Prediction model from spectral intensity to spectral polarization based on polarization spectral bidirectional...

Area of Science:

• Optics and Photonics
• Materials Science
• Computer Vision

Background:

• Acquiring multi-angle polarization spectral data rapidly is a significant challenge.
• Existing methods struggle with unknown detection angles and cannot predict spectral polarization.
• Polarization Spectral Bidirectional Reflectance Distribution Function (PSBRDF) is crucial for understanding light-material interactions.

Purpose of the Study:

• To develop a prediction model for spectral polarization from spectral intensity.
• To enable efficient acquisition of detailed polarization spectral data.
• To improve the generalization and adaptability of spectral prediction models.

Main Methods:

• A prediction model was developed based on PSBRDF, transforming spectral intensity prediction into eigenvalue prediction.
• The model involves three key steps: dataset preparation, eigenvalue extraction, and prediction model construction.
• Predicted eigenvalues are used to recover spectral intensity and spectral polarization.

Main Results:

• The proposed model achieves high accuracy, with Mean Relative Error (MRE) below 0.03 and R-squared (R²) above 0.99.
• It demonstrates superior generalization and adaptability for unknown detection angles compared to traditional methods.
• The model successfully predicts spectral polarization, a capability lacking in conventional approaches.

Conclusions:

• The developed prediction model offers a viable solution for obtaining challenging polarization spectral data.
• It enhances the ability to predict spectral intensity and polarization, especially under varying conditions.
• This approach advances the field by enabling efficient and comprehensive spectral analysis.