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: ODS Point Cloud demo not detecting stationary people

Part Number: IWR6843ISK-ODS
Other Parts Discussed in Thread: IWR6843ISK

We are using the ODS Point Cloud demo for tracking people inside an office where the sensor is mounted at the ceiling at about 2,5 meters facing down. The point cloud data is used by a PC based tracking application to do the tracking.

The problem is that there are no detection points coming from the sensor for stationary people, they only show up when they start moving. This happens even with the clutter removal option turned off and CFAR threshold settings for both range and doppler turned down. The mmWave SDK demo combined with the IWR6843ISK is capable of detecting stationary people in its field of view when mounted in the same position. This makes me wonder if the problem is caused by the large field of view of the ODS sensor picking up too many objects at similar distances to the person, which is what the range/doppler heatmap and range profile output seems to indicate. Are there any ways to solve this issue and allow the ODS sensor to also detect stationary people?

  • Hello Wouter,

    Have you tried disabling the Peak grouping settings in the ODS Point Cloud Demo configuration file? Please disable peak grouping in both range and Doppler domains. Unlike the ISK demo, the ODS point cloud demo is not compatible with the mmWave Demo Visualizer and does not support real-time tuning for these options so you will need to disable peak grouping in the configuration file manually and reload it as also discussed in the following thread:

    e2e.ti.com/.../799402

    Besides this, can you detect stationary humans with your original configuration if you mount the sensor sideways (i.e. a wall or tripod)?

    Regards
    -Nitin
  • Hello Nitin,

    Yes, I have had peak grouping disabled for both range and doppler via the config file. Turning the grouping off did increases the number of points shown when a person is moving, but it hasn't caused stationary people to show up in the detections.

    I will try a setup with the sensor mounted sideways on a tripod and get back to you with the results.

    Regards,

    Wouter

  • With the sensor mounted sideways on a tripod there do appear to be a few detections for stationary people, but this is far less when compared to when the person is moving.
  • Hi Wouter,

    This is expected since you get more reflections from the side (larger reflection surface area as compared to overhead mounting). There are a few things you can do to improve the people detection performance in overhead (as well as sideways) mounting:

    1. Increase the velocity resolution of the chirp: to detect fine human motions such as breathing. You can try the Fine Motion Detection Chirp (6843 version) available in the mmWave Chirp Database. This chirp was tested with side mounting (with the IWR6843-ISK antenna). While we have not tested it with the ODS module, it should work equally well. Please look at the Fine Motion Detection Chirp.

    2. Use the Overhead People Counting Lab: The People counting lab uses a different detection chain aka the "Range-Azimuth" chain (as compared to "Range-Doppler") which is better suited for people detection. It also uses a very high velocity resolution chirp. Please look at the Overhead People Counting Demo.

    Regards

    -Nitin

  • Adding to my last response, with regards to item #1, please note that you will need to adapt the FIne Motion Chirp to the ODS Point Cloud Lab. The Fine motion detection Chirp is based on 2TX antennas while the ODS Point Cloud Lab uses all 3 TX channels. Use the chirp config provided with the ODS Point Cloud Lab as base and update the profile and Frame config parameters based on the Fine Motion Detection Chirp.

    Thanks
    -Nitin
  • Hi Nitin,

    Thanks for the response, I'll go ahead and try your suggestions.

    Thanks,
    Wouter