VNIR/SWIR Multispectral Polarimetric Imager for Polymer Discrimination and Identification.

Area of Science:

• Remote Sensing
• Optical Engineering
• Materials Science

Background:

• Polarimetric multispectral imaging (PMSI) offers rich data for material characterization.
• Existing systems may lack portability or comprehensive calibration.
• Need for advanced imaging for target detection and material identification in diverse environments.

Purpose of the Study:

• To present a portable PMSI system operating in the visible to shortwave infrared (VNIR-SWIR) range.
• To evaluate its efficacy in target detection, discrimination from aquatic backgrounds, and polymer identification.
• To develop and implement a robust radiometric and polarimetric calibration framework.

Main Methods:

• Developed a portable PMSI system with two synchronized cameras, motorized filters, and piezoelectric polarization control.
• Acquired 48 wavelength-polarization measurements per capture.
• Implemented a comprehensive calibration framework including system response, polarization correction, and reflectance normalization.
• Utilized spectral unmixing techniques (VCA, N-FINDR, PPI) for material recoverability assessment.
• Conducted underwater detectability studies under natural illumination.

Main Results:

• Polarimetric information significantly improved polymer class separability (mean gain of 6.9) over intensity-only features.
• Intensity and Degree of Linear Polarization (DoLP) features showed moderate correlation but complementary identification capabilities.
• VCA spectral unmixing provided the best accuracy-complexity trade-off.
• SWIR bands had limited underwater penetration (4 cm), while VNIR bands (430-550 nm) allowed detection up to 20 cm, with DoLP enhancing edges.
• Identification of chemically similar polymers remained challenging.

Conclusions:

• The portable PMSI system effectively extracts intensity and polarimetric features for enhanced material identification and target detection.
• Polarimetric data provides complementary information to spectral data, improving classification accuracy.
• VNIR bands are more suitable for underwater target detection than SWIR due to water's optical properties.
• Future work should focus on increased spectral dimensionality and validation in complex aquatic environments.