Angle-independent spectropolarimetric target classification using machine learning

Area of Science:

• Remote Sensing
• Optical Engineering
• Material Science

Background:

• Light polarization differs between manmade and natural objects.
• Polarization measurements are valuable for remote sensing applications.
• Differentiating artificial and natural targets is crucial in various fields.

Purpose of the Study:

• To present a novel method for characterizing unknown targets.
• To utilize hyper-spectral polarimetric LiDAR for material classification.
• To develop a robust algorithm for distinguishing artificial from natural objects.

Main Methods:

• Implementation of a hyper-spectral polarimetric light detection and ranging (LiDAR) system.
• Application of a k-nearest neighbors (KNN) algorithm for material classification.
• Testing with incident angles from -10° to 60° without target orientation assumptions.

Main Results:

• Achieved a balanced accuracy of 86.9% in material classification.
• Demonstrated effective differentiation of artificial and natural objects using polarization.
• Validated the KNN algorithm's performance across a range of incident angles.

Conclusions:

• Hyper-spectral polarimetric LiDAR is effective for target characterization.
• Polarization signatures provide a reliable means to distinguish manmade from natural materials.
• The developed KNN method offers a robust solution for remote sensing material identification.