System and method of new understandable training for AI as applied to electromagnetic sensor detection
AssigneeThe Government of the United States of America, as represented by the Secretary of Homeland Security
InventorMark A. Fry
Life of this patent
Estimating the term…
Abstract
Systems and methods are described, and an example method includes a training an artificial intelligence (AI) classifier of electromagnetic (EM) scanned items, including obtaining a training set of sample raw EM scans. The set includes a population of sample in-class raw EM scans, which include blocks of EM sensor data from EM scans of regions having in-class objects, and the set includes a population of sample not-in-class raw EM scans, which include blocks of EM sensor data from EM scan of regions without in-class objects. The example includes applying the AI classifier to sample raw EM scans in the training set, measuring errors in the results, and updating classifier parameters based on the errors, until detecting a training completion state.
Claims20
A method for training an artificial intelligence (AI) classifier of electromagnetic (EM) scanned items, comprising:
- obtaining a training set of sample raw EM scans, comprising a population of sample in-class raw EM scans and a population of sample not-in-class raw EM scans, the sample in-class raw EM scans including a corresponding block of EM sensor data generated in an EM scan of a spatial region that includes one or more in-class objects, and the sample not-in-class raw EM scan including a corresponding block of EM sensor data generated in an EM scan of a spatial region that includes no in-class objects;
- initializing AI classifier parameters of the AI classifier, and
- applying the AI classifier to a plurality of sample raw EM scans in the training set, measuring errors in results of applying the AI classifier, and updating at least one of the AI classifier parameters based on the errors.
A method for artificial intelligence (AI) classification of electromagnetic (EM) scanned items, comprising:
- EM scanning a subject, including transmitting an EM scan energy toward the subject, receiving an EM return, and, based on the EM return, generating an EM return reception signal, and generating a raw EM sample block, based on a sampling of the EM return reception signal;
- AI classifying the raw EM sample block between likely including and likely not including samples of EM return from an in-class object;
- and
- generating an alarm, based at least in part on a positive result of the AI classifying.
A tangible, non-transitory computer-readable storage medium, embodying computer-readable instructions that cause a hardware processor to:
- perform an artificial intelligence (AI) classification of electromagnetic (EM) scanned items, including:
- receiving samples of an EM scanning a subject, including samples of an EM return reception signal, which corresponds to transmitting an EM scan energy toward the subject, receiving an EM return, and, based on the EM return, generating the EM return reception signal,
- generating a raw EM sample block, based on the samples of the EM return reception signal, and
- perform an AI classifying of the raw EM sample block between likely including and likely not including samples of EM return from an in-class object;
- and
- generating an alarm, based at least in part on a positive result of the AI classifying.
Dependencies are inferred from the claim text.