Jump to content

Anyone Using Artificial Intelligence With Their Gold Detectors?


Recommended Posts


At the risk of demonstrating just how little I know and understand any of this discussion l believe that it would be possible to build a detector capable of seeing and interpreting accurately signatures from an unknown object and identify through this a gold nugget or any other object, but I suspect not by doing the same thing as we do now gathering  data in the same way we always have.

A larger database of more of the same information is constrained by its own limitations, much like determining an objects color from looking at a variety of black and white photographs at different angels and different resolution it can tell you only so much the color always a guess interpolating an answer from incomplete information. The current state of the art technology is well exploited, refined and developed advances are baby steps bringing improvement fine tuning information the same way it’s always done even GZP is a signal timing differences not that different from PI I suspect just processed slightly different.

We will need to add something new to the game, we have one finger now pointing at the ground, VLF, PI and GPZ  our black and white images from these technologies all see objects but due to the variety in objects  their relative location in the ground  the state they are in even within the same element cause so much overlap the best we can do is guess. New led old led, bits of led snow flake after snow flake all looking similar all different. The technology works amazingly at what it does just separating an object out from a huge volume of surrounding noise but something more we’ll need like looking with multi frequency only the frequency will need to gather clues from a different perspective that adds significantly different information?

OK, I have no idea of what I’m talking about and I’ll leave this to you smart guys to figure out, but I’m sure glad someone invented metal detectors.

  • Like 2
  • Thanks 1
Link to comment
Share on other sites

27 minutes ago, 1515Art said:

At the risk of demonstrating just how little I know and understand any of this discussion l believe that it would be possible to build a detector capable of seeing and interpreting accurately signatures from an unknown object and identify through this a gold nugget or any other object, but I suspect not by doing the same thing as we do now gathering  data in the same way we always have.

A larger database of more of the same information is constrained by its own limitations, much like determining an objects color from looking at a variety of black and white photographs at different angels and different resolution it can tell you only so much the color always a guess interpolating an answer from incomplete information. The current state of the art technology is well exploited, refined and developed advances are baby steps bringing improvement fine tuning information the same way it’s always done even GZP is a signal timing differences not that different from PI I suspect just processed slightly different.

We will need to add something new to the game, we have one finger now pointing at the ground, VLF, PI and GPZ  our black and white images from these technologies all see objects but due to the variety in objects  their relative location in the ground  the state they are in even within the same element cause so much overlap the best we can do is guess. New led old led, bits of led snow flake after snow flake all looking similar all different. The technology works amazingly at what it does just separating an object out from a huge volume of surrounding noise but something more we’ll need like looking with multi frequency only the frequency will need to gather clues from a different perspective that adds significantly different information?

OK, I have no idea of what I’m talking about and I’ll leave this to you smart guys to figure out, but I’m sure glad someone invented metal detectors.

Like I said previously, we are not going to see a step change in capability without moving away from Induction-based technology.  It's pretty much tapped out and all we are doing is polishing a cannonball while adding bells and whistles to it.

  • Like 4
  • Thanks 1
Link to comment
Share on other sites

I'd be curious to see how AI could be applied to geology and mapping of ancient gold bearing rivers? From what we have now to what geologists have previously discovered and published their findings. Possibly could be a useful tool in future gold patch hunting. 

  • Like 2
Link to comment
Share on other sites

As for accurately mapping where a detector coil has been, here’s a paper describing using a PDA and a camera to precisely map out the coil position and it creates a low resolution outline image of the target as the coil scans it. The primary use for the system was intended for UXO detecting, but it may be useful for relict detecting as well.  They even use the system for object recognition, but it would be useful for small targets though.  As for hardware, I’d think any new smartphone has the capability to do this, as the researchers who made the system were using a basic PDA and camera to do it.

I posted about it here previously:


https://www.ndt.net/article/ecndt2006/doc/Tu.4.5.4.pdf


More Here:

 

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

43 minutes ago, Chase Goldman said:

Like I said previously, we are not going to see a step change in capability without moving away from Induction-based technology.  It's pretty much tapped out and all we are doing is polishing a cannonball while adding bells and whistles to it.

Sorry if I’m restating a view previously expressed, not my intention… I picked back up on a thought from last night that needed more time in the oven before it was ready to slice and serve, only skimming new entries.

  • Like 1
Link to comment
Share on other sites

I think as soon as phones become as powerful as many already believe they are the tools to do real time local 3D comparative scanning already exists.

https://www.opentopography.org/blog/iphone-lidar-applications-geosciences

Now all we have to do is figure out how to strap a $1000 iPhone to our coils along with a cooling tower for the coprocessor and a case for the extra batteries and data storage.

 

 

  • Like 1
  • Haha 1
Link to comment
Share on other sites

56 minutes ago, Chase Goldman said:

Like I said previously, we are not going to see a step change in capability without moving away from Induction-based technology.  It's pretty much tapped out and all we are doing is polishing a cannonball while adding bells and whistles to it.

Bingo!

Link to comment
Share on other sites

11 hours ago, Ben201000 said:

Hi BrokeInBendigo,

Thank you for both such in depth posts. I'll do my best to try and summarize what I understand from them and then add/ask anything more. 

