• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 31 subscribers
  • Views 217 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.

Original question:

IWR1642BOOST: Editing .cfg Files Based on mmWave Sensing Estimator

IWR6843ISK-ODS: Fall detection overhead mount or side mount

Danh Pham
Intellectual 295 points
Part Number: IWR6843ISK-ODS

Hi, I've been testing the sensor for fall detection and people counting in both Side mount and Overhead mount. However, despite using an ODS sensor (Overhead detection sensor), the Side mount has proven to perform better. To increase the accuracy of overhead mounting, what should I change in the config files? Or better, what is the optimal config for overhead fall detection?

Thank you very much!

  • 0
    TI__Genius 14285 points

    Hello,

    While the ODS sensor is often used for overhead detection, this is only because the higher elevation angle resolution design of the antenna. The antenna on the ODS board is also well suited for side mount people counting or fall detection, just with lower maximum range than the antenna on the side mount EVM.

    Our testing and examples of fall detection usually use the ODS antenna as well, in the side mount configuration.  Is your requirement to run fall detection with the sensor mounted overhead? Our visualizer and fall detection algorithm has not been optimized for this. It should be possible but we have not verified the accuracy.

    Can you please indicate which binary file, cfg file, and visualizer you are using?

    Regards,

    Jackson

  • 0 in reply to Jackson Thomas
    Intellectual 295 points

    Hello Jackson, thank you very much for your reply.

    I am currently testing with the bin file from /mmwave_industrial_toolbox_4_7_0/labs/people_counting/68xx_3D_people_counting/prebuilt_binaries, using the default people counting config for ODS and the fall detection visualizer.

    So I've been testing fall detection for side mount on the ODS, and I noticed that false positives happened quite a lot. The delta height values usually get sudden plummet, dropping down to around -0.5 and more, while for normal falling, delta height is around -0.2 to -0.4.  This usually happened when people are standing close to the sensor (only 1 in the boundary box, so that fall detection works)

    Do you know why this occurs and if so, how can I stop this from happening? 

    Thank you for your help.

  • +1 in reply to Danh Pham
    TI__Genius 14285 points

    Hello,

    The fall detection logic is fairly simple, and meant as an example, so it doesn't try to account for any of these types of errors. Improvements can surely be made, I have 2 suggestions.

    These errors also seem to happen when the track is lost, usually when someone is outside the static boundary box. The visualizer right now will report that as zero height and lose the height history. The example is only valid when tracks are actively moving. This could be improved in your implementation.

    The elevation angle measurement (height of the person) will be difficult at close range. The demo uses the centroid point of the output track to estimate height, so if this changes because some points are out of the FOV, it may change quickly. You can modify the code to use the talles point in the pointcloud associated with the track instead of the centroid point. This may give you a better estimate of height and might not change so quickly.

    Regards,

    Jackson