3D Threat Model Rendering for X-Ray Training Image Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Obtaining a large quantity of radiographic images of illicit materials and/or threat items in various inspection environments and scenarios for training machine learning vision models is difficult and expensive, and existing methods for generating synthetic images are inaccurate and costly.
Innovation Solution
Generating two-dimensional radiographic-like images from three-dimensional virtual models of illicit materials and threat items using computer graphics, modifying orientations, and rendering these images with simulation software to simulate x-ray interactions, adding noise, and resizing to fit machine learning tools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real radiographic images of illicit materials and threat items are obtained for training machine learning models, then the training data accuracy is improved, but the cost and difficulty of obtaining such images increases significantly
Solution Approach 1:
The patent creates synthetic copies of radiographic images by rendering 3D models of illicit materials and threat items. Instead of obtaining actual radiographic images of real contraband (which is difficult and expensive), the system generates photorealistic synthetic images that replicate the visual characteristics and radiographic properties of real images, providing accurate training data without the logistical challenges of acquiring genuine samples
Solution Approach 2:
The system varies multiple parameters in the synthetic image generation process, including material composition, density, orientation, positioning within containers, and radiographic imaging parameters. By changing these parameters across multiple generated images, the system creates diverse training datasets that accurately represent the variability found in real inspection scenarios
2Adaptability or versatility
If diverse orientations and configurations of threat items are imaged for comprehensive model training, then the model robustness is improved, but the number of images required and processing complexity increases
Solution Approach 1:
The system dynamically generates images with varying orientations, positions, and configurations of threat items within containers. Rather than manually setting up each scenario, the software automatically varies parameters such as rotation angles, placement positions, and container configurations to produce diverse training images, enabling comprehensive model training without manual intervention for each variant
Solution Approach 2:
The system performs preliminary actions by pre-defining 3D models, material properties, and container geometries before image generation. These pre-configured digital twins and material databases allow rapid generation of diverse scenarios without repeated setup, reducing processing complexity while maintaining model robustness through comprehensive orientation and configuration coverage
3Ease of manufacture
If synthetic images are generated to reduce cost and improve accessibility of training data, then the ease of manufacture is improved, but the accuracy and realism of the images may deteriorate
Solution Approach 1:
The patent replaces the mechanical and logistical system of physically obtaining radiographic images of real illicit materials with a computational rendering system. Instead of using actual X-ray machines and physical contraband samples, the system uses computer graphics rendering with physics-based models to simulate radiographic imaging, achieving both ease of image generation and high realism through accurate physical simulations
Solution Approach 2:
The system uses composite approaches by combining multiple technical elements: 3D modeling, material science databases, radiographic physics simulations, and computer graphics rendering. This composite methodology integrates various disciplines to produce synthetic images that maintain high realism and accuracy while being easily generated through software, overcoming the trade-off between ease of manufacture and image quality
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Provides an accurate and inexpensive method to generate synthetic images of illicit materials and threat items in multiple orientations and configurations, enhancing the training of AI-based vision models for security inspection systems.
Implementation Method 1
scanning the virtual 3D model by using a simulation software application
Data Source
AI summary
A method for generating two dimensional (2D) radiographic like images of a threat item from a 3D model of the item includes: constructing a 3D model of an illicit material and/or threat item by using a computer graphics (CG) process; making the 3D model at least partially transparent to light; changing orientation of the 3D model one or more times; and rendering the 3D model to generate a plurality of 2D radiographic like images, wherein a number of 2D images rendered is dependent upon the number of times the orientation of the 3D model is changed.