AI Detector

[The Beep Goes On]

Asked Grok about AI detecting. The root benefit is training on large signal data sets that make probable ID more accurate - might help in some cases - if it can tell the difference between a gold ring and pull tab with the same signal, consistently, it may be interesting.

~~~

AI will transform hobby metal detecting from a skill-heavy, trial-and-error pursuit into a more intuitive, efficient, and rewarding experience, while preserving the core thrill of discovery.

Operational Changes​

• Smarter Target Identification: Traditional detectors rely on tones, basic IDs, and user interpretation to distinguish trash (like pull tabs) from treasure (coins, rings, relics). AI integration, using machine learning trained on vast signal datasets, will analyze coil data in real-time—factoring in ground conditions, EMI interference, depth, and multi-frequency responses—to provide highly accurate classifications (e.g., "95% likelihood: silver coin" or "gold ring"). This reduces false positives and unnecessary digs, as seen in emerging concepts where AI achieves near-expert-level discrimination.
• App Integration and Connectivity: The detector will pair seamlessly with a smartphone app via Bluetooth/Wi-Fi. AI processes data on-device or in the cloud, offering live feedback, automatic settings adjustments (e.g., optimizing for soil mineralization), and predictive suggestions like "strong signal pattern matches known relic hotspots."
• Mapping and Data Logging: Built-in GPS combined with AI will auto-map scans, create 3D visualizations of searched areas, highlight probable targets, and even suggest optimal grid patterns or return paths based on prior finds. Over time, the AI learns your preferences and local sites, improving efficiency.
• Augmented Reality (AR) Overlays: Future models could project AR via phone or headset, overlaying virtual markers on the ground for detected objects, showing estimated shape/depth before digging.

Experiential Changes​

• More Accessible for Beginners: New users often struggle with interpreting signals and settings. AI acts as an on-device coach, explaining detections ("This tone indicates iron—likely trash"), suggesting adjustments, and building confidence faster—lowering the entry barrier while accelerating the learning curve.
• Deeper Engagement for Experts: Veterans get advanced insights, like pattern recognition across hunts ("This site resembles Civil War-era camps—prioritize ferrous signals"), historical context integration (cross-referencing finds with databases), or community-shared AI models for specific regions.
• Increased Success and Fun: Fewer junk digs mean more meaningful finds, sustaining motivation. The hobby becomes gamified—apps could track stats, award "achievements," or integrate social features to share mapped hunts ethically.
• Immersive and Relaxing: Voice guidance, haptic feedback, or AR turns solitary sweeps into an interactive adventure, blending outdoor exploration with tech-assisted treasure hunting.

Overall, AI won't replace the joy of fresh air, exercise, and unearthing history—it will amplify it, making every outing more productive and personalized. As of late 2025, we're seeing AI breakthroughs in professional/security detectors, and hobby models will follow soon, potentially revolutionizing the field within the next 5–10 years.

Traditional Target Identification in Metal Detectors​

In hobby metal detectors, target identification (Target ID or TID) is the process of determining what a buried object might be before digging. It relies on the electromagnetic signals returned from the metal object when the detector's coil passes over it.

Key elements:

• Conductivity: High-conductive metals (e.g., silver, copper coins) produce stronger, higher signals. Low-conductive ones (e.g., gold, nickel) produce weaker signals.
• Ferrous vs. Non-Ferrous: Iron and steel (ferrous) often give lower tones or negative IDs due to their magnetic properties.
• Visual Display: Most modern detectors show a numerical Target ID (e.g., 0-99 scale). For example:
  • Iron/foil: Low numbers (0-40)
  • Pull tabs/nickel: Mid (40-60)
  • Coins (quarters, dimes): High (80-95)
• Audio Tones: Low tone for trash (iron), high for valuables (coins).
• Limitations: Influenced by depth, orientation, soil mineralization, nearby trash, and EMI (electromagnetic interference). Signals can fluctuate, leading to "iffy" reads and false positives (e.g., pull tabs mimicking gold rings).

Traditional systems use fixed algorithms and multi-frequency processing (like Minelab's Multi-IQ or Nokta's SMF) for better separation, but interpretation still requires user experience.

How AI Enhances Target Identification​

AI, particularly machine learning, takes target ID to the next level by analyzing complex data patterns in real-time that traditional algorithms struggle with. Here's how it works:

1. Data Processing:
   • The detector collects raw signals from multiple sweeps (e.g., 30+ passes over a target).
   • AI algorithms factor in variables like ground conditions, mineralization, depth, EMI, signal decay, shape hints, and multi-frequency responses.
2. Pattern Recognition and Learning:
   • Trained on vast datasets of known targets (coins, rings, trash in various soils).
   • Machine learning models (e.g., neural networks) recognize subtle signatures—distinguishing a gold ring from an aluminum pull tab despite similar conductivity.
   • Reduces false alarms by "learning" to ignore interfering signals (e.g., product effect in industrial detectors, adaptable to hobby use).
3. Improved Accuracy:
   • Provides probabilistic IDs (e.g., "92% silver coin" instead of just a number).
   • Better discrimination in trashy or mineralized sites—unmasking good targets near iron.
   • Adaptive: The detector can "learn" from your hunts, refining suggestions over time (e.g., local soil patterns).
4. User Benefits:
   • Fewer junk digs → More efficient hunts.
   • Beginner-friendly: Acts like an expert coach with voice/text feedback.
   • Advanced features: 2D ID maps (like Minelab Manticore's ferrous/conductive plot), predictive suggestions, or AR overlays.

​

Current State (as of December 2025)​

True AI/machine learning is emerging more in professional/security detectors (e.g., Garrett Paragon's 2025 ML update for better threat discrimination; Sesotec THiNK for industrial food inspection). In hobby models:

• Advanced processing mimics AI benefits (e.g., Minelab Manticore's enhanced 2D ID and target trace; Nokta Legend's FerroCheck and stability updates).
• Discussions in communities highlight AI's potential for near-perfect trash rejection.
• Full AI integration (on-device learning, cloud-trained models) is on the horizon, with prototypes and investments focusing on smarter hobby detectors within 5-10 years.

AI won't eliminate the need for skill entirely, but it will make target ID far more reliable, turning "guesswork" into confident decisions while keeping the excitement of the dig alive.

 

[Chilli]

Asked Grok about AI detecting. The root benefit is training on large signal data sets that make probable ID more accurate - might help in some cases - if it can tell the difference between a gold ring and pull tab with the same signal, consistently, it may be interesting.

