Related Experiment Video
Updated: Jun 9, 2025
07:13
Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
Published on: October 27, 2023
1.1K
Near-field millimeter-wave and visible image fusion via transfer learning
1School of Automation Engineering, University of Electronic Science and Technology of China, Xiyuan street 2006, Chengdu, 611731, Sichuan, China.
Summary
A new deep learning strategy fuses millimeter wave (mmWave) and visible images for enhanced internal defect detection in obstructed environments. This approach improves accuracy and robustness for real-world penetrating imaging applications.
Area of Science:
- Engineering
- Computer Science
- Materials Science
Background:
- Nondestructive internal defect detection is crucial for various industries.
- Existing imaging techniques face limitations in obstructed environments.
- Millimeter wave (mmWave) imaging offers penetration capabilities, while visible imaging provides surface details.
Purpose of the Study:
- To propose a novel pixel-level information fusion strategy for mmWave and visible images.
- To enhance the capabilities of penetrating-imaging applications, particularly for internal defect detection and localization.
- To address challenges posed by obstructed environments in imaging.
Main Methods:
- A deep transfer learning strategy is employed to capture information from both mmWave and visible images.
- Fine-tuning strategies are implemented to optimize the model's performance.
- An improved bilateral filter is utilized to robustly exploit information from both imaging modalities.
Main Results:
- The proposed fusion strategy effectively integrates information from near-field mmWave and visible light fields.
- Experimental results demonstrate superior performance in terms of accuracy and robustness.
- The method shows significant potential for real-world penetrating imaging applications.
Conclusions:
- The developed deep learning-based fusion strategy offers a robust solution for internal defect detection.
- This approach enhances the reliability and accuracy of imaging under obstructed conditions.
- The findings pave the way for advanced nondestructive testing and evaluation methods.
