Calculation model of the Hyperion reflected light scattering efficiency.

Area of Science:

• Remote Sensing
• Optical Physics
• Machine Learning

Background:

• Space-borne sensors provide crucial spatial image data for ground target observation.
• Existing methods for target classification require robust physical characteristic analysis.

Purpose of the Study:

• To introduce and validate the scattering efficiency per unit area (SEA) as a characteristic signal for ground target classification.
• To enhance classification accuracy using advanced data processing techniques.

Main Methods:

• Derivation of the scattering efficiency per unit area (SEA) parameter.
• Application of principal component analysis (PCA) to select noise-unaffected spectral bands.
• Integration of PCA-selected bands with a backpropagation neural network (BPNN) for media identification.

Main Results:

• Scattering efficiency per unit area (SEA) values vary across Hyperion wavelengths, serving as a unique target signature.
• Principal component analysis effectively identified characteristic bands, reducing noise interference.
• The combined PCA-BPNN model achieved high accuracy in identifying five distinct ground media with minimal residuals.

Conclusions:

• Scattering efficiency per unit area (SEA) is a viable physical parameter for characterizing ground target scattering.
• The developed methodology demonstrates the potential of SEA and machine learning for accurate remote sensing-based target classification.