PI Target Identification

[Tinkerer]

Designing a better PI detector
Every technology has its limitations, however, daily there are new and improved components available that make it possible to overcome some of the limitations. But it also needs inventive minds and much effort to come up with real improvements.
I believe that the joined resources of the Geotech forum has the capability to come up with many ideas and improvements.
Lets look at Target Identification:
The decay curve decays at a rate that is dependent on the TC or Time Constant of the target.
If we take several samples along the decay curve, we can define the TC of the target by a mathematical formula, calculated by the MCU.
We could then compare the result with a “Look up Table” (ROM Array) that shows the probable identity of the target.
Any ideas of how to implement this?????
Tinkerer

[Mrand]

I recall that somebody was trying already the statistical method. The scatter of parameters of real objects was so big as he gave this method up. The description was probably 2-3 year back in the Eric Foster forum. I think the Adonis is utilizing the similar method. Look this site:
http://www.titanium-system.com/en/index.htm

Mrand

[Max]

Designing a better PI detector
Every technology has its limitations, however, daily there are new and improved components available that make it possible to overcome some of the limitations. But it also needs inventive minds and much effort to come up with real improvements.
I believe that the joined resources of the Geotech forum has the capability to come up with many ideas and improvements.
Lets look at Target Identification:
The decay curve decays at a rate that is dependent on the TC or Time Constant of the target.
If we take several samples along the decay curve, we can define the TC of the target by a mathematical formula, calculated by the MCU.
We could then compare the result with a “Look up Table” (ROM Array) that shows the probable identity of the target.
Any ideas of how to implement this?????
Tinkerer

Hi,
it's not so difficault to implement having right hardware... fast MCU and ADC stuff.. but problem is that's totally unreliable.

It's not like phase shift with which you could REALLY discriminate targets in a relative large range in the boundary of bare detection range of detector.

Kind regards,
Max

[Aziz]

Designing a better PI detector
Every technology has its limitations, however, daily there are new and improved components available that make it possible to overcome some of the limitations. But it also needs inventive minds and much effort to come up with real improvements.
I believe that the joined resources of the Geotech forum has the capability to come up with many ideas and improvements.
Lets look at Target Identification:
The decay curve decays at a rate that is dependent on the TC or Time Constant of the target.
If we take several samples along the decay curve, we can define the TC of the target by a mathematical formula, calculated by the MCU.
We could then compare the result with a “Look up Table” (ROM Array) that shows the probable identity of the target.
Any ideas of how to implement this?????
Tinkerer

Hi Tinkerer,

this seams a true challange for me. Developing a highly computerized PI technology using sophisticated numerical methods.

Unfortunately, I have not all instruments what I need for this yet. I want to investigate in identifying targets (true PI discrimination) later. And if this is possible, then it could be implemented quite easily.

You can use neural networks for analyzing the data sets for all possible targets. If they build a distinguishable cluster, then they can be savely classifiyed. They also help to find the proof of an unknown problem. The unknown problem ("black-box") may not allways be linear nature. Neural networks need no look-up tables but numerical weights for neuron links. They can be used for finding new knowledge, analyzing the quality of input data and also analyzing black-boxes.

Or:

If you want to make this very easy, then you could use the auto-correlation method. This is quite easy to implement and it is fast and efficient.

Aziz

[Tinkerer]

Discriminating PI

I recall that somebody was trying already the statistical method. The scatter of parameters of real objects was so big as he gave this method up. The description was probably 2-3 year back in the Eric Foster forum. I think the Adonis is utilizing the similar method. Look this site:
http://www.titanium-system.com/en/index.htm

Mrand

I agree with you.
Just a simple statistic is not good enough.
Now, what if we recognize the decay curve in its various components and analyze each component separately?
The decay curve is composed of several components:
1.Remnants of the TX field decay and ground response
2.Skin effect eddy current decay
3.Core eddy currents
4.Magnetic reactance or hysteresis
Removing #1 will help recognizing the others
Skin effect eddy currents are of very short duration and are related to the surface area of the target. Foil is a good example.
Core eddy current TC’s are more related to the mass of the target. Example, a sphere has maximum mass and minimum surface.
Core eddy current TC’s are also related to the conductivity of the target.
Differentiating magnetic response will help to eliminate lots of FE trash as well as bricks and many hot rocks.
Tinkerer

[Tinkerer]

Target ID for PI

Hi Tinkerer,

this seams a true challange for me. Developing a highly computerized PI technology using sophisticated numerical methods.

