旋光法
散射
计算机科学
合成孔径雷达
遥感
雷达
可视化
雷达成像
表征(材料科学)
极化(电化学)
反向散射(电子邮件)
逆合成孔径雷达
人工智能
模式识别(心理学)
雷达截面
计算
双基地雷达
物理
计算机视觉
作者
Haoliang Li,SiWei Chen
标识
DOI:10.1109/tgrs.2025.3647123
摘要
Polarimetric radar enables the acquisition of fully polarimetric information from targets. The electromagnetic scattering response of radar targets exhibits strong dependencies on the polarization basis, time, radar frequency, and incident angle. Most existing methods for characterizing target scattering mechanisms are limited to single-domain features or independent multi-domain features, thereby hindering the achievement of fine target understanding and investigation. This work is dedicated to addressing this issue, with the core idea of transforming from the traditional paradigm of independent single-domain representation to a joint-domain processing framework. First, a polarimetric scattering tensor characterization approach is constructed to integrate target scattering information across multiple domains, including the polarimetric, time, frequency, and spatial domains. Then, a visualization and characterization tool named general polarimetric correlation pattern (GPCP) is proposed for mining and utilizing target joint-domain scattering characteristics. This tool enables the visualization of scattering diversity variations across different dimensions, which encompass rich information rarely considered in existing studies. Investigations are conducted to reveal canonical structures’ latent scattering patterns using electromagnetic computation data. On this basis, a set of polarimetric features, including 8 geometric features and 8 statistical features, are derived to quantitatively characterize these patterns. Quantitative investigations based on the target-to-clutter ratio (TCR) index with Radarsat-2 and Gaofen-3 polarimetric synthetic aperture radar (PolSAR) datasets demonstrate the superiority of the proposed polarimetric features in enhancing ship target contrast. Finally, potential future applications of the proposed GPCP are discussed.
科研通智能强力驱动
Strongly Powered by AbleSci AI