US-12625233-B1 - Measurement of electromagnetic spectral properties

Abstract

Disclosed are methods, systems, and computer-readable medium to perform operations to classifying airfield debris. The operations include generating a series of frequency-modulated chirps having respective frequencies across a frequency range and directed toward a location of an object in an airfield. The operations can include receiving a series of reflections corresponding to a plurality of frequency-modulated chirps in the series of frequency-modulated chirps and generating a two-dimensional frequency-swept spectral signature representing properties of the plurality of received reflections across a frequency range. The operations can include providing the two-dimensional frequency-swept spectral signature as input to a machine learning classifier that is trained to predict characteristics of objects from frequency-swept spectral signatures and receiving an output representing a predicted classification of the object as airfield debris or not airfield debris.

Inventors

• Jon Williams

Assignees

• General Radar Corporation

Dates

Publication Date

20260512

Application Date

20221208

Claims (20)

1. 1 . A method for classifying airfield debris comprising: generating, using a radar transmit subsystem directed toward a location of a same object in an airfield, a series of frequency-modulated chirps across a same frequency range, each frequency-modulated chirp from the series of frequency-modulated chirps having a different respective portion of frequencies from the same frequency range than another respective portion of frequencies for another frequency-modulated chirp in the series of frequency-modulated chirps; receiving, using a radar receive subsystem, a series of reflections reflected from the same object and that correspond to a plurality of frequency-modulated chirps in the series of frequency-modulated chirps; generating, from the series of reflections across the same frequency range, a two-dimensional frequency-swept spectral signature representing frequency-dependent features of the same object derived from the series of reflections across the same frequency range, wherein the two-dimensional frequency-swept spectral signature is an image comprising a first dimension representing the same frequency range and a second dimension representing powers of the series of reflections measured at the respective frequencies across the same frequency range; providing the two-dimensional frequency-swept spectral signature of the same object as input to a machine learning classifier that is trained to predict characteristics of objects from frequency-swept spectral signatures; and receiving, from the machine learning classifier, an output representing a predicted classification of the same object as airfield debris or not airfield debris.
2. 2 . The method of claim 1 , wherein the series of frequency-modulated chirps having the respective portions of frequencies across the same frequency range are modulated linearly with respect to time.
3. 3 . The method of claim 1 , wherein generating the two-dimensional frequency-swept spectral signature comprises: determining a product of the series of frequency-modulated chirps and the series of reflections; and analyzing, using an analog to digital converter, the product at the respective portions of frequencies across the same frequency range.
4. 4 . The method of claim 3 , wherein the analog to digital converter uses a fast Fourier transform.
5. 5 . The method of claim 3 , wherein the product of the series of frequency-modulated chirps and the series of reflections comprises a single tone beat frequency resulting from the same object.
6. 6 . The method of claim 1 , wherein receiving the output representing the predicted classification of the same object comprises: scaling at least one reference signal; and matching, using least squares, the two-dimensional frequency-swept spectral signature to the at least one reference signal.
7. 7 . The method of claim 1 , wherein the output representing a predicted classification of the same object comprises a certainty that the same object is airfield debris.
8. 8 . The method of claim 1 , wherein the output representing a predicted classification of the same object comprises a severity of the same object for a given state of the airfield.
9. 9 . The method of claim 8 , wherein the severity of the same object is based on at least one of (i) size or (ii) material, of the same object.
10. 10 . The method of claim 1 , wherein generating the two-dimensional frequency-swept spectral signature comprises: determining, for a particular chirp in the plurality of frequency-modulated chirps reflected from the same object, that a power of a beat frequency generated by the same object for a duration of the particular chirp is proportional to a power of frequency content in a transmitted chirp from the series of frequency modulated chirps.
11. 11 . A radar system comprising one or more processors and one or more storage devices storing instructions that when executed by the one or more processors cause the one or more processors to perform operations comprising: generating a series of frequency-modulated chirps i) having respective frequencies across a frequency range across a same frequency range and ii) directed toward a location of a same object in an airfield, each frequency-modulated chirp from the series of frequency-modulated chirps having a different respective portion of frequencies from the same frequency range than another respective portion of frequencies for another frequency-modulated chirp in the series of frequency-modulated chirps; receiving a series of reflections reflected from the same object and that correspond to a plurality of frequency-modulated chirps in the series of frequency-modulated chirps; generating, from the series of reflections across the same frequency range, a two-dimensional frequency-swept spectral signature representing frequency-dependent features of the same object derived from the series of reflections across the same frequency range, wherein the two-dimensional frequency-swept spectral signature is an image comprising a first dimension representing the same frequency range and a second dimension representing powers of the series of reflections measured at the respective frequencies across the same frequency range; and providing the two-dimensional frequency-swept spectral signature of the same object as input to a machine learning classifier that is trained to predict characteristics of objects from frequency-swept spectral signatures; and receiving, from the machine learning classifier, an output representing a predicted classification of the same object as airfield debris or not airfield debris.
12. 12 . The radar system of claim 11 , wherein the series of frequency-modulated chirps having the respective portions of frequencies across the same frequency range are modulated linearly with respect to time.
13. 13 . The radar system of claim 11 , wherein generating the two-dimensional frequency-swept spectral signature comprises: determining a product of the series of frequency-modulated chirps and the series of reflections; and analyzing, using an analog to digital converter, the product at the respective portions of frequencies across the frequency same range.
14. 14 . The radar system of claim 13 , wherein the analog to digital converter uses a fast Fourier transform.
