Pulse Induction metal detector with AI [idea]

[eclipse]

I have this crazy idea of adding AI features to pulse induction metal detector:
*) AI analyses signals and applies filter based on network knowledge uploaded to central database
*) Users are able to work with AI features like if they dig some signal they can enter what they have found like 1) metal JUNK 2)Good target 3) mineralization (no real target) and so on
*) When they get home they upload their target database to the server which processes the information (if it determines it adds to the database which can be reused by other users).

Applied for metal detectors (PI) with some extra processing of the signal like VoooDoo or Multiple period sampling or similar (plain PIs don't have good enough data to work with).

What do you think ?

[Detectorist#1]

Hi eclipse,
This idea is not so crazy. MD creates file with current GPS coordinates and data from direct sampling ( with very fast 24bits ADC) and this information will be processed by AI. The result will be owned by who?

[eclipse]

No detectorist I know will want to share his GPS coordinates /must be kept private/. I was thinking more like no GPS information to be collected /unless user agrees with this but must be kept private/.
Just full target response (10-20) samples to help create predetermined target filter response.
A ton of free information if you will. No company has an army of detectorists working day and night for free, worldwide.
There will be a point where database is so full that no more useful information needs to be gathered.
Result is owned by the guy/s that made the whole crazy thing going.

[eclipse]

Database could contain 100s thousands of target responses (stored on the API server)
Those will be processed / crunched by similarity to some low number like 20... or even 100 .. 200 or whatever the metal detector can handle.
Target response will be compared with the database based on similarity points (where the AI comes in - it will have to determine which response is similar to our current response) in real time... (the hard part)

[Detectorist#1]

The hope ever gain!

[lucifer]

That would require many people working with the same device. Even so, they will operate in different locations, e.g. different ground conditions, different noise conditions, likely different coils as well.. so it would be difficult to get standardized responses to the database.

[Detectorist#1]

Hi lucufer,
Don't warry - for now it's only a dream!

[moodz]

... and this guy has been reading your mind

https://patents.google.com/patent/US...oq=20240078408

[Detectorist#1]

Ha-ha-ha. Very now days patent! This patent claims for any kind of input data-base, for any kind of data processing (well learned machine) and for any kind of output information. This is the end new ideas - this is the final patent in PI metal detecting!!! Ha-ha-ha.

[Aziz]

Ha-ha-ha. Very now days patent! This patent claims for any kind of input data-base, for any kind of data processing (well learned machine) and for any kind of output information. This is the end new ideas - this is the final patent in PI metal detecting!!! Ha-ha-ha.

Yep! AI hype is feeding the patent trolls.
I'm sure, the patent holder does not know how to realise such a technolgy.

Btw, we don't need AI for such a trivial task.
Aziz