Synthetic 3D X-Ray Bag Images for Threat Detection Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing aviation security protocols face challenges in training human scanners and computer-aided systems to identify threats in luggage due to the scarcity of 3D image data of bags with and without threats, necessitating costly and dangerous real-world scans.
Innovation Solution
A system generates synthetic 3D X-ray images by combining pre-captured 3D CT images of non-threat bags with separately scanned threat images, creating threat-carrying bag images for training and testing detection algorithms and human screeners.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real-world 3D scanning of bags with and without threats is performed to train scanners, then training data quality is improved, but time, cost, and safety risk increase
Solution Approach 1:
The patent creates synthetic 3D CT images by copying and combining existing 3D bag images with threat objects. The system extracts threat objects from training datasets and inserts them into virtual bag models, generating realistic synthetic images that replicate real-world scenarios without requiring actual scanning of threat-containing bags. This copying approach maintains training data quality while eliminating the time-consuming and dangerous real-world scanning process.
Solution Approach 2:
The patent segments the bag and threat objects into separate 3D components. By dividing the complex task of scanning entire threat-containing bags into separate operations (scanning bags, scanning threats, then combining), the system enables efficient data generation. The segmentation allows independent processing and combination of bag and threat 3D models, significantly reducing acquisition time while maintaining realism.
2Reliability
If real-world 3D scanning of threat-containing bags is performed, then training data authenticity is improved, but safety risk and cost increase
Solution Approach 1:
The system generates synthetic training data by copying and combining existing 3D bag models with threat object models. This approach maintains authenticity by using real 3D scanned components while eliminating the harmful act of physically scanning threat-containing bags. The synthetic images faithfully represent real-world scenarios without exposing personnel to safety risks.
Solution Approach 2:
The patent introduces a computational intermediary process that combines 3D bag models with threat object models to create synthetic images. This intermediary digital combination process replaces the direct physical scanning of threat-containing bags, maintaining data authenticity while removing safety hazards associated with handling and scanning actual threats.
3Measurement precision
If a large number of 3D images with threats are collected for training, then detection accuracy is improved, but data acquisition complexity and cost increase
Solution Approach 1:
The patent segments the data acquisition process into independent components: collecting 3D bag images, collecting threat object images, and then computationally combining them. This segmentation transforms a complex acquisition task into simpler, more manageable steps that can be performed independently and scaled efficiently, reducing overall complexity while enabling large dataset generation.
Solution Approach 2:
The system uses copying to generate synthetic training data by combining existing 3D bag models with threat object models. This approach eliminates the need for complex real-world scanning operations and allows unlimited generation of training images by simply combining pre-existing 3D models, significantly reducing acquisition complexity while maintaining detection accuracy.
Data Source
AI summary
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.