基于弱监督的超表面传感器设计方法、装置和电子设备
By combining a dual-branch encoder network and a global context injection module with a progressive resolution decoder, a weakly supervised method is used to solve the problems of insufficient training samples and insufficient prediction accuracy in the design of metasurface sensors. This enables accurate prediction of structural parameters and spectral responses, simplifies the design process, and reduces costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 浙江优众新材料科技有限公司
- Filing Date
- 2026-04-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing metasurface sensor design methods suffer from problems such as difficulty in obtaining training samples, insufficient high-precision spectral feature capture, and lack of model interpretability, resulting in inaccurate sensor performance evaluation. In particular, peak position shifts and inaccurate linewidth predictions often occur when dealing with sharp resonance peak optical responses.
We adopt a weakly supervised metasurface sensor design method, which extracts local and global features through a dual-branch encoder network, and improves prediction accuracy step by step by combining a global context injection module and a progressive resolution decoder. Furthermore, we simplify the design process and reduce the dependence on large-scale labeled datasets by iteratively optimizing the total task loss.
It achieves accurate prediction of structural parameters and spectral response, simplifies the design process, improves design efficiency, reduces costs, solves the problems of separation between local details and global laws and insufficient prediction accuracy, and enhances the model's generalization ability and interpretability.
Smart Images
Figure CN121997776B_ABST