System and method for transferring electro-optical (EO) knowledge for synthetic-aperture-radar (SAR)-based object detection

1. A system for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the system comprising:

one or more processors and a non-transitory computer-readable medium having executable instructions encoded thereon such that when executed, the one or more processors perform an operation of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain,

wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity,

wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

2. The system as set forth in claim 1 , wherein the one or more processors further perform operations of:

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain.

3. The system as set forth in claim 2 , wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution.

4. A computer implemented method for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the method comprising an act of:

causing one or more processers to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain,

wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity,

wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

5. The method as set forth in claim 4 , wherein the one or more processors further perform operations of:

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain.

6. The method as set forth in claim 5 , wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution.

7. A computer program product for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the computer program product comprising:

computer-readable instructions stored on a non-transitory computer-readable medium that are executable by a computer having one or more processors for causing the processor to perform operations of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain,

wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity,

wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

8. The computer program product as set forth in claim 7 , further comprising instructions for causing the one or more processors to further perform operations of:

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain.

9. The computer program product as set forth in claim 8 , wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution.

10. A system for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the system comprising:

one or more processors and a non-transitory computer-readable medium having executable instructions encoded thereon such that when executed, the one or more processors perform an operation of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain;

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain,

wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

11. The system as set forth in claim 10 , wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity.

12. The system as set forth in claim 11 , wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders.

13. A computer implemented method for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the method comprising an act of:

causing one or more processers to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain;

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain,

wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

14. The method as set forth in claim 13 , wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity.

15. The method as set forth in claim 14 , wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders.

16. A computer program product for transferring learned knowledge from an electro-optical (EO) domain to a synthetic-aperture-radar (SAR) domain, the computer program product comprising:

computer-readable instructions stored on a non-transitory computer-readable medium that are executable by a computer having one or more processors for causing the processor to perform operations of:

using a measured similarity between the EO domain and the SAR domain, training a model for classifying SAR images using knowledge previously learned from the electro-optical (EO) domain;

training the model with a set of labeled data points in the EO domain; and

adapting the trained model using a set of labeled data points in the SAR domain, where a number of data points in the set of labeled data points in the SAR domain is significantly fewer than a number of data points in the set of labeled data points in the EO domain,

wherein the set of labeled data points in the SAR domain and a set of unlabeled data points in the SAR domain are used to align an EO probability distribution and a SAR domain probability distribution;

using the trained model, processing a SAR image to determine regions of interest in the SAR image;

classifying a region of interest to determine whether the region of interest corresponds to an object of interest;

outputting classified regions of interest that contain the object of interest;

displaying the object of interest on a visualization map; and

automatically updating the visualization map to reflect a change in position of the object of interest.

17. The computer program product as set forth in claim 16 , wherein Sliced-Wasserstein Distance (SWD) is used to determine the measured similarity.

18. The computer program product as set forth in claim 17 , wherein SWD is used to minimize a discrepancy between an EO domain probability distribution and a SAR domain probability distribution in a latent space of a neural auto-encoder in order to supervise training of domain-specific encoders.