• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 31 subscribers
  • Views 203 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

IWR1843: 3D Traffic monitoring with 1843 - confusing Z coordinate

Veljko Papic
Prodigy 40 points
Part Number: IWR1843

Hi,

We are trying to obtain height of a vehicle approacihng or departing ramp on public parking. Detection and counting of cars is fine, only we get confusing results with measuring height of the object. Sensor is at approx. 3m height with 5 degrees tilt toward the ground. The scenary parameters are pasted below.

staticBoundaryBox -12 12 5 14 0 3
boundaryBox -10 10 6 12 0 3
gatingParam 6 5 10 5 10
stateParam  20 20 60 40 10 100
allocationParam 60 80 0.1 6 40 5
maxAcceleration 1 2 1
trackingCfg 1 2 200 10 135 422 100
sensorPosition  3 0 5
presenceBoundaryBox -10 10 6 12 0 3

The strange values for Z coordinate we obtain from TLV7 as the mean value of the Z coordinate through the frames the specific target appears in the scene. Often we obtain the negative values for the Z coordinate, X and Y coordinate are OK. Can you explain me in detail how the sensor calcuates the Z coordinate  and please share with me scene geometry in this specific case. We use the modified chirp configuration file from the D:\mmwave_industrial_toolbox_4_11_0\labs\Traffic_Monitoring_chirp_configs\18xx_traffic_monitoring_70m_MIMO_3D.cfg.

Please, we need to solve this issue as urgent as possible because we have customer presentation in 3 days.

Thanks in advance, Veljko

  • 0
    TI__Expert 8620 points

    Hello.

    It is possible that the sensor position was correct in the configuration, but incorrect in the GUI as there is a place to input sensor position.  Can you confirm that you have updated the sensor position on the GUI as well?

    Sincerely,

    Santosh

  • 0 in reply to Santosh Krishnan
    Prodigy 40 points

    Hi Santosh,

    Thank you for the prompt reply. We use CLI and afterwars only UART raw data. Software is later ours and we designed visualizstion based on the point cloud and target ids from specific TLVs. So, we don't use at all GUI designed by TI. I have to repeat, the results in estimating X and Y coordinates are very good, but for Z we get sometimes negative values.

    The results are very confusing. Is it possible that some compensation of the sensor height is made in the sensor software that TI provides in this lab (Traffic Monitoring). We checked the code several times and cannot find any mistake in our software.

    So please, we still need help.

    Thanks in advance,

    Veljko

  • 0 in reply to Veljko Papic
    Prodigy 40 points

    So, my basic question is:

    Are the raw data that we collect through UART in the sensor coordinate system with the origin in the sensor or is it in world coordinates and how the tracking parameters sensorPosition (that we set to 3 0 5) affects the scene geometry.

    Thanks,Veljko

  • +1 in reply to Veljko Papic
    TI__Expert 8620 points

    Hello.

    The raw data that you collect is with respect to the sensor as the origin and not the world coordinates.  So if there is tilt in the sensor position you will have to account for that when calculating the Z coordinate.  The sensor position essentially sets the origin point for the sensor, so in your case by setting it at 3, 0, 5, you are setting the origin point at 3 meters height and with a tilt of 5 degrees.  This means when the signal reflects off a surface below the sensor and the adc data is parsed, the Z coordinate will be negative.  You can get more information on the sensorPosition command here.

    Sincerely,

    Santosh

  • 0 in reply to Santosh Krishnan
    Prodigy 40 points

    Hi again Santosh,

    It is OK now with the explanation of the sensor and world coordinates. Anyway, we performed the testing, but still get not so satisfactory results. When we detect the vehicles, and there are many, as we observe the public parking, we can not obtain some relieble difference between motorbikes, normal small cars, SUVs, vans and big tracks or buses in matter of the vehicle height. We would like to classify the object according to the height of the vehicle, but the results are not promising. Is there any way that you can help us, maybe with the tuning of the detection or tracking parameters, or anything else.

    I am repeating, the x and y coordinates of the point cloud and targets' data, so TLVs 06 and 07, are very good but in 3D traffic monitoring lab with IWR 1843 the results are very poor in getting reliable Z coordinate.

    Thanks in advance, Veljko

  • +1 in reply to Veljko Papic
    TI__Expert 8620 points

    Hello.

    We have documentation on the tracking and detection layer that you can use to tune to your specifications.  However, I do not have any specific suggestions to do this as this is not a case which we have done any testing/development for.  For classification, we typically find it best to extract features from the processing chain(i.e. range profile) rather than the point cloud, but I'm not sure what features you could use to do this.

    Sincerely,

    Santosh