Multimodal Image Fusion for Noisy Terahertz Security Scans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Terahertz/millimeter wave imaging technology faces limitations such as limited penetration capability, large image noise, low image definition, and inability to accurately identify materials, leading to false alarms and unidentifiable objects during security inspections.
Innovation Solution
A multi-modal image processing apparatus and method that combines terahertz/millimeter wave, visible light, and infrared imaging technologies for enhanced image registration, noise filtering, and fusion, utilizing image enhancement, background noise removal, and deep learning algorithms to improve image quality and material identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Temperature
If terahertz/millimeter wave imaging technology is used for security inspection, then penetration capability is improved, but image noise increases and image definition deteriorates
Solution Approach 1:
The patent combines terahertz/millimeter wave images with visible light images and infrared images into a fused image. The visible light and infrared images provide high-definition structural information that compensates for the poor image definition in terahertz/millimeter wave images, while the terahertz/millimeter wave component maintains penetration capability through clothing.
2Reliability
If terahertz/millimeter wave imaging technology is used, then material detection capability is improved, but image noise increases
Solution Approach 1:
The patent extracts and removes background noise from the terahertz/millimeter wave image before fusion. By separating the useful material detection information from the noisy background, the system maintains material detection capability while eliminating the harmful noise that would otherwise interfere with image quality.
3Productivity
If single-band terahertz/millimeter wave imaging is used, then imaging speed is improved, but identification accuracy deteriorates
Solution Approach 1:
The patent performs preliminary image preprocessing operations including noise removal, image enhancement, and feature extraction on the terahertz/millimeter wave images before fusion. This preparation work ensures that when multiple images are combined, the identification accuracy is maximized without requiring repeated imaging, thus maintaining imaging speed.
4Measurement precision
If multi-modal image fusion is implemented, then identification accuracy is improved, but device complexity increases
Solution Approach 1:
The patent divides the image processing system into distinct functional modules: terahertz/millimeter wave imaging module, visible light imaging module, infrared imaging module, image preprocessing module, and image fusion module. Each module performs a specific function, making the complex system more manageable and maintainable while achieving high identification accuracy through coordinated operation of all modules.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In the disclosure, an image processing apparatus (104) and method thereof are provided. The image processing apparatus comprises an image obtaining module configured to obtain a terahertz/millimeter wave image in a terahertz/millimeter wave band and at least two images which are imaged in at least two light/radiation bands different from terahertz/millimeter wave band and have the same scene as the terahertz/millimeter wave image, an image registration module (107) configured to perform image registration between the terahertz/millimeter wave image and the at least two images by using the terahertz/millimeter wave image as a reference image, an image fusion module (109) configured to successively perform multi-level fusion on the registered terahertz/millimeter wave image and each of the registered at least two images using at least one fusion method to obtain a multi-level fused image, and an image identification module configured to identify an item carried on an object in the multi-level fused image using an image identification method.