Techniques for generating synthetic three-dimensional representations of threats disposed within a volume of a bag
In an approach to generating a synthetic three-dimensional (3D) X-ray volume, a first bag image of the plurality of bag images that includes an associated bag subvolume greater than a volume of a threat represented in a first threat image of the plurality of threat images is selected. An image is created based on inserting the threat of the first threat image into the associated bag subvolume of the first bag image.
1. A system for generating a synthetic three-dimensional (3D) image, the system comprising:
a memory having a plurality of bag images and a plurality of threat images;
one or more controllers;
one or more non-transitory computer readable storage media; and
program instructions stored on the one or more non-transitory computer readable storage media for execution by at least one of the one or more controllers, the stored program instructions including instructions to:
determine a threat image of the plurality of threat images having a smallest threat volume;
determine at least one bag image of the plurality of bag images having a bag subvolume greater than the smallest threat volume;
store in the memory a first identifier of the at least one bag image having the bag subvolume greater than the smallest threat volume;
store an entry in a lookup table that maps the first identifier with a second identifier of at least one threat image of the plurality of threat images with a first threat volume less than the bag subvolume of the at least one bag image having the bag subvolume greater than the smallest threat volume;
select a second bag image and a second threat image, the second bag image having the bag subvolume greater than a second threat volume of the second threat image based on the lookup table; and
responsive to determining that the bag subvolume of the second bag image is greater than the second threat volume of the second threat image, create a new image based on inserting the second threat image into the first second bag image.
2. The system of claim 1 , wherein the one or more controllers are configured to determine the at least one bag image of the plurality of bag images having the bag subvolume greater than the smallest threat volume based on connected component labeling (CCL).
3. The system of claim 1 , wherein the one or more controllers are further configured to store in the memory an identifier of the at least one bag image of the plurality of bag images having the bag subvolume greater than the smallest threat volume.
4. The system of claim 3 , wherein the one or more controllers are further configured to store in the memory an associated identifier of at least one bag subvolume for the at least one bag image of the plurality of bag images having the bag subvolume greater than the smallest threat volume.
5. The system of claim 1 , wherein the one or more controllers are further configured to identify whether the second threat image contains a metal based on an associated threshold density or atomic number, and wherein the one or more controllers are further configured to generate an artifact image with integrated threat and metal artifacts based on the metal.
6. The system of claim 5 , wherein the one or more controllers are further configured to output the artifact image with the integrated threat and the metal artifacts to a machine learning system.
7. The system of claim 1 , wherein the one or more controllers are configured to determine a plurality of orientations for the second threat image within the second bag image.
8. The system of claim 7 , wherein the one or more controllers are configured to randomly select an orientation of the plurality of orientations to generate the image.
9. The system of claim 1 , wherein each of the plurality of threat images include a representation of a threat object, the threat object being an explosive device or a weapon.
10. The system of claim 1 , wherein the plurality of bag images and the plurality of threat images comprise 3D image data.
11. The system of claim 1 , wherein the plurality of threat images and the plurality of bag images comprise 3D image data captured by a 3D CT scanning device.
12. A computer-implemented method for generating a synthetic three-dimensional (3D) image comprising:
determining, by one or more controllers, a first 3D X-ray volume of a first object having a smallest object volume;
determining, by the one or more controllers, at least one second 3D X-ray volume of a second object having a subvolume greater than the smallest object volume;
storing, by the one or more controllers, an entry in a lookup table that maps a first identifier of the at least one second 3D X-ray volume of the second object with a second identifier of at least one object of a plurality of objects with a 3D X-ray volume less than the second 3D X-ray volume of the second object;
selecting, by the one or more controllers, a third object and a fourth object, the third object having a subvolume greater than a volume of the fourth object based on the lookup table; and
creating, by the one or more controllers, a synthetic 3D X-ray image by inserting a third 3D X-ray volume of the third object into a fourth 3D X-ray volume of the fourth object.
13. The computer-implemented method of claim 12 , wherein:
the fourth 3D X-ray volume of the fourth object is a bag image and the third 3D X-ray volume of the third object is a threat image, and
a bag subvolume of the bag image is greater than a smallest threat volume.
14. The computer-implemented method of claim 13 , further comprising:
determining, by the one or more controllers, a plurality of orientations for the threat image within the bag image.
15. The computer-implemented method of claim 13 , wherein the lookup table associates the bag image with any threat image that is insertable into the bag image.
16. The computer-implemented method of claim 12 , wherein creating the synthetic 3D X-ray image by inserting the third 3D X-ray volume of the third object into fourth 3D X-ray volume of the fourth object further comprises:
determining, by the one or more controllers, whether the third 3D X-ray volume of the third object contains a metal based on at least one of an associated average density and an atomic number; and
creating, by the one or more controllers, the synthetic 3D X-ray image with artifacts based on the metal.