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.

AWR1642: Training new objects

Part Number: AWR1642

Hi,

I am so excited to use MMWAVE solutions for some other use-cases apart from the automotive and industrial applications. As we are closely working with the forest department, we are planning to use it to detect animals especially elephants and wild boars. How to train new objects/animals and improve the range up-to 500m. Highly appriciate if you can share the new object training tutorial and reference links available.

Thanks in advance.

Best Regards,

Krishna 

  • Hello Krishna,

    Typically object classification is done by measuring the data from these objects in different scenarios and looking for possible signatures that can help clarify them. For example the signatures could be RCS (reflected signal level), implement clustering to recognize the size of the object, velocity of the object etc. and using this information in the processing algorithum. We do not have exact examples to show classification as such.

    Getting 500m range for animals could be a challenge. You might need a very high gain antennas, but that would make them very directional. ie you field of view would be very narrow.

    regards,
    Vivek

  • Hello Vivek,

    Thanks for your valuable reply. What you said is correct. From object size, velocity and distance we can approximate to the classification. Do IWR 1642 BOOST, give us these parameters  - size, velocity and distance via SPI?

     Thanks and best regards,

    Krishna 

  • Hello Krishna,

    The mmwave SDK has a mmw demo application for 1642 device that does the object range, velocity computation and sends the data out over the UART.

    To determine 'size' you would have to implement so kind of clustering algorithm using the multiple points from a single object and use that estimate the size.

    The SRR TI design has some clustering implemented : http://www.ti.com/tool/TIDEP-0092. You can use that as a starting point for your application development.

    Regards,

    Vivek