~~~

AI will transform hobby metal detecting from a skill-heavy, trial-and-error pursuit into a more intuitive, efficient, and rewarding experience, while preserving the core thrill of discovery.

Operational Changes​

• Smarter Target Identification: Traditional detectors rely on tones, basic IDs, and user interpretation to distinguish trash (like pull tabs) from treasure (coins, rings, relics). AI integration, using machine learning trained on vast signal datasets, will analyze coil data in real-time—factoring in ground conditions, EMI interference, depth, and multi-frequency responses—to provide highly accurate classifications (e.g., "95% likelihood: silver coin" or "gold ring"). This reduces false positives and unnecessary digs, as seen in emerging concepts where AI achieves near-expert-level discrimination.
• App Integration and Connectivity: The detector will pair seamlessly with a smartphone app via Bluetooth/Wi-Fi. AI processes data on-device or in the cloud, offering live feedback, automatic settings adjustments (e.g., optimizing for soil mineralization), and predictive suggestions like "strong signal pattern matches known relic hotspots."
• Mapping and Data Logging: Built-in GPS combined with AI will auto-map scans, create 3D visualizations of searched areas, highlight probable targets, and even suggest optimal grid patterns or return paths based on prior finds. Over time, the AI learns your preferences and local sites, improving efficiency.
• Augmented Reality (AR) Overlays: Future models could project AR via phone or headset, overlaying virtual markers on the ground for detected objects, showing estimated shape/depth before digging.

