IWR6843: Differentiate human from other objects

Sowmya P

Prodigy 60 points

Part Number: IWR6843

Hi,

Is there a way we can differentiate between a person and other objects detected by sensor/radar?

Thanks and Regards,

Sowmya P

over 5 years ago

0 Akash Gondalia over 5 years ago

TI__Guru 50240 points

Hi Sowmya,

I've asked an expert to look into this and we should have an answer for you in the next few days.

Cheers,

Akash

0 zigang Yang over 5 years ago

TI__Mastermind 40115 points

Hi, Sowmya:

Can you explain your scenario a little bit more?

People will never completely static, they always move a little. If other objects are static objects, then you can remove static object and what left is people. You can look at the people counting demo for this idea. http://dev.ti.com/tirex/explore/node?node=ANoGbRd1GQzjC9YA7IUpPQ__VLyFKFf__LATEST

If you you are trying to differentiate moving fans, moving trees branches from walking people, then you need to consider classification. You may need to use some micro-doppler information to help classification. You can look this lab with classification integrated: http://dev.ti.com/tirex/explore/node?node=AC63JY6F3gmrVCyLfpS.GQ__VLyFKFf__LATEST.

Best,

Zigang

0 Sowmya P over 5 years ago in reply to zigang Yang

Prodigy 60 points

Hi Zigang,

Thanks for your reply.

The scenario is similar to the second scenario that you have explained. We want to detect humans in a room and eliminate all other objects as shown in this demo http://dev.ti.com/tirex/explore/node?node=AC63JY6F3gmrVCyLfpS.GQ__VLyFKFf__LATEST. But when we run this demo at our side, it is not as precise as it is shown in this link. It is detecting other objects as humans and displays multiple color circles when only one person is in the vicinity. How to rectify these issues?

Regards,

Sowmya P

0 Justin Curewitz over 5 years ago in reply to Sowmya P

TI__Mastermind 28055 points

Hi Sowmya,

We have had the best success by putting a filter on the output of the classifier that will require an object to be classified as human a certain percentage of the time before accepting that as the classification. I also recommend using this configuration for the classifier:

classifierCfg 1 1 3 500 0.8 1.0 0.95 10

Replace the classifier config line in your chirp config with the one above.

Regards,

Justin

Sensors

Sensors forum

IWR6843: Differentiate human from other objects