Abstract
The polarized bidirectional reflectance distribution function (pBRDF) describes the changes between the Stokes vectors of incident and reflected light. Existing models typically describe the spatial distribution of a target's polarization characteristics at a single wavelength, leaving a gap in describing polarized spectral characteristics. This paper proposes a modified three-component polarized spectral bidirectional reflectance distribution function (pSBRDF). By combining the Fresnel equation in the specular reflection component with Cauchy's empirical dispersion equation, the model introduces wavelength variables and wavelength-independent dispersion constants. Validated using coated fabric samples across the 400 to 760 nm range, the model controlled spectral and spatial distribution errors within 0.00743 and 0.0757, respectively. This research provides a theoretical and model-based foundation for advanced polarization detection and target analysis.
Introduction
In optical sensing, the Bidirectional Reflectance Distribution Function (BRDF) is the standard for describing how light interacts with a surface. Adding polarization (pBRDF) provides a "fingerprint" of a material’s texture and chemical composition. However, most existing pBRDF models are monochromatic, failing to account for how polarization evolves across the visible spectrum. This "polarization spectral correlation" is vital for identifying targets in complex environments. This paper introduces a unified pSBRDF model that handles both spatial and spectral variables simultaneously.
The pSBRDF Model
The proposed model utilizes a three-component approach to characterize light reflection, focusing on the complex refractive index of the material.
- Specular Reflection: The core innovation integrates Cauchy’s empirical dispersion equation into the specular component. This allows the refractive index ($n$) and the extinction coefficient ($k$) to be treated as functions of wavelength ($\lambda$). By solving for stable dispersion constants, the model predicts polarization at any visible wavelength.
- Multiple Reflection: This accounts for light bouncing between surface microfacets, modeled using a modified Minnaert function that incorporates surface roughness.
- Volume Scattering: This describes light that penetrates the material, scatters internally, and re-emerges, modeled using a simpler cosine function.
Experimental Setup and Datasets
To validate the model, researchers used the Northeast Normal University Laboratory Goniospectrometer System (NENULGS). The setup included an ASD spectrometer and a Glan–Thompson polarizer to measure the Degree of Linear Polarization (DoLP).
- Samples: Two coated fabrics ("Desert" and "Soil" colors) were tested for their relevance in camouflage and remote sensing applications.
- Spectral Range: 400 nm to 760 nm.
- Geometry: Measurements were taken at incident angles of 30°, 40°, and 50°, with detection angles spanning the upper hemisphere.
Key Insights and Results
The researchers discovered that while refractive indices vary with color, the material's dispersion constants remain stable. By extending Cauchy’s equation to describe the extinction coefficient ($k$), the model accurately captures how a material absorbs light across different wavelengths.
Experimental data confirmed that DoLP generally increases with the incident angle but decreases as the wavelength increases. The model demonstrated high precision:
- Spectral Accuracy: The Residual Sum of Squares (RSS) for the desert sample was limited to 0.00743.
- Spatial Accuracy: The spatial distribution error for the soil sample was kept within 0.0757.
- Predictive Power: The model successfully predicted that forward scattering produces higher polarization than backscattering, with maximum DoLP appearing at high detection zenith angles (60°).
Conclusion
This pSBRDF model successfully bridges the gap between spatial and spectral polarization data. While currently validated on coated fabrics in the visible band, future work will involve testing highly metallic or translucent materials and extending the spectral range into the Short-Wave (SWIR) and Long-Wave Infrared (LWIR) bands. Optimizing the inversion algorithm for real-time detection remains a priority for practical engineering applications.
Summary updated