Experiential Changes​

• More Accessible for Beginners: New users often struggle with interpreting signals and settings. AI acts as an on-device coach, explaining detections ("This tone indicates iron—likely trash"), suggesting adjustments, and building confidence faster—lowering the entry barrier while accelerating the learning curve.
• Deeper Engagement for Experts: Veterans get advanced insights, like pattern recognition across hunts ("This site resembles Civil War-era camps—prioritize ferrous signals"), historical context integration (cross-referencing finds with databases), or community-shared AI models for specific regions.
• Increased Success and Fun: Fewer junk digs mean more meaningful finds, sustaining motivation. The hobby becomes gamified—apps could track stats, award "achievements," or integrate social features to share mapped hunts ethically.
• Immersive and Relaxing: Voice guidance, haptic feedback, or AR turns solitary sweeps into an interactive adventure, blending outdoor exploration with tech-assisted treasure hunting.

Overall, AI won't replace the joy of fresh air, exercise, and unearthing history—it will amplify it, making every outing more productive and personalized. As of late 2025, we're seeing AI breakthroughs in professional/security detectors, and hobby models will follow soon, potentially revolutionizing the field within the next 5–10 years.

Traditional Target Identification in Metal Detectors​

In hobby metal detectors, target identification (Target ID or TID) is the process of determining what a buried object might be before digging. It relies on the electromagnetic signals returned from the metal object when the detector's coil passes over it.

Key elements:

• Conductivity: High-conductive metals (e.g., silver, copper coins) produce stronger, higher signals. Low-conductive ones (e.g., gold, nickel) produce weaker signals.
• Ferrous vs. Non-Ferrous: Iron and steel (ferrous) often give lower tones or negative IDs due to their magnetic properties.
• Visual Display: Most modern detectors show a numerical Target ID (e.g., 0-99 scale). For example:
  • Iron/foil: Low numbers (0-40)
  • Pull tabs/nickel: Mid (40-60)
  • Coins (quarters, dimes): High (80-95)
• Audio Tones: Low tone for trash (iron), high for valuables (coins).
• Limitations: Influenced by depth, orientation, soil mineralization, nearby trash, and EMI (electromagnetic interference). Signals can fluctuate, leading to "iffy" reads and false positives (e.g., pull tabs mimicking gold rings).

Traditional systems use fixed algorithms and multi-frequency processing (like Minelab's Multi-IQ or Nokta's SMF) for better separation, but interpretation still requires user experience.

How AI Enhances Target Identification​

AI, particularly machine learning, takes target ID to the next level by analyzing complex data patterns in real-time that traditional algorithms struggle with. Here's how it works:

1. Data Processing:
   • The detector collects raw signals from multiple sweeps (e.g., 30+ passes over a target).
   • AI algorithms factor in variables like ground conditions, mineralization, depth, EMI, signal decay, shape hints, and multi-frequency responses.
2. Pattern Recognition and Learning:
   • Trained on vast datasets of known targets (coins, rings, trash in various soils).
   • Machine learning models (e.g., neural networks) recognize subtle signatures—distinguishing a gold ring from an aluminum pull tab despite similar conductivity.
   • Reduces false alarms by "learning" to ignore interfering signals (e.g., product effect in industrial detectors, adaptable to hobby use).
3. Improved Accuracy:
   • Provides probabilistic IDs (e.g., "92% silver coin" instead of just a number).
   • Better discrimination in trashy or mineralized sites—unmasking good targets near iron.
   • Adaptive: The detector can "learn" from your hunts, refining suggestions over time (e.g., local soil patterns).
4. User Benefits:
   • Fewer junk digs → More efficient hunts.
   • Beginner-friendly: Acts like an expert coach with voice/text feedback.
   • Advanced features: 2D ID maps (like Minelab Manticore's ferrous/conductive plot), predictive suggestions, or AR overlays.

​

Current State (as of December 2025)​

True AI/machine learning is emerging more in professional/security detectors (e.g., Garrett Paragon's 2025 ML update for better threat discrimination; Sesotec THiNK for industrial food inspection). In hobby models:

• Advanced processing mimics AI benefits (e.g., Minelab Manticore's enhanced 2D ID and target trace; Nokta Legend's FerroCheck and stability updates).
• Discussions in communities highlight AI's potential for near-perfect trash rejection.
• Full AI integration (on-device learning, cloud-trained models) is on the horizon, with prototypes and investments focusing on smarter hobby detectors within 5-10 years.