First it isn't very useful (potentially impossible) to detect between metals. It is far more useful to balance the ground and reduce EMI. Is this because EMI produces a lot of noise and leads to digging holes without any metal in? Or is it mostly distracting and mentally tiring over the hours? 

Secondly the ground conditions can vary enormously, essentially making a single model that knows everything impossible. Do the ground conditions vary hugely within one detecting session or is it more in different geographical areas? 

Something I hadn't considered was the scale of the area. I.e would a very experienced detector walk through a 5 kilometre square area and be able to narrow down a few 100 meter square areas that are worth detecting? If so, could AI be useful attached to a drone that scans a 2km area with various sensors attached, and then tries to narrow a few 100m areas to look in. As I write this I realise getting training data for this would be extremely hard. 

I really like the ideas around having a screen showing the raw waveforms. 

It seems from everyone's response that conditions can be so variable that the AI would have to be trained on each separate detection session. This presents a few challenges but could be possible. YouTube/Spotify etc all train customised AI models for each person. Right now it's possible to take about 30 photos of a bird vs not-bird on a hike, and train an AI in about two minutes on a standard laptop, with about 90% accuracy (not a hugely relevant example, but this was unthinkable 7 years ago). Of course no one wants to carry a laptop around, but there are AI specific training chips which are tiny. One on the iPhone as mentioned, but also more specialised ones like the Google Coral and Nvidia Jetson Nano. There is also the possibility of sending the waveforms over to servers to quickly train, but I imagine there isn't any 4G where everyone is detecting. 

Latest thought: There is a small AI training chip (no larger than a smartphone), with a screen (probably use smartphone app for it). The raw waves from the coil somehow connect into the chip. There are 3 buttons on the screen (label 1, label 2, train AI). When someone reaches the specific site they begin by choosing the labels. Maybe metal vs not-metal, but it could be anything depending what the detector thinks could work. As you dig holes you click which label it turned out to be. After say 30 holes you click 'train' and wait about 2 minutes. It then starts giving you a real-time prediction in a percentage of which label it thinks the coil is detecting. 
In no way does this replace the detector, you'd still use your usual method, there would just be something extra to test against. The approach of this method would be to have little AIs that are personalised to each persons local knowledge, and these AIs are each persons property unless they decide to share them. They may find that it works well distinguishing between two local features and they use it for that. Or that training after 10 samples is handy but they only trust the results about 40% of the time. Or that 500 samples is required, but that's okay because they regularly detect on similar ground. Or they upload and combine data with others in similar ground conditions. 

My last thought was that to collect data wouldn't it be possible to bury a few known items and scan over them in the specific soil? Or on a larger scale bury lots of samples in an area and scan over at different heights, add more soil above and scan again etc? 

To clarify how 30 samples could possibly work. We could use transfer learning on an AI trained on millions of images (called ImageNet), and then it is taught the new samples. Some people have had a lot of success turning waveforms into images and then training them with this AI. 

Thank you so much for the Dall E prompt booklet! I've got access to it so will definitely have a read through. 

Also are you in Bendigo, Victoria, Australia? I'm in Melbourne 🙂 

 

You're most welcome. I've a keen interest in this, so please excuse if I go on a bit long about this stuff.

Balancing and EMI

It's most important to balance the ground mineralisation because a target on the edge of detection cannot be differentiated by the detector operator from the response from the ground.

To elaborate, I'll just describe a bit of how pulse induction detectors work. The coil is energised with varied intensity and at varied intervals. The pattern constitutes the "timing" involved. The detector transmits a pulse or pattern of pulses and then is set to a receive mode to get a response. TX/RX cycles are very fast.

The timing you select is largely dependent on the ground type. Some timings are higher-energy than others. The more energy you put into the coil, the better the response on gold/metal, so you want the highest-energy timing that will work in the area.  

Problem is, different ground responds to different timings in different ways. For example, the ground may be saturated to a significant degree with iron-rich minerals (generally in an area where quartz reefs occur) or salt (dry salt lakes), or have far less mineralisation (granitic sands). Some ground types will produce responses very similar to a piece of gold with a certain timing. The higher-energy timings may be unusable in these areas because the ground looks more or less the same as a metal object at depth. In other words, it sounds like there are targets everywhere (on current detectors). 

You then switch to a lower-energy timing, or perhaps a timing with a different pattern of pulses, which does not produce as much of a response for that particular ground type. In doing so, you lose detection depth. 

Regarding EMI, it is a similar situation. When near sources of EMI (e.g. power lines) at frequencies and amplitudes such that they energise the coil, again you get a response that is very similar to a real target. 

In both cases, there are a variety of settings on the detector, some of which process the data stream before it is converted into audio and some afterwards. You could use a higher-energy timing but reduce the gain at a particular point a the processing chain, or apply whatever kind of DSP method. A common one is "stabilisation", where you ignore low-amplitude variations in the signal. Reducing the amount of data being converted to audio reduces audible responses of course. Sometimes it's better to have a quiet detector with less sensitivity, other times a noisier detector with more sensitivity. The operator and their experience is a major factor in setting choice.

