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.

IWR1843BOOST: Discrepency between IWR1843BOOST and IWR1642BOOST's traffic monitoring demo

Part Number: IWR1843BOOST
Other Parts Discussed in Thread: IWR1843, IWR1642

Hi I have purchased both the IWR1642 and the IWR1843 traffic monitoring demo, However the performence between these two are very different and so is the interface, here are the description of these discrepencies and some questions I have.

The IWR1843 demo in Industrial toolbox 4.5.1 detects a lot of objects that aren't cars espacially when I am moving with the radar , The screenshot below is when I put the radar stationary and having a car passing by, as you can see the passing of the car triggered the radar to detect a lot of different tracked objects that are neither there or moving. while it one of the tracked objects IS indeed the movement of the car, a lot of extra unwanted objects appears:

However for the IWR1642's demo in toolbox version 4.1.0, it doesn't detect any of the random objects as cars and only shows the actual cars, however it is not very reliable, as shown in the picture below it detects the car's orientation, however, the car should be a lot more to the left than its position on the picture:

I am wondering what exactly is causing such a discrepency between the two versions? are they using two different tracking algorithms, if so how are they different? Also how is it that the IWR1642 demo can see the orientation of the car while the IWR1843 can only see a circle as the car? And is it possible to have the IWR1843 demo's correct object pattern without all those unwanted falsely tracked objects?

Thanks!

  • Hello

    Can you please confirm which versions of demo are you using for the two EVM's

    Our latest demos have one Software which can be compiled across platforms to get the required image for targeted device - are you using the same software code to compile for both the devices.

    Thank you,

    Vaibhav

  • No I am using different pre-build binaries for the two, as mentioned in my question, I am using the toolbox 4.5.1 for IWR1843 and 4.1.0 for IWR1642, since in the newest toolbox it only supports 18xx and 68xx and 4.1.0 is the last one with 16xx support

  • Hello

    In order to get identical performance we would recommend using the same code and binary generated from that specific to the target device.

    Difference in detection and tracker contribute to differences.

    Please see how you can tune the thresholds on the demo in order to change the detection for cars. Have you tried raising the thresholds.

    Thank you,

    Vaibhav

  • Hi thanks for the reply, I understand that but 16xx is not supported in the newest toolbox and I wanted to get the best performence for 18xx that is why I used the newest version of the toolbox for 18xx, but regardless there is a huge difference in the performence as mentioned and my question was what major change occurred between the two versions? Also I noticed that the Gtrack tuning guide is no longer present in the newest toolbox (ie. 4.5.1) I am wondering why that is the case, do the new toolbox not use those algorithms mentioned in there?

    Which thresholds are you referring to? There seems a lot and I have changed a few of them but probably not the ones you have in mind in your reply

    Thanks

  • Hello

    Please take a look at the tuning guide called out in the following  Traffic monitoring guide.

    Allocation Params and allied sections would be of help. 

    Increase/decrease number of points in detection data -- CFAR threshold - please see SDK USer;s guide

    CHange attribution of points to generate tracks  - Tuning guide - Mentioned in Tirex.

    THank you,

    Vaibhav