AI won't eliminate the need for skill entirely, but it will make target ID far more reliable, turning "guesswork" into confident decisions while keeping the excitement of the dig alive.

Yep one would think that AI in detecting is inevitable.
The main negative for me is it logging my finds and location and that being accessible being it will be internet connected.
I think by the time its rolled out ill have hung up my detector though so probably not gonna be an issue to me.

 

[AstralDruid]

In my book , the above jargon translates to AI will effectively remove all fun from our hobby, and introduce the element of being snooped upon by them that snoop............
I will never own such a piece of equipment...

 

[robertk]

AI won't eliminate the need for skill entirely, but it will make target ID far more reliable, turning "guesswork" into confident decisions while keeping the excitement of the dig alive.

I'm not holding my breath. My interactions with AI so far have been pretty underwhelming. It is very often wrong, and sometimes just completely makes stuff up. The only thing it would be useful for in this situation is pattern matching -- matching the details of the signal data received to the actual target found. But there's a pretty big real world gap at the end there -- you have to give it the details of what you found. What kind of metal, how big, how much corrosion, what shape, what orientation in the ground, and so on. Without that data to match up with the signal data, it's still going to be guessing just as much as the human operator would.

What I'd like to see is a detector with ground penetrating radar built into it, so you could (sort of) "see" the target before you dig it up. That might reduce the amount of trash dug. Or it might not. I'd probably still dig everything. Except maybe nails. I think I have enough of those.

 

[Chilli]

What I'd like to see is a detector with ground penetrating radar built into it, so you could (sort of) "see" the target before you dig it up. That might reduce the amount of trash dug. Or it might not. I'd probably still dig everything. Except maybe nails. I think I have enough of those.

I know that Minelab was having a crack at that about 20 yrs ago. Wether they are still pursuing it today I don't know. Id put money on them being the first to release one if it ever did eventuate.

 

[The Beep Goes On]

Yep one would think that AI in detecting is inevitable.
The main negative for me is it logging my finds and location and that being accessible being it will be internet connected.
I think by the time its rolled out ill have hung up my detector though so probably not gonna be an issue to me.

Yeah, me too. I've been a programmer for 40 years; got out just in time for the AI takeover. The connection has potential security as well as response time issues, unless the training can be distilled down into a chip/memory and be local only - connect for updates ... who knows.

 

[The Beep Goes On]

In my book , the above jargon translates to AI will effectively remove all fun from our hobby, and introduce the element of being snooped upon by them that snoop............
I will never own such a piece of equipment..

I hear you ... snooping is one thing, relinquishing dig decisions to a machine is another. The V3i does this - provides likely target and probability - but it's hard-coded without machine learning/training. Unless AI can dig for you we'll always have the last word.

 

[The Beep Goes On]

I'm not holding my breath. My interactions with AI so far have been pretty underwhelming. It is very often wrong, and sometimes just completely makes stuff up. The only thing it would be useful for in this situation is pattern matching -- matching the details of the signal data received to the actual target found. But there's a pretty big real world gap at the end there -- you have to give it the details of what you found. What kind of metal, how big, how much corrosion, what shape, what orientation in the ground, and so on. Without that data to match up with the signal data, it's still going to be guessing just as much as the human operator would.

What I'd like to see is a detector with ground penetrating radar built into it, so you could (sort of) "see" the target before you dig it up. That might reduce the amount of trash dug. Or it might not. I'd probably still dig everything. Except maybe nails. I think I have enough of those.

Yes, the ability to quickly see what's there with decent resolution would kind of remove the need for all the complexity. AI, as you say, is crap in a lot of ways - it does make shit up, experienced that many times. There's an infinite number of signals so there's always going to be some guessing.

 

[The Beep Goes On]

I know that Minelab was having a crack at that about 20 yrs ago. Wether they are still pursuing it today I don't know. Id put money on them being the first to release one if it ever did eventuate.

It may be here sooner than specified... AI chips are becoming prevalent. With Minelab's de-mining and other efforts that could possibly benefit from AI you're probably correct.

