Autonomous detection of concealed weapons at a distance by sensor fusion

Patent metadata

Field	Value
Publication number	US-12078770-B2
Application number	US-202217750838-A
Country	US
Kind code	B2
Filing date	May 23, 2022
Priority date	May 23, 2022
Publication date	Sep 3, 2024
Grant date	Sep 3, 2024

Abstract

An apparatus for distant detection of ferrous metallic objects includes a plurality of magnetic field sensors and a fingerprint processing system. The plurality of magnetic field sensors each measure an ambient magnetic field. The fingerprint processing system receives, from the plurality of magnetic field sensors, magnetic field data corresponding to the ambient magnetic field. The fingerprint processing system also generates, from the magnetic field data, a magnetic fingerprint image of one or more perturbations of the ambient magnetic field. The plurality of magnetic field sensors is separated from the ferrous metallic objects being detected by a detection distance that is greater than or equal to 0.5 meters.

First claim

What is claimed is:

1. An apparatus for distant detection of ferrous metallic objects, the apparatus comprising:
a plurality of magnetic field sensors that each measure an ambient magnetic field; and
a fingerprint processing system, wherein the fingerprint processing system:
receives, from the plurality of magnetic field sensors, magnetic field data corresponding to the ambient magnetic field,
generates, from the magnetic field data, a magnetic fingerprint image of one or more perturbations of the ambient magnetic field, and
classifies, using a pretrained magnetic fingerprint image classifier, the magnetic fingerprint image into one of a set of classes that each correspond to a different type of ferrous metallic object, wherein

the plurality of magnetic field sensors is separated from the ferrous metallic objects being detected by a detection distance that is greater than or equal to 0.5 meters, and
the magnetic fingerprint image classifier is trained using labeled training data that includes a set of standard labeled fingerprint images selected from a group consisting of:
a set of artificially-generated magnetic fingerprint images based on simulations of magnetic field perturbations that correspond to ferrous metallic objects,
a set of empirically-generated magnetic fingerprint images, of known ferrous metallic objects, that are physically measured using the plurality of magnetic field sensors, and
combinations of the set of artificially-generated magnetic fingerprint images and the set of empirically-generated magnetic fingerprint images.

2. The apparatus according to claim 1 , wherein the labeled training data further comprises:
variational labeled fingerprint images that are based on the set of standard labeled fingerprint images and are generated using a generative adversarial network.

3. The apparatus according to claim 1 , wherein
the plurality of magnetic field sensors is organized as a magnetic sensor array comprising rows and columns of magnetic field sensors, and
each of the rows is orthogonal with each of the columns.

4. The apparatus according to claim 3 , wherein
the magnetic sensor array comprises at least four magnetic field sensors, disposed in a square array, and
a separation distance, between adjacent magnetic field sensors, is less than or equal to 20 centimeters.

5. The apparatus according to claim 1 , wherein each of the plurality of the magnetic field sensors is an anisotropic magnetoresistance sensor that has a sensitivity sufficient to measure perturbations of 10 nanoTesla in the ambient magnetic field.

6. A method for distant detection of ferrous metallic objects, the method comprising:
receiving, from a plurality of magnetic field sensors, magnetic field data corresponding to an ambient magnetic field;
generating, from the magnetic field data, a magnetic fingerprint image of one or more perturbations of the ambient magnetic field;
classifying, using a pretrained magnetic fingerprint image classifier, the magnetic fingerprint image into one of a set of classes that each correspond to a different type of ferrous metallic object; and
training the magnetic fingerprint image classifier using labeled training data that includes a set of standard labeled fingerprint images selected from a group consisting of:
a set of artificially-generated magnetic fingerprint images based on simulations of magnetic field perturbations that correspond to ferrous metallic objects,
a set of empirically-generated magnetic fingerprint images, of known ferrous metallic objects, that are physically measured using the plurality of magnetic field sensors, and
combinations of the set of artificially-generated magnetic fingerprint images and the set of empirically-generated magnetic fingerprint images.

7. The method according to claim 6 , wherein
the plurality of magnetic field sensors is separated from the ferrous metallic objects being detected by a detection distance that is greater than or equal to 0.5 meters.

8. The method according to claim 6 , wherein
the plurality of magnetic field sensors is organized as a magnetic sensor array comprising rows and columns of magnetic field sensors, and
each of the rows is orthogonal with each of the columns.

9. The method according to claim 8 , wherein
the magnetic sensor array comprises at least four magnetic field sensors, disposed in a square array, and
a separation distance, between adjacent magnetic field sensors, is less than or equal to 20 centimeters.

10. The method according to claim 6 , wherein each of the plurality of the magnetic field sensors is an anisotropic magnetoresistance sensor that has a sensitivity sufficient to measure perturbations of 10 nanoTesla in the ambient magnetic field.

11. The method according to claim 6 , wherein the labeled training data further comprises:
variational labeled fingerprint images that are based on the set of standard labeled fingerprint images and are generated using a generative adversarial network.

12. A non-transitory computer readable medium (CRM) storing computer readable program code for distant detection of ferrous metallic objects, the computer readable program code causes a computer to:
receive, from a plurality of magnetic field sensors, magnetic field data corresponding to an ambient magnetic field;
generate, from the magnetic field data, a magnetic fingerprint image of one or more perturbations of the ambient magnetic field;
classify, using a pretrained magnetic fingerprint image classifier, the magnetic fingerprint image into one of a set of classes that each correspond to a different type of ferrous metallic object; and
train the magnetic fingerprint image classifier using labeled training data that includes a set of standard labeled fingerprint images selected from a group consisting of:
a set of artificially-generated magnetic fingerprint images based on simulations of magnetic field perturbations that correspond to ferrous metallic objects,
a set of empirically-generated magnetic fingerprint images, of known ferrous metallic objects, that are physically measured using the plurality of magnetic field sensors, and
combinations of the set of artificially-generated magnetic fingerprint images and the set of empirically-generated magnetic fingerprint images.

13. The non-transitory CRM according to claim 12 , wherein
the plurality of magnetic field sensors is separated from the ferrous metallic objects being detected by a detection distance that is greater than or equal to 0.5 meters.

14. The non-transitory CRM according to claim 12 , wherein
the plurality of magnetic field sensors is organized as a magnetic sensor array that comprises at least four magnetic field sensors, disposed in a square array, and
a separation distance, between each magnetic field sensor, is less than or equal to 20 centimeters.

15. The non-transitory CRM according to claim 12 , wherein each of the plurality of the magnetic field sensors is an anisotropic magnetoresistance sensor that has a sensitivity sufficient to measure perturbations of 10 nanoTesla in the ambient magnetic field.

16. The non-transitory CRM according to claim 12 , wherein the labeled training data further comprises:
variational labeled fingerprint images that are based on the set of standard labeled fingerprint images and are generated using a generative adversarial network.

Assignees

• Konica Minolta Business Solutions Usa Inc

Inventors

• Amano Jun

Classifications

• G06V10/774

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G01R33/096

  anisotropic magnetoresistance sensors · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V2201/05

  Recognition of patterns representing particular kinds of hidden objects, e.g. weapons, explosives, drugs · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title