Researchers have developed CV-SSMNet, a novel physics-aware complex-valued state-space network designed for polarimetric synthetic aperture radar (PolSAR) image classification. This model effectively captures long-range spatial dependencies while preserving crucial amplitude-phase information inherent in PolSAR data. By integrating seven physically meaningful scattering priors as modulation signals, CV-SSMNet adaptively recalibrates complex-valued representations throughout the feature evolution process. Experiments on L-band and P-band datasets demonstrate that CV-SSMNet achieves competitive accuracy, enhanced regional consistency, and improved boundary preservation, validating the approach of embedding polarimetric scattering mechanisms into deep learning for GeoAI representation learning. AI
IMPACT This model could improve the accuracy and consistency of GeoAI applications by better integrating physical scattering mechanisms into deep learning.
RANK_REASON The cluster contains an academic paper detailing a new model for image classification. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