 

[Terry Soloman]

Anything is possible.

 

[philber]

I'm not holding my breath. My interactions with AI so far have been pretty underwhelming. It is very often wrong, and sometimes just completely makes stuff up. The only thing it would be useful for in this situation is pattern matching -- matching the details of the signal data received to the actual target found. But there's a pretty big real world gap at the end there -- you have to give it the details of what you found. What kind of metal, how big, how much corrosion, what shape, what orientation in the ground, and so on. Without that data to match up with the signal data, it's still going to be guessing just as much as the human operator would.

What I'd like to see is a detector with ground penetrating radar built into it, so you could (sort of) "see" the target before you dig it up. That might reduce the amount of trash dug. Or it might not. I'd probably still dig everything. Except maybe nails. I think I have enough of those.

Thats kind of what the nokta Invenio is supposed to do. And uses AI

 

[michael NY]

Asked Grok about AI detecting. The root benefit is training on large signal data sets that make probable ID more accurate - might help in some cases - if it can tell the difference between a gold ring and pull tab with the same signal, consistently, it may be interesting.

~~~

AI will transform hobby metal detecting from a skill-heavy, trial-and-error pursuit into a more intuitive, efficient, and rewarding experience, while preserving the core thrill of discovery.

Operational Changes​

• Smarter Target Identification: Traditional detectors rely on tones, basic IDs, and user interpretation to distinguish trash (like pull tabs) from treasure (coins, rings, relics). AI integration, using machine learning trained on vast signal datasets, will analyze coil data in real-time—factoring in ground conditions, EMI interference, depth, and multi-frequency responses—to provide highly accurate classifications (e.g., "95% likelihood: silver coin" or "gold ring"). This reduces false positives and unnecessary digs, as seen in emerging concepts where AI achieves near-expert-level discrimination.
• App Integration and Connectivity: The detector will pair seamlessly with a smartphone app via Bluetooth/Wi-Fi. AI processes data on-device or in the cloud, offering live feedback, automatic settings adjustments (e.g., optimizing for soil mineralization), and predictive suggestions like "strong signal pattern matches known relic hotspots."
• Mapping and Data Logging: Built-in GPS combined with AI will auto-map scans, create 3D visualizations of searched areas, highlight probable targets, and even suggest optimal grid patterns or return paths based on prior finds. Over time, the AI learns your preferences and local sites, improving efficiency.
• Augmented Reality (AR) Overlays: Future models could project AR via phone or headset, overlaying virtual markers on the ground for detected objects, showing estimated shape/depth before digging.

Experiential Changes​

• More Accessible for Beginners: New users often struggle with interpreting signals and settings. AI acts as an on-device coach, explaining detections ("This tone indicates iron—likely trash"), suggesting adjustments, and building confidence faster—lowering the entry barrier while accelerating the learning curve.
• Deeper Engagement for Experts: Veterans get advanced insights, like pattern recognition across hunts ("This site resembles Civil War-era camps—prioritize ferrous signals"), historical context integration (cross-referencing finds with databases), or community-shared AI models for specific regions.
• Increased Success and Fun: Fewer junk digs mean more meaningful finds, sustaining motivation. The hobby becomes gamified—apps could track stats, award "achievements," or integrate social features to share mapped hunts ethically.
• Immersive and Relaxing: Voice guidance, haptic feedback, or AR turns solitary sweeps into an interactive adventure, blending outdoor exploration with tech-assisted treasure hunting.

Overall, AI won't replace the joy of fresh air, exercise, and unearthing history—it will amplify it, making every outing more productive and personalized. As of late 2025, we're seeing AI breakthroughs in professional/security detectors, and hobby models will follow soon, potentially revolutionizing the field within the next 5–10 years.

Traditional Target Identification in Metal Detectors​

In hobby metal detectors, target identification (Target ID or TID) is the process of determining what a buried object might be before digging. It relies on the electromagnetic signals returned from the metal object when the detector's coil passes over it.

