Related Experiment Video
Updated: Mar 19, 2026
10:53
Image-guided, Laser-based Fabrication of Vascular-derived Microfluidic Networks
Published on: January 3, 2017
10.4K
Supervised learning for roughness reconstruction under different scanning modes using a confocal laser scanning
Applied Optics
|March 17, 2026
Summary
Artificial intelligence (AI) enhances industrial confocal laser scanning microscopy (CLSM) by improving image quality in fast scanning modes. This AI-driven approach achieves high precision comparable to slower methods, boosting efficiency in surface metrology.
Area of Science:
- Metrology
- Microscopy
- Artificial Intelligence
Background:
- Confocal laser scanning microscopy (CLSM) faces a trade-off between acquisition speed and precision.
- Existing methods struggle to balance rapid data capture with high-quality imaging in industrial applications.
Purpose of the Study:
- To develop an AI-driven framework to overcome the speed-precision limitations in industrial CLSM.
- To enhance signal-to-noise ratio (SNR) and accuracy in fast-acquired microscopy data.
Main Methods:
- A novel neural network, leaky ReLU residual-in-residual neural network (LRIRN), was developed.
- A dataset was created using fast scanning mode (FaSM) and fine scanning mode (FiSM).
- A task-specific loss function and a comprehensive evaluation scheme with novel metrics were employed.
Main Results:
- The AI framework significantly improved the SNR of FaSM data to levels comparable with FiSM.
- The method effectively corrected displacement deviations in the y and z directions.
- Qualitative and quantitative validation confirmed the framework's performance.
Conclusions:
- The proposed AI method offers a solution for high-speed, high-precision industrial surface metrology using CLSM.
- This approach has the potential to revolutionize efficiency and accuracy in industrial measurements.
