Object verification using radar images

Patent metadata

Field	Value
Publication number	US-11391819-B2
Application number	US-201916272991-A
Country	US
Kind code	B2
Filing date	Feb 11, 2019
Priority date	Jul 18, 2018
Publication date	Jul 19, 2022
Grant date	Jul 19, 2022

Abstract

Techniques and systems are provided for performing object verification using radar images. For example, a first radar image and a second radar image are obtained, and features are extracted from the first radar image and the second radar image. A similarity is determined between an object represented by the first radar image and an object represented by the second radar image based on the features extracted from the first radar image and the features extracted from the second radar image. A determined similarity between these two sets of features is used to determine whether the object represented by the first radar image matches the object represented by the second radar image. Distances between the features in the two radar images can optionally also be compared and used to determine object similarity. The objects in the radar images may optionally be faces.

First claim

What is claimed is:

1. A method of performing object verification using radar images, the method comprising:
obtaining a first radar image and a second radar image;
extracting features from the first radar image, wherein the features extracted from the first radar image include a respective magnitude for each pixel in the first radar image, a magnitude for a pixel in the first radar image being calculated based on at least a root sum of squares of an amplitude and a phase of at least a first radio signal corresponding to the pixel in the first radar image;
extracting features from the second radar image, wherein the features extracted from the second radar image include a respective magnitude for each pixel in the second radar image, a magnitude for a pixel in the second radar image being calculated based on at least a root sum of squares of an amplitude and a phase of at least a second radio signal corresponding to the pixel in the second radar image;
determining distances between first respective magnitudes corresponding to the features extracted from the first radar image and second respective magnitudes corresponding to the features extracted from the second radar image;
determining a similarity between an object represented by the first radar image and an object represented by the second radar image based on the distances between the first respective magnitudes corresponding to the features extracted from the first radar image and the second respective magnitudes corresponding to the features extracted from the second radar image; and
determining whether the object represented by the first radar image matches the object represented by the second radar image based on the determined similarity.

2. The method of claim 1 , wherein the first radar image and the second radar image are generated using radio signals from an array of antennas, wherein the radio signals include the first radio signal and the second radio signal.

3. The method of claim 2 , wherein each pixel in the first radar image corresponds to at least one antenna from the array of antennas, and wherein each pixel in the second radar image corresponds to at least one antenna from the array of antennas.

4. The method of claim 1 , wherein the features extracted from the first radar image include at least the amplitude and the phase for the pixel in the first radar image, and wherein the features extracted from the second radar image include at least the amplitude and the phase for the pixel in the second radar image.

5. The method of claim 1 , wherein determining the distances between the first respective magnitudes corresponding to the features from the first radar image and the second respective magnitudes corresponding to the features from the second radar image includes:
determining a first respective distance between a respective amplitude for each pixel in the first radar image and a corresponding amplitude for a corresponding pixel in the second radar image; and
determining a second respective distance between a respective phase for each pixel in the first radar image and a corresponding phase for a corresponding pixel in the second radar image.

6. The method of claim 1 , wherein determining the distances between the first respective magnitudes corresponding to the features from the first radar image and the second respective magnitudes corresponding to the features from the second radar image further includes:
determining a respective distance between the respective magnitude for each pixel in the first radar image and a corresponding magnitude for a corresponding pixel in the second radar image.

7. The method of claim 1 , wherein at least the amplitude and the phase are extracted for each range bin of a plurality of range bins corresponding to each pixel in the first radar image, and wherein at least the amplitude and the phase are extracted for each range bin of a plurality of range bins corresponding to each pixel in the second radar image.

8. The method of claim 7 , wherein the respective magnitude for each pixel in the first radar image is extracted for each range bin of the plurality of range bins corresponding to each pixel in the first radar image, and wherein the respective magnitude for each pixel in the second radar image is extracted for each range bin of the plurality of range bins corresponding to each pixel in the second radar image.

9. The method of claim 1 , wherein the similarity between the object represented by the first radar image and the object represented by the second radar image is determined using a mapping function between matching labels and distances between the features from the first radar image and the features from the second radar image.

10. The method of claim 9 , wherein the mapping function is determined using a support vector machine (SVM).

11. The method of claim 9 , wherein the mapping function is determined using a support vector machine (SVM) and principal component analysis (PCA).

12. The method of claim 9 , wherein the mapping function is determined using a Partial Least Squares Regression (PLSR).

13. The method of claim 9 , wherein the mapping function is determined using a deep neural network.

14. The method of claim 1 , wherein the object represented by the first radar image is determined to match the object represented by the second radar image when the determined similarity is greater than a pre-determined matching threshold.

15. The method of claim 1 , wherein the object represented by the first radar image is determined not to match the object represented by the second radar image when the determined similarity is less than a pre-determined matching threshold.

16. The method of claim 1 , wherein the first radar image is an input image obtained from a radar measurement device, and wherein the second radar image is an enrolled image from an enrolled database.

17. The method of claim 1 , wherein the object represented by the first radar image is a first face, and wherein the object represented by the second radar image is a second face.

18. The method of claim 1 , wherein determining a similarity between the object represented by the first radar image and the object represented by the second radar image is based on an output of one or more trained machine learning models in response to input of at least a portion of the distances into the one or more trained machine learning models.

19. The method of claim 1 , wherein the root sum of squares of the amplitude and the phase of at least the first radio signal is an absolute value of the amplitude and the phase of at least the first radio signal, and wherein the root sum of squares of the amplitude and the phase of at least the second radio signal is an absolute value of the amplitude and the phase of at least the second radio signal.

20. An apparatus for performing object verification using radar images, comprising:
a memory configured to store one or more radar images; and
a processor configured to:
obtain a first radar image and a second radar image;
extract features from the first radar image, wherein the features extracted from the first radar image include a respective magnitude for each pixel in the first radar image, a magnitude for a pixel in the first radar image being calculated based on at least a root sum of squares of an amplitude and a phase of at least a first radio signal corresponding to the pixel in the first radar image;
extract features from the second radar image, wherein the features extracted from the second radar image include

Assignees

• Qualcomm Inc
• Qualcomm Incorporate

Inventors

• Sarkis Michel Adib
• Bi Ning
• Qi Yingyong
• Sanderovich Amichai
• Hemo Evyatar

Classifications

• G01S13/87 Primary

  Combinations of radar systems, e.g. primary radar and secondary radar · CPC title

• G01S13/90

  using synthetic aperture techniques {, e.g. synthetic aperture radar [SAR] techniques} · CPC title

• G01S13/89

  for mapping or imaging · CPC title

• G01S7/417 Primary

  involving the use of neural networks · CPC title

• G01S7/412

  based on a comparison between measured values and known or stored values · CPC title