Using AWR1843BOOST to detect animals

[guruvishnuvardan]

Dear Ibai,

Good Day!

Can you please let me know, if I can use AWR1843BOOST for detecting animals on the road? (example: Cattle).

Thanks
Guru

[mhatch06]

Not Ibai, but hopefully I can provide some help.

It depends on your implementation and needs, but you can definitely use this to detect 'objects' on the road(within a certain range), but if you need to differentiate animals from other objects(e.g. cars, people, etc) I think that an actual camera would be better for your case.

[guruvishnuvardan]

Not Ibai, but hopefully I can provide some help.

It depends on your implementation and needs, but you can definitely use this to detect 'objects' on the road(within a certain range), but if you need to differentiate animals from other objects(e.g. cars, people, etc) I think that an actual camera would be better for your case.

@mhatch06 --- Thanks a ton for your advise.

[guruvishnuvardan]

@mhatch06 -- Can you please clarify me the following

Can we detect larger animals like Cows based on Signal to Noise ratio ? Or Can we detect the presence of animals based on Vital Signs (using range Doppler)

Please correct me, if any of my assumptions are wrong.

Thanks
Guru

[guruvishnuvardan]

@mhatch06

@ibaiGorordo

Can you please clarify my above comments.

[adithyasunil26]

@guruvishnuvardan I did attempt object detection and classification using the IWR1843BOOST board.

As @mhatch06 suggested, FMCW radars are great for object or obstacle detection, but when it comes to classification, it would be better to go with a camera, lidar or some form of sensor fusion rather than depending on the radar board.

From my experiments and observations, the point clouds generated from radar boards aren't dense enough to actually classify objects that aren't very different. For example, if you were to classify between a person and a truck, it might work well, but if u want to classify between a hatchback and a sedan, it probably won't. So, if you are dealing with objects that are very different from each other in size and shape, lets say a cow and a person, then it makes sense to proceed with the radar-based approach, but if you need high accuracy and the objects aren't very characteristically different like a cow and other animals of similar characteristic features, then you will need to re-evaluate your approach.