极化SAR（合成孔径雷达）凭借其全天时、全天候的优势，在作物分类中具有重要作用。目前，基于机器学习方法构建极化分解特征与作物之间的关联是实现极化SAR分类的重要途径。然而，现有双极化分解方法常依赖于固定的散射模型，仅能获取离散的散射特征，限制了其对作物复杂散射信息的细致表征。为此，本文提出一种基于旋转变换的双极化SAR散射特征提取方法，该方法利用表征双极化空间的两个极化参数构建旋转矢量，观测矩阵在旋转矢量上的连续投影可获得海量连续散射特征，从而更全面地反映不同作物的散射特性，增强类间差异性。本文选取中国甘肃省金昌市与美国加利福尼亚州两个农田实验区进行实验。实验结果显示，与现有双极化分解方法相比，在金昌实验区，新方法分类精度达96.10%，较C_2矩阵、Stokes分解和Cloude分解分别提升4.73%，2.23%和2.83%；在加利福尼亚实验区，分类精度为93.01%，较上述方法分别提升5.07%，3.19%和3.13%，证明本文方法在不同农业场景下均具有稳定的分类性能提升。

Abstract: Objective: In the field of polarimetric synthetic aperture radar (PolSAR) remote sensing, the capability to conduct all-weather, day-and-night observations makes it a pivotal tool for crop classification. Currently, establishing correlations between decomposition features and crop types via machine learning algorithms has become a mainstream approach. However, existing dual-polarimetric decomposition methodologies, such as Stokes or Cloude decompositions, are constrained by their reliance on predefined and fixed scattering models. These conventional approaches are limited to extracting discrete scattering features, which often fail to sufficiently characterize the geometric structures and scattering processes of complex agricultural vegetation, thereby hindering the achievement of higher classification accuracy. Method: To effectively address these issues, the paper proposes a rotation transformation methodology specifically designed for dual-polarimetric data. By using the polarimetric scattering angle α and the polarimetric phase angle δ, a dynamic and continuous rotation vector is constructed. This method projects the dual-polarization covariance matrix C_2 onto this vector, transforming static observations into a continuous power response distribution. Then, by utilizing α, δ, and the continuous scattering power P at these rotation vectors as coordinate axes, a three-dimensional polarimetric space is constructed. Within this visualization framework, the research generates corresponding three-dimensional power response surfaces for different crops. Based on this, a series of key scattering features like skewness and smoothness are extracted. These features capture subtle physical differences that conventional fixed decomposition models fail to detect, thereby enhancing classification robustness in complex agricultural scenes. Result: The efficacy of this approach was rigorously validated through comprehensive experiments in Jinchang, China, and California, USA. These two regions are characterized by distinct climatic conditions and diverse crop planting structures. Quantitative analysis demonstrates that the proposed method achieved an outstanding overall classification accuracy of 96.10% in the Jinchang study area, outperforming C_2 matrix, Stokes decomposition, and Cloude decomposition by 4.73%, 2.23%, and 2.83%, respectively. Similarly, in the California agricultural site, the method yielded a high classification accuracy of 93.01%, providing performance improvements ranging from 3.13% to 5.07% over baseline methods. These results illustrate that the new method consistently delivers superior classification performance across different agricultural landscapes. Conclusion: By constructing a rotation matrix that allows the scattering model to vary continuously, the method transforms discrete dual-polarimetric observations into a continuous power response. This enables the mining of richer geometric and physical information than conventional models. Experimental results confirm that this approach achieves superior and stable classification accuracy across diverse agricultural landscapes. While the current study utilizes single-temporal, single-band data, future work will integrate multi-band SAR and long-term time-series observations. This will better capture phenological growth trajectories, ultimately enhancing ability to distinguish between structurally similar crops across different developmental stages.