Concealed substance detection with hyperspectral imaging
A drug screening system and methods of using the same are disclosed. The drug screening system identifies and highlights drugs or other contraband substances enclosed within mailpieces based on score imagery computed from selected wavelengths based on chemical signatures. A wide-field, multispectral short-wave infrared imaging system allows for dissolved, concealed drug screening of mailpieces that improve inspection efficiency and accuracy.
1. A method of distinguishing a target from a background in a sample, the method comprising:
generating a score image from a hyperspectral image, the hyperspectral image having been formed from photons that have interacted with a sample that includes the target and the background;
preprocessing the score image, the preprocessing comprising:
removing isolated outlier pixels from the score image,
performing substrate flat fielding of the score image,
performing hole detection and filling for the score image,
preforming image orientation of the score image, and
cropping the score image;
performing anomaly detection on the score image to remove confusants and identify one or more anomalies in the score image, the confusants including one or more of text, graphics, or images; and
classifying one or more anomalies as target or non-target.
2. The method of claim 1 , wherein the target is one or more of body tissues, body fluids, explosives, drugs, toxins, fruit, cheese, meat, alcohol, flavor compounds, Chemical Warfare Agents (CWA), Toxic Industrial Chemicals (TIC), or Non-Traditional Agents (NTA).
3. The method of claim 2 , wherein the target is a drug and the non-target is a substance that is not a drug.
4. The method of claim 1 , wherein the text, graphics, or images are removed by one or more of Niblack's approach, Sauvola's approach, or Bataineh's approach.
5. The method of claim 1 , wherein classifying is performed with a trained neural network.
6. The method of claim 1 , wherein classifying is performed based on a You Only Look Once (YOLO) detector.
7. A system for distinguishing a target from a background in a sample, the system comprising a processor and a non-transitory computer readable medium that when executed, causes the processor to:
generate a score image from a hyperspectral image, the hyperspectral image having been formed from photons that have interacted with a sample that includes the target and the background;
preprocess the score image by removing isolated outlier pixels from the score image, performing substrate flat fielding of the score image, performing hole detection and filling for the score image, performing image orientation of the score image, and cropping the score image;
perform anomaly detection on the score image to remove confusants and identify one or more anomalies in the score image, the confusants including one or more of text, graphics, or images; and
classify one or more anomalies as target or non-target.
8. The system of claim 7 , wherein the target is one or more of body tissues, body fluids, explosives, drugs, toxins, fruit, cheese, meat, alcohol, flavor compounds, Chemical Warfare Agents (CWA), Toxic Industrial Chemicals (TIC), or Non-Traditional Agents (NTA).
9. The system of claim 8 , wherein the target is a drug and the non-target is a substance that is not a drug.
10. The system of claim 7 , wherein the text, graphics, or images are removed by one or more of Niblack's approach, Sauvola's approach, or Bataineh's approach.
11. The system of claim 7 , wherein the processor classifies the one or more anomalies with a trained neural network.
12. The system of claim 7 , wherein the processor classifies the one or more anomalies based on a You Only Look Once (YOLO) detector.
13. A method of distinguishing a target from a background in a sample, the method comprising:
forming a hyperspectral image from photons that have interacted with a sample that includes the target and the background;
generating a score image from the hyperspectral image;
preprocessing the score image, the preprocessing comprising:
removing isolated outlier pixels from the score image,
performing substrate flat fielding of the score image,
performing hole detection and filling for the score image,
performing image orientation of the score image, and
cropping the score image;
performing anomaly detection on the score image to remove confusants and identify one or more anomalies in the score image, the confusants including one or more of text, graphics, or images; and
classifying one or more anomalies as target or non-target.
14. The method of claim 13 , wherein the target is one or more of body tissues, body fluids, explosives, drugs, toxins, fruit, cheese, meat, alcohol, flavor compounds, Chemical Warfare Agents (CWA), Toxic Industrial Chemicals (TIC), or Non-Traditional Agents (NTA).
15. The method of claim 14 , wherein the target is a drug and the non-target is a substance that is not a drug.
16. The method of claim 13 , wherein the text, graphics, or images are removed by one or more of Niblack's approach, Sauvola's approach, or Bataineh's approach.
17. The method of claim 13 , wherein classifying is performed with a trained neural network.
18. The method of claim 13 , wherein classifying is performed based on a You Only Look Once (YOLO) detector.
19. A system for distinguishing a target from a background in a sample, the system comprising a processor and a non-transitory computer readable medium that when executed, causes the processor to:
form a hyperspectral image from photons that have interacted with a sample that includes the target and the background;
generate a score image from the hyperspectral image;
preprocess the score image by removing isolated outlier pixels from the score image, performing substrate flat fielding of the score image, performing hole detection and filling for the score image, performing image orientation of the score image, and cropping the score image,
perform anomaly detection on the score image to remove confusants and identify one or more anomalies in the score image, the confusants including one or more of text, graphics, or images; and
classify one or more anomalies as target or non-target.
20. The system of claim 19 , wherein the target is one or more of body tissues, body fluids, explosives, drugs, toxins, fruit, cheese, meat, alcohol, flavor compounds, Chemical Warfare Agents (CWA), Toxic Industrial Chemicals (TIC), or Non-Traditional Agents (NTA).
21. The system of claim 20 , wherein the target is a drug and the non-target is a substance that is not a drug.
22. The system of claim 19 , wherein the text, graphics, or images are removed by one or more of Niblack's approach, Sauvola's approach, or Bataineh's approach.
23. The system of claim 19 , wherein the processor classifies the one or more anomalies with a trained neural network.
24. The system of claim 19 , wherein the processor classifies the one or more anomalies based on a You Only Look Once (YOLO) detector.