Land mine detector

• US 7,173,560 B2
• Filed: 08/30/2004
• Issued: 02/06/2007
• Est. Priority Date: 08/28/2003
• Status: Expired due to Fees

• Current Assignee: University of Florida Research Foundation, Incorporated (State University System of Florida)
• Original Assignee: University of Florida Research Foundation, Incorporated (State University System of Florida)
• Inventors: Li, Jian, Sun, Yijun
• Primary Examiner(s): Tarcza; Thomas H.
• Assistant Examiner(s): Barker; Matthew M.
• Application Number:
  US10/929,269
• Publication Number:
  US 20050128125A1
• Time in Patent Office: 890 Days
• Field of Search: 342/22
• US Class Current: 342/22
• CPC Class Codes: F41H 11/12 Means for clearing land minefields; Systems specially adapted for detection of landmines
• CPC Class Codes: G01V 3/12 operating with electromagnetic waves operating with millimetre waves G01V8/005
• CPC Class Codes: G01V 3/15 specially adapted for use during transport, e.g. by a person, vehicle or boat

Abstract

A forwarding looking ground penetrating mine detection apparatus includes a radiation source for irradiating a sample of ground suspected of containing at least one mine with a plurality of frequency swept ground penetrating radar signals. A detector receives target signals backscattered from the ground responsive to the radar signal. The detector includes a time-frequency analyzer which transforms the target signals into a time-frequency image representation (TFR). In a preferred embodiment, the detector can include a wavelet packet transformer (WPT) for extracting time-frequency localized information from the TFR in the form of feature set constructed from a wavelet table. The apparatus can also include a data dimensionality reducer for selecting features to form a feature subset from the feature set, preferably based on reference to a training data set. A multilayer neural network classifier can be based on the feature subset, and be adaptable to the surrounding environment through learning.

Claims

1. A forwarding looking ground penetrating mine detection apparatus, comprising:

a movable vehicle;

a radiation source comprising dual transmitting antennas mounted on said movable vehicle for irradiating a sample of ground suspected of containing at least one mine with a plurality of frequency swept ground penetrating radar signals, each of said radar signals having a discrete frequency, the discrete frequencies being evenly spaced over a predetermined frequency range; and

a detector comprising a plurality of evenly spaced-apart receiving antennas mounted on said movable vehicle for receiving target signals backscattered from said ground responsive to said radar signal, said detector including a time-frequency analyzer, said time-frequency analyzer transforming said target signals into a time-frequency image representation (TFR);

wherein said transmitting antennas operate sequentially in irradiating the sample, and wherein said receiving antennas operate simultaneously in receiving the target signals.

2. The apparatus of claim 1, wherein said detector further comprises a wavelet packet transformer (WPT), said WPT extracting time-frequency localized information from said TFR.

3. The apparatus of claim 2, wherein said time-frequency localized information comprises a feature set constructed from a wavelet table provided by said WPT, said feature set having a first data dimensionality, further comprising a data dimensionality reducer, said dimensionality reducer selecting features to form a feature subset from said feature set, said feature subset having a reduced dimensionality as compared to said first dimensionality.

4. The apparatus of claim 3, wherein said dimensionality reducer implements a sequential forward selector (SFS) with a cost function to compile said feature subset.

5. The apparatus of claim 4, wherein said cost function comprises a linear discriminant analysis (LDA) cost function and said SFS comprises sequential floating forward selector.

6. The apparatus of claim 5, wherein said data dimensionality reducer selects said feature subset based on reference to a training data set.

7. The apparatus of claim 4, further comprising a multilayer neural network classifier based on said feature subset.

8. The apparatus of claim 7, wherein said neural network is a trained network.

9. The apparatus of claim 8, wherein said training comprises minimizing a cross entropy function.

10. The apparatus of claim 7, wherein said neural network implements a boosting based adaptive feature selector.

11. The apparatus of claim 1, wherein said time-frequency analyzer implements a Choi-Williams Distribution (CWD).

21. The apparatus of claim 1, wherein said radar signals operate over 1024 discrete frequencies and wherein the discrete frequencies are evenly spaced at 2.5 MHz intervals over a frequency range from 442.5 MHz to 3 GHz.

22. The apparatus of claim 21, wherein the apparatus irradiates a sample and receives backscattered signals at 2 meter intervals as the moveable vehicle advances.

23. The apparatus of claim 22, further comprising a global positioning system to measure the location of the apparatus at each 2 meter interval.

12. A mine detection method, comprising the steps of:

irradiating a sample of ground suspected of containing at least one mine with a plurality of frequency swept ground penetrating radar signals using dual transmitting antennas mounted on a movable vehicle, said transmitting antennas being operated sequentially to irradiate the sample with frequency swept ground penetrating radar signals, each of said radar signals having one of a plurality of discrete frequencies evenly spaced over a predetermined frequency range;

receiving at a plurality of evenly spaced-apart receiving antennas mounted on said movable vehicle a signal backscattered from a ground penetrating swept radar radiated ground sample;

generating a time-frequency representation based on said backscattered signal, and

analyzing said time-frequency representation to indicate whether a mine is in said ground.

13. The method of claim 12, further comprising the step of wavelet packet transforming to extract time-frequency localized information from said TFR, said time-frequency localized information comprising a feature set constructed from a wavelet table provided by said wavelet packet transforming.

14. The method of claim 12, further comprising the step of data dimensionality reducing said feature set to provide a reduced dimensionality feature subset.

15. The method of claim 14, wherein said dimensionality reducing comprises a sequential forward selector (SFS) with a cost function.

16. The method claim 15, wherein said cost function comprises a linear discriminant analysis (LDA) cost function and said SFS comprises sequential floating forward selector.

17. The method of claim 16, further comprising extracting time-frequency localized information from said TFR based on reference to a training data set.

18. The method of claim 14, further comprising the step of configuring a multilayer neural network classifier based on said feature subset.

19. The method of claim 18, further comprising the step of training said neural network to a surrounding environment adjacent to said ground.

20. The method of claim 19, wherein said neural network implements a boosting based adaptive feature selector.