Granted Patent B2

US 12,737,587 · App. 18/240,554 · Granted Sep 15, 2026

Inventor: Matthew Miller (Edgewater, CO)

G06N3/044G01V3/10G01V3/38G06N3/09

Abstract

Claims (68)

1 . A system comprising:

a pulsed induction (PI) metal detector configured to output a voltage signal responsive to being proximate to an object, wherein the PI metal detector comprises a power supply, a coil, and a waveform generating component; and

a processing device configured to:

receive the voltage signal outputted from the PI metal detector, wherein the voltage signal comprises a square wave including an on period and an off period;

determine, based on the voltage signal and position information associated with the PI metal detector, a plurality of features;

provide the plurality of features as input to a trained machine learning model (MLM);

receive output from the trained MLM, wherein the output comprises a confidence value associated with a class of a metal, and wherein the confidence value comprises a percentage;

determine whether the confidence value in the output satisfies a threshold criterion; and

responsive to determining that the confidence value satisfies the threshold criterion, cause output of a notification indicating a detection of the class of the metal in the object,

wherein determining the plurality of features further comprises:

selecting data points in the voltage signal satisfying a criterion;

identifying a first feature of the plurality of features by using a linear regression on the voltage signal;

identifying a second feature of the plurality of features by comparing predefined data points associated with the voltage signal;

identifying a third feature of the plurality of features by using a Fast Fourier transform (FFT) on the voltage signal;

using a proximity to the object of the position information as a fourth feature of the plurality of features;

using signal intensity information of the voltage signal as a fifth feature of the plurality of features; and

using a heat map representing detection confidence of the voltage signal as a sixth feature of the plurality of features.

2 . The system of claim 1 , wherein the voltage signal comprises a voltage measured over a time period.

3 . The system of claim 1 , wherein the trained MLM comprises a binary classification model.

4 . The system of claim 1 , wherein the trained MLM comprises a multi-class classification model.

5 . The system of claim 1 , wherein the trained MLM comprises a recurrent neural network with at least one hidden layer.

6 . The system of claim 1 , wherein the trained MLM is trained using data input comprising historical features of a type common with a type of the plurality of features associated with the voltage signal.

7 . The system of claim 1 , wherein the processing device is further configured to:

preprocess the voltage signal to detect a second plurality of features; and

select the plurality of features from the second plurality of features.

8 . The system of claim 1 , wherein the voltage signal is associated with a voltage measured across the coil.

9 . The system of claim 1 , wherein the output of the notification comprises providing an alert to a user device or making a physical marking proximate to the object.

10 . A method comprising:

receiving a voltage signal outputted from a PI metal detector responsive to the PI metal detector being proximate to an object, wherein the PI metal detector comprises a power supply, a coil, and a waveform generating component, wherein the voltage signal comprises a square wave including an on period and an off period;

determining, based on the voltage signal and position information associated with the PI metal detector, a plurality of features;

providing the plurality of features as input to a trained machine learning model (MLM);

receiving output from the trained MLM, wherein the output comprises a confidence value associated with a class of a metal, and wherein the confidence value comprises a percentage;

determining whether the confidence value in the output satisfies a threshold criterion; and

responsive to determining that the confidence value satisfies the threshold criterion, causing output of a notification indicating a detection of the class of the metal in the object,

wherein determining the plurality of features further comprises:

selecting data points in the voltage signal satisfying a criterion;

identifying a first feature of the plurality of features by using a linear regression on the voltage signal;

identifying a second feature of the plurality of features by comparing predefined data points associated with the voltage signal;

identifying a third feature of the plurality of features by using a Fast Fourier transform (FFT) on the voltage signal;

using a proximity to the object of the position information as a fourth feature of the plurality of features;

using signal intensity information of the voltage signal as a fifth feature of the plurality of features; and

using a heat map representing detection confidence of the voltage signal as a sixth feature of the plurality of features.

11 . The method of claim 10 , wherein the voltage signal comprises a voltage measured over a time period.

12 . The method of claim 10 , wherein the trained MLM comprises a recurrent neural network with at least one hidden layer.

13 . A non-transitory computer-readable medium storing instructions thereon, wherein the instructions, when executed by a processing device, cause the processing device to:

receive a voltage signal outputted from a pulsed induction (PI) metal detector responsive to the PI metal detector being proximate to an object, wherein the PI metal detector comprises a power supply, a coil, and a waveform generating component, wherein the voltage signal comprises a square wave including an on period and an off period;

determine, based on the voltage signal and position information associated with the PI metal detector, a plurality of features;

provide the plurality of features as input to a trained machine learning model (MLM);

receive output from the trained MLM, wherein the output comprises a confidence value associated with a class of a metal, and wherein the confidence value comprises a percentage;

determine whether the confidence value in the output satisfies a threshold criterion; and

responsive to determining that the confidence value satisfies the threshold criterion, cause output of a notification indicating a detection of the class of the metal in the object,

wherein determining the plurality of features further comprises:

selecting data points in the voltage signal satisfying a criterion;

identifying a first feature of the plurality of features by using a linear regression on the voltage signal;

identifying a second feature of the plurality of features by comparing predefined data points associated with the voltage signal;

identifying a third feature of the plurality of features by using a Fast Fourier transform (FFT) on the voltage signal;

using a proximity to the object of the position information as a fourth feature of the plurality of features;

using signal intensity information of the voltage signal as a fifth feature of the plurality of features; and

using a heat map representing detection confidence of the voltage signal as a sixth feature of the plurality of features.

14 . The non-transitory computer-readable medium of claim 13 , wherein the voltage signal comprises a voltage measured over a time period.

15 . The non-transitory computer-readable medium of claim 13 , wherein the trained MLM comprises a binary classification model.

16 . The non-transitory computer-readable medium of claim 13 , wherein the trained MLM comprises a multi-class classification model.

17 . The non-transitory computer-readable medium of claim 13 , wherein the trained MLM comprises a recurrent neural network with at least one hidden layer.

18 . The non-transitory computer-readable medium of claim 13 , wherein the trained MLM is trained using data input comprising historical features of a type common with a type of the plurality of features associated with the voltage signal.

19 . The non-transitory computer-readable medium of claim 13 , wherein the processing device is further configured to:

preprocess the voltage signal to detect a second plurality of features; and

select the plurality of features from the second plurality of features.

20 . The non-transitory computer-readable medium of claim 13 , wherein the output of the notification comprises providing an alert to a user device or making a physical marking proximate to the object.

Continuity (2)

References Cited (59)