So what we need to improve in current detectors is the handling of the ground response, when using the highest-energy timings. We need to separate the signal from noise, which I believe is a task very well-suited to an AI.

Ground Conditions

Ground conditions can be and very often is variable in very small areas. For example, you may have an iron rich area of bedrock that is visible at the surface and then slopes downward, at which point it is covered in a mostly organic soil. Swinging the detector across this boundary of soil to ironstone will induce a response. Another example is a quantity of soil which has higher capacity to retain water than an adjacent area (again, this could be within a square metre) - this will cause a response. Of course, across larger regions you have trends as well. 

The timing used can reduce/eliminate the response at these boundaries, at the cost of detection depth.

Training/implementation

Your idea is fine, but I'll reiterate that you absolutely need an existing algorithm processing the ground for this to work. Nobody wants to dig X holes just to get started in a particular spot, it's not feasible. You'll understand when you recover 30 targets 🙂

Burying targets

Not gonna work. Metal objects in the ground have been there from a few decades to literally millions of years. The presence of the objects in addition to all that time create profoundly different conditions than what you'd get if you buried something. A common technique to somewhat obviate this is to drill holes from below the desired target depth and insert a test target. This allows the ground above the target to be undisturbed and provides a more realistic testing scenario. 

Still, it won't really work because of the extremely diverse range of ground conditions in which gold is found. The only real use of testing on buried targets is comparing different detectors relative detection depths, and even then it's a very smooshy, vague kind of testing. 

Waveforms to Images

Sounds like a reasonable method. I had a lot of fun coding an app to interpret an image as raw audio, then apply audio effects to it, and finally re-interpret it back to an image. I mean, a waveform is just a 1 by X image, right? 

Yes - In Bendigo at the moment. I think the majority of this site is focused on the US but plenty of Australian fossickers and prospectors here.

 

11 hours ago, Chase Goldman said:

No, that is really easily sorted and works well right now.  Tracking algorithms work well, also.  I am sure AI could be applied but really you have a solution looking for a problem when it comes to ground tracking, to be frank.

...

 

Hard disagree. The increase in targets recovered if we could use the highest-energy timings in all areas, with perfect ground and EMI handling (meaning silent threshold), would be IMMENSE. 

 

11 hours ago, Ben201000 said:

Hi Redz,

Excuse me for the extremely dumb questions, but are hot rocks rocks that read similar to metal? Are there certain areas where there are lots which makes it harder to look for metals? 

 

Hot rocks are metal-rich rocks (sometimes also mineralised with gold!) which sound like a nugget. Some areas are covered in hot rocks and your options are to either tune down your detector to ignore them (and necessarily miss some gold) or dig them all (and spend a lot more time digging).

 

7 hours ago, jasong said:

Inertial navigation using accelerometers and then dead reckoning is far more accurate than GPS for moving systems. It's sufficient for things like general ground sensing. When used in combination with mag sensors and coil data, you could achieve sub-inch resolution easily. 

Essentially you would make the ground it's own reference frame while swinging the coil using inertial navigation, and then tie that high resolution work "chunk" to a traditional lat/lon reference frame via GPS. 

 

I've suggested this a few times, and there are other applications - an accelerometer could be used to know the detector is in a "target recovery" mode and it should not do any ground tracking. It could also be used with an AI to recognise small-scale geological boundaries and not give one-way responses when crossing them. Would I want this? No, I'd rather hear the ground changes. But still interesting.

 

6 hours ago, Dutchman4 said:

...

Do you really want an AI based, all knowing, detector to reduce you down to the mere unintelligent functions of swinging and digging? 

 

I absolutely want my detector to handle as much extraneous information processing as possible. I want to use my prospecting knowledge to determine where to swing and turn my detector sensitivity as high as it can go.

 

5 hours ago, Sourdough Scott said:

I'd be curious to see how AI could be applied to geology and mapping of ancient gold bearing rivers? From what we have now to what geologists have previously discovered and published their findings. Possibly could be a useful tool in future gold patch hunting. 

 

You may be interested in looking up geophysical interpretations. I don't know if AI is being used in this area but such datasets are extremely interesting.

  • Like 3
  • Thanks 1
Link to comment
Share on other sites

1 hour ago, BrokeInBendigo said:

Hard disagree. The increase in targets recovered if we could use the highest-energy timings in all areas, with perfect ground and EMI handling (meaning silent threshold), would be IMMENSE. 

Immense?  Um, ok. :rolleyes:

Not sure what you mean by silent threshold, but OK.

Induction-based detecting tech has plateaued.  But I can't fault folks for wanting to squeeze more blood out of a rock.  More power to you and your unbridled optimism and enthusiasm.  Would be more than happy to be proven wrong.

Meanwhile I'll focus most of my energy into site identification and using tech and research to find sites likely to produce and methods to thoroughly exploit them for coverage.  Much greater return on investment than micro incremental improvements in an applied principle that has been fundamentally unchanged for 100 plus years despite the latest advancements that have delivered incremental improvements through SMF and signal processing.

  • Like 5
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...