -1 0
calibDcRangeSig -1 0 -5 8 256
extendedMaxVelocity -1 0
bpmCfg -1 0 0 1
lvdsStreamCfg -1 0 0 0
compRangeBiasAndRxChanPhase 0.0762179 -0.60397 -0.64542 -0.66522 -0.66946 -0.64209 -0.70471 -0.37708 -0.86667 -0.76633 -0.37451 -0.84061 -0.35657 -0.78491 -0.30798 -0.61557 -0.57355 -0.83316 0.55045 -0.82367 0.56708 -0.82932 0.52661 -0.97696 0.17557
measureRangeBiasAndRxChanPhase 0 1.5 0.2
CQRxSatMonitor 0 3 4 79 0
CQSigImgMonitor 0 105 6
analogMonitor 0 0
aoaFovCfg -1 -90 90 -90 90
cfarFovCfg -1 0 0 49.99
cfarFovCfg -1 1 -15 15.00
calibData 0 0 0
sensorStart

The detection result using visualizer is as follows.

A moving vehicle could be detected at a greater distance (40m), but it does not make sense that it is not detected when it is stopped.

Any advice on why the static object is not being detected would be appreciated.

Thanks.

  • +1
    TI__Mastermind 19520 points

    Hello Jinhyeong.

    The reason why the vehicle is not being detected, it is very difficult to detect static object due to the way it works.  When a chirp is sent out and reflects off a moving object, it returns with an added doppler shift, and this change of frequency is what is used to detect an object.  However, if an object is static, it is no different from any other static object in the field of view, such as a tree or a lamp post, as the reflected wave is the same as the transmitted wave except with a potential phase shift.  And since it is very far away, there may not be enough points within the range bin for it to deem that something has been detected in that area.  However, the moving object shouldn't be picked up by the zero doppler graph as it no longer has zero doppler and shouldn't be detected, so it may be that the second point at around 35 meters may be something else and not the car.  The traffic monitoring demo and the long range people detection demo, in comparison to out of box, may be more helpful in trying to pick up objects as they use a tracker which will continue to pick up on static objects better.  The area scanner demo also has additional processing that can improve static detections in a given scene.

    Hope this helps.

    Sincerely,

    Santosh

  • 0 in reply to Santosh Krishnan
    Intellectual 740 points

    Thank you for reply,

    Reading your answer and the OOB Demo doxygen documentation, I understand that the algorithm used in OOB Demo, it goes in the following order: Range FFT -> Doppler FFT -> CFAR -> Angle FFT. So if the object is static, then CFAR because of the surrounding static object It is understood that the probability of becoming a detected point is reduced, and therefore the detection performance is inferior to that of a moving object.

     

    I read the Traffic monitoring Demo Design guide documentation, and the detection algorithm was the same as the OOB Demo.

    https://www.ti.com/lit/ug/tidud31b/tidud31b.pdf?ts=1657591369705&ref_url=https%253A%252F%252Fwww.ti.com%252Ftool%252FTIDEP-0090 

    You mentioned tracker, are you referring to gtracker and not the detection algorithm?

     

    I also looked at the documentation of the 3D People Counting Demo, and it is stated that the detection algorithm of 3D People Counting is performed in the following order: Range FFT -> capon beamforming, resulting in a range-azimuth(wall mount) or range-azimuth-elevation(cell mount) matrix.(mmwave_industrial_toolbox_4_10_0/labs/People_Counting/docs/3D_people_counting_detection_layer_tuning_guide.pdf)

    Could this algorithmic difference make it better to detect static objects?

    Thanks.

  • +1 in reply to JINHYEONG LEE
    TI__Mastermind 19520 points

    Hello Jinhyeong.

    In regards to the tracker, yes it is the gtracker. The tracker is different from the detection algorithm in that it keeps track of moving objects even after they stop moving. In regards to the algorithmic difference, there may be a slight performance difference between the two, but it is more useful to use the tracker to detect static objects. There is also a static retention feature in area scanner that would be useful.  I recommend you tune the device to make sure you don't get any false detections.  Overall, I recommend that you use the people counting, traffic monitoring, or area scanner labs over the Out of Box demo to do this kind of detection as it has the tracker feature available to do such detection.  The out of box demo is not designed to track or distinguish between static and dynamic objects, but the other demos build on this base demo to provide greater functionality.

    Hope this helps.

    Sincerely,

    Santosh