Unfortunately, I have not all instruments what I need for this yet. I want to investigate in identifying targets (true PI discrimination) later. And if this is possible, then it could be implemented quite easily.

You can use neural networks for analyzing the data sets for all possible targets. If they build a distinguishable cluster, then they can be savely classifiyed. They also help to find the proof of an unknown problem. The unknown problem ("black-box") may not allways be linear nature. Neural networks need no look-up tables but numerical weights for neuron links. They can be used for finding new knowledge, analyzing the quality of input data and also analyzing black-boxes.

Or:

If you want to make this very easy, then you could use the auto-correlation method. This is quite easy to implement and it is fast and efficient.

Aziz

Could you tell me a bit more about the "auto-correlation"?

The neural network part is just a bit too much for the few neurons I have left in my brain.

About the PI target ID, it is definitely possible even I can do it. But I still have a lot of work to be done on it until it is ready.
I welcome all the help I can get.
Tinkerer

[Aziz]

Could you tell me a bit more about the "auto-correlation"?

The neural network part is just a bit too much for the few neurons I have left in my brain.

About the PI target ID, it is definitely possible even I can do it. But I still have a lot of work to be done on it until it is ready.
I welcome all the help I can get.
Tinkerer

Hi Tinkerer,

see wiki:
http://en.wikipedia.org/wiki/Autocorrelation
http://en.wikipedia.org/wiki/Cross-correlation
http://en.wikipedia.org/wiki/Correlation

This background will help to find best matches of signal decay curve to the stored signal curves for identifiying the target.

Aziz

[Tinkerer]

PI discrimination

Hi Tinkerer,

see wiki:
http://en.wikipedia.org/wiki/Autocorrelation
http://en.wikipedia.org/wiki/Cross-correlation
http://en.wikipedia.org/wiki/Correlation

This background will help to find best matches of signal decay curve to the stored signal curves for identifiying the target.

Aziz

Thanks for the links. I am looking forward to when you will have the time to write this software.

In the meantime I would be happy to just find a good way to make a user interface that lets the operator see or feel or hear the variations that I can clearly see on the scope.

Tinkerer

[1843]

I think a ferrous nonferrous discrimination is possible on a PI. There are several commerical PIs in the world capable of it.
Could you describe the decay curves of ferrous metals vs. nonferrous?

Thanks,
1843

[Tinkerer]

Target differentiation

I think a ferrous nonferrous discrimination is possible on a PI. There are several commerical PIs in the world capable of it.
Could you describe the decay curves of ferrous metals vs. nonferrous?

Thanks,
1843

I will look for some scope pictures and post them. Give my a few days, I am a bit busy now.

Tinkerer

[Tinkerer]

Target discrimination

I think a ferrous nonferrous discrimination is possible on a PI. There are several commerical PIs in the world capable of it.
Could you describe the decay curves of ferrous metals vs. nonferrous?

Thanks,
1843

OK, here are the scope pictures. I will post them separately so the legend does not get mixed up.
Picture #1, NO TARGET

[Tinkerer]

Target identification

Picture # 2, Silver Dollar

[Tinkerer]

Target identification

Picture # 3
Steel nail.
You can see how the signal goes in opposite polarity than the silver dollar.

[Tinkerer]

Target identificaation

Picture # 4.
1/2" lead minie ball. This represents a target with lesser conductivity and minimal surface area versus mass.

[Unregistered]

Targets

Hi Tinkerer,

What does a piece of ferrite do to the curve?

What happens if you run a nail vertical across the coil and then horizontal?

Great work!
Mark