Key elements:

• Conductivity: High-conductive metals (e.g., silver, copper coins) produce stronger, higher signals. Low-conductive ones (e.g., gold, nickel) produce weaker signals.
• Ferrous vs. Non-Ferrous: Iron and steel (ferrous) often give lower tones or negative IDs due to their magnetic properties.
• Visual Display: Most modern detectors show a numerical Target ID (e.g., 0-99 scale). For example:
  • Iron/foil: Low numbers (0-40)
  • Pull tabs/nickel: Mid (40-60)
  • Coins (quarters, dimes): High (80-95)
• Audio Tones: Low tone for trash (iron), high for valuables (coins).
• Limitations: Influenced by depth, orientation, soil mineralization, nearby trash, and EMI (electromagnetic interference). Signals can fluctuate, leading to "iffy" reads and false positives (e.g., pull tabs mimicking gold rings).

Traditional systems use fixed algorithms and multi-frequency processing (like Minelab's Multi-IQ or Nokta's SMF) for better separation, but interpretation still requires user experience.

How AI Enhances Target Identification​

AI, particularly machine learning, takes target ID to the next level by analyzing complex data patterns in real-time that traditional algorithms struggle with. Here's how it works:

1. Data Processing:
   • The detector collects raw signals from multiple sweeps (e.g., 30+ passes over a target).
   • AI algorithms factor in variables like ground conditions, mineralization, depth, EMI, signal decay, shape hints, and multi-frequency responses.
2. Pattern Recognition and Learning:
   • Trained on vast datasets of known targets (coins, rings, trash in various soils).
   • Machine learning models (e.g., neural networks) recognize subtle signatures—distinguishing a gold ring from an aluminum pull tab despite similar conductivity.
   • Reduces false alarms by "learning" to ignore interfering signals (e.g., product effect in industrial detectors, adaptable to hobby use).
3. Improved Accuracy:
   • Provides probabilistic IDs (e.g., "92% silver coin" instead of just a number).
   • Better discrimination in trashy or mineralized sites—unmasking good targets near iron.
   • Adaptive: The detector can "learn" from your hunts, refining suggestions over time (e.g., local soil patterns).
4. User Benefits:
   • Fewer junk digs → More efficient hunts.
   • Beginner-friendly: Acts like an expert coach with voice/text feedback.
   • Advanced features: 2D ID maps (like Minelab Manticore's ferrous/conductive plot), predictive suggestions, or AR overlays.

​

Current State (as of December 2025)​

True AI/machine learning is emerging more in professional/security detectors (e.g., Garrett Paragon's 2025 ML update for better threat discrimination; Sesotec THiNK for industrial food inspection). In hobby models:

• Advanced processing mimics AI benefits (e.g., Minelab Manticore's enhanced 2D ID and target trace; Nokta Legend's FerroCheck and stability updates).
• Discussions in communities highlight AI's potential for near-perfect trash rejection.
• Full AI integration (on-device learning, cloud-trained models) is on the horizon, with prototypes and investments focusing on smarter hobby detectors within 5-10 years.

AI won't eliminate the need for skill entirely, but it will make target ID far more reliable, turning "guesswork" into confident decisions while keeping the excitement of the dig alive.

Yes i asked ai same question kinda funny i was wondering how many years out is this technology before consumer sees it?

 

[sibbley]

Thats kind of what the nokta Invenio is supposed to do. And uses AI

From what I've seen so far the Invenio isn't very good at doing what it's supposed to do. At least not on smaller targets anyway.

 

[99thpercentile]

Far more advanced metal detectors with automatic target recognition have existed for 20+ years at this point. In the geophysics world when looking for unexploded ordnance (UXO) we utilize a process called advanced geophysical classification (AGC). Even pre 2010 under the AGC process we as an industry achieved a 100% receiver operating characteristic (ROC). This means that on an entire site 100% of the UXO in the ground were detected, all of the targets were correctly discriminated as either UXO or non-UXO, and all of the items correctly discriminated as UXO were then correctly classified as the correct model, at the correct location, correct depth, inclination, and declination.

These same advanced time domain electromagnetic induction (TDEM) metal detectors can be used to detect/discriminate/classify any know objects like coins the same way. Several of the instruments did this for demonstration.

You don't need AI for this, you just need more money than most hobbyists want to spend.

 

[The Beep Goes On]