15. 15 . The radar system of claim 11 , wherein receiving the output representing the predicted classification of the same object comprises: scaling at least one reference signal; and matching, using least squares, the two-dimensional frequency-swept spectral signature to the at least one reference signal.
16. 16 . The radar system of claim 11 , wherein the output representing a predicted classification of the same object comprises a certainty that the same object is airfield debris.
17. 17 . The radar system of claim 11 , wherein the output representing a predicted classification of the same object comprises a severity of the same object for a given state of the airfield.
18. 18 . The radar system of claim 17 , wherein the severity of the same object is based on at least one of (i) size or (ii) material, of the same object.
19. 19 . A non-transitory computer storage medium encoded with instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: generating a series of frequency-modulated chirps i) having respective frequencies across a same frequency range and ii) directed toward a location of a same object in an airfield, each frequency-modulated chirp from the series of frequency-modulated chirps having a different respective portion of frequencies from the same frequency range than another respective portion of frequencies for another frequency-modulated chirp in the series of frequency-modulated chirps; receiving a series of reflections reflected from the same object and that correspond to a plurality of frequency-modulated chirps in the series of frequency-modulated chirps; generating, from the series of reflections across the same frequency range, a two-dimensional frequency-swept spectral signature representing frequency-dependent features of the same object derived from the series of reflections across the same frequency range, wherein the two-dimensional frequency-swept spectral signature is an image comprising a first dimension representing the same frequency range and a second dimension representing powers of the series of reflections measured at the respective frequencies across the same frequency range; and providing the two-dimensional frequency-swept spectral signature of the same object as input to a machine learning classifier that is trained to predict characteristics of objects from frequency-swept spectral signatures; and receiving, from the machine learning classifier, an output representing a predicted classification of the same object as airfield debris or not airfield debris.
20. 20 . The non-transitory computer storage medium of claim 19 , wherein the series of frequency-modulated chirps having the respective portions of frequencies across the same frequency range are modulated linearly with respect to time.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims benefit of U.S. Provisional Application No. 63/289,366, filed Dec. 14, 2021, the contents of which is hereby incorporated by reference in its entirety. BACKGROUND Radio detection and ranging (radar) and other range-finding systems determine the presence, distance and/or velocity of objects (e.g., aircraft). Radars determine these properties by transmitting electromagnetic waves and receiving a reflection off of the objects. SUMMARY Radar can be used to detect and classify foreign object debris (FOD) on airfields. Objects can be classified by controlling frequencies of waves (e.g., through sweeps or chirps) transmitted from a frequency modulated continuous wave (FMCW) radar system. The radar system can receive reflections corresponding to the transmitted radio frequency (RF) waves (e.g., microwaves), and determine a spectral signature of the objects. The spectral signature can be used to determine properties of the objects. In some examples, an empty runway (e.g., free from vehicles, people, aircraft) can be scanned with RF waves, and the radar system can classify the object (e.g., as debris or not debris). In general, one innovative aspect of the subject matter described in this specification can be embodied in methods that include the actions of generating, using a radar transmit subsystem directed toward a location of an object in an airfield, a series of frequency-modulated chirps having respective frequencies across a frequency range; receiving, using a radar receive subsystem, a series of reflections corresponding to a plurality of frequency-modulated chirps in the series of frequency-modulated chirps; generating, from the plurality of received reflections across a frequency range, a two-dimensional frequency-swept spectral signature representing properties of the plurality of received reflections across the frequency range; providing the two-dimensional frequency-swept spectral signature as input to a machine learning classifier that is trained to predict characteristics of objects from frequency-swept spectral signatures; and receiving, from the machine learning classifier, an output representing a predicted classification of the object as airfield debris or not airfield debris. The previously-described implementation can be performed using a computer-implemented method; a non-transitory, computer-readable medium storing computer-readable instructions to perform the computer-implemented method; a processor including circuitry to execute one or more instructions that, when executed, cause the processor to perform the computer-implemented method; and a computer system including a computer memory interoperably coupled with a hardware processor configured to perform the computer-implemented method or the instructions stored on the non-transitory, computer-readable medium. These and other embodiments may each optionally include one or more of the following features. In some implementations, the series of frequency-modulated chirps having the respective frequencies across the frequency range are modulated linearly with respect to time. In some implementations, the two-dimensional frequency-swept spectral signature includes powers of the series of reflections measured at the respective frequencies across the frequency range. In some implementations, generating the two-dimensional frequency-swept spectral signature includes determining a product of the series of frequency-modulated chirps and the series of reflections; and analyzing, using an analog to digital converter, the product at the respective frequencies across the frequency range. In some implementations, the analog to digital converter uses a fast Fourier transform. In some implementations, the receiving the output representing the predicted classification of the object includes scaling at least one reference signal; and matching, using least squares, the two-dimensional frequency-swept spectral signature to the at least one reference signal. In some implementations, the output representing a predicted classification of the object includes a certainty that the object is airfield debris. In some implementations, the output representing a predicted classification of the object includes a severity of the object for a given state of the airfield. In some implementations, the severity of the object is based on at least one of the size or material of the object. The subject matter described in this specification can be implemented to realize one or more of the following advantages. The disclosed method enables improvement to the detection and classification of small changes (e.g., angularly small relative to the system) to an environment by using RF reflection. For example, small objects can be identified in complicated and/or noisy environments. In some implementations, properties of objects can be more precisely classified by using specially trained machine learning models. In some impleme