Surface identification from multiband LADAR reflectance with varied incidence angle via database mapping.

Area of Science:

• Remote Sensing
• Data Science
• Materials Science

Background:

• LADAR reflection characteristics are influenced by material properties and surface textures.
• These dependencies can be leveraged for effective surface identification.
• Multiband LADAR offers potential for enhanced surface characterization.

Purpose of the Study:

• To assess surface identification using multiband LADAR reflected radiance.
• To investigate the effectiveness of stochastic neighborhood embedding (t-SNE) in analyzing LADAR data.
• To determine if t-SNE can overcome challenges posed by unknown incident angle dependencies.

Main Methods:

• Utilized the nonconventional exploitation factors data system database.
• Applied a statistics-based dimension reduction algorithm, t-SNE, to LADAR reflected radiance and band ratios.
• Analyzed data clouds generated from monostatic LADAR measurements.

Main Results:

• t-SNE effectively separated data clouds from multiband LADAR reflected radiance and band ratios.
• This separation enabled surface identification despite unknown incident angle dependencies.
• Significant improvements in surface identification were observed using t-SNE mapping.

Conclusions:

• Multiband LADAR reflectance, analyzed with t-SNE, provides a viable method for surface identification.
• t-SNE enhances the robustness of LADAR-based surface identification against variable incident angles.
• This approach facilitates accurate surface identification in complex environments.