Far more advanced metal detectors with automatic target recognition have existed for 20+ years at this point. In the geophysics world when looking for unexploded ordnance (UXO) we utilize a process called advanced geophysical classification (AGC). Even pre 2010 under the AGC process we as an industry achieved a 100% receiver operating characteristic (ROC). This means that on an entire site 100% of the UXO in the ground were detected, all of the targets were correctly discriminated as either UXO or non-UXO, and all of the items correctly discriminated as UXO were then correctly classified as the correct model, at the correct location, correct depth, inclination, and declination.

These same advanced time domain electromagnetic induction (TDEM) metal detectors can be used to detect/discriminate/classify any know objects like coins the same way. Several of the instruments did this for demonstration.

You don't need AI for this, you just need more money than most hobbyists want to spend.

Would be interesting to somehow scale down TDEM with a modified PI detector using different selectable parameter sets for various target types. You'd think someone would have tried this, so assumption is that there are limiting factors.

 

[geowizard]

Would be interesting to somehow scale down TDEM with a modified PI detector using different selectable parameter sets for various target types. You'd think someone would have tried this, so assumption is that there are limiting factors.

The limiting factor:

TDEM (Time Domain EM) and PI (Pulse Induction) are based on the same concept, pulses! No discrimination of metal type!

FDEM (Frequency Domain EM) uses a method that provides "discrimination" of targets according to metal type!

It HAS been tried and tested. There are THINK TANKS in the industry. Have been for a long time, ROOMS full of THINKERS paid for thinking about the POSSIBILITIES that haven't been thunk!

Stick around! There's more!

- Geowizard

 

[geowizard]

Is AI "Intelligence"?

The short answer is "No".

Intelligence is defined as "The ability to acquire and APPLY knowledge or skills."

Think tanks leverage knowledge of many deep thinkers to extrapolate or expand the known into the unknown. Unknown naturally covers new discoveries, new ideas.

- Geowizard

 

[geowizard]

Far more advanced metal detectors with automatic target recognition have existed for 20+ years at this point. In the geophysics world when looking for unexploded ordnance (UXO) we utilize a process called advanced geophysical classification (AGC). Even pre 2010 under the AGC process we as an industry achieved a 100% receiver operating characteristic (ROC). This means that on an entire site 100% of the UXO in the ground were detected, all of the targets were correctly discriminated as either UXO or non-UXO, and all of the items correctly discriminated as UXO were then correctly classified as the correct model, at the correct location, correct depth, inclination, and declination.

These same advanced time domain electromagnetic induction (TDEM) metal detectors can be used to detect/discriminate/classify any know objects like coins the same way. Several of the instruments did this for demonstration.

You don't need AI for this, you just need more money than most hobbyists want to spend.

Can you cite just one tangible reference please?

Thanks,

- Geowizard

 

[Carl-NC]

TDEM (Time Domain EM) and PI (Pulse Induction) are based on the same concept, pulses! No discrimination of metal type!

Tony Barringer had a discriminating pulse design in the 1960s, for aerial surveys at that. I built a handheld hobby version when I was at White's.

 

[geowizard]

Awesome. You are a genuine asset to TreasureNet and the Metal Detecting Community at large!

I have read every available Document, technical paper, including Canadian and US papers patent on Barringer going back as far as records go. I am not aware of "discrimination" per se related to PI or TDEM. Those (Pulse) "INPUT" systems back in the day, preceded modern age discrimination of metals in the frequency domain.

Without IQ, in the Frequency Domain, discrimination of metals is not possible as a product of phase unless a time domain pulse is transformed to frequency or frequencies in a Fourier transform. Admittedly, a Time Domain Pulse may be transformed into a multiplicity of frequencies... however, the phase relationship of two sets of Time domain transforms of I and Q with known muli-core DSP processing in real time operating systems is asking for a quantum leap in processing outside of a full blown laboratory at MIT. (IMHO)

Are we talking the use of XILINX FPGA cores? Interesting!

Please feel free to enlighten me.!

- Geowizard