测绘学报 ›› 2026, Vol. 55 ›› Issue (10): 1780-1796.doi: 10.11947/j.AGCS.2026.20260108

• 大地测量学与导航 • 上一篇    下一篇

顾及频域量化误差补偿的圆方差迭代干涉相位滤波算法

李坚1(), 范洪冬1(), 高晓雄2, 王俊1   

  1. 1.中国矿业大学环境与测绘学院,江苏 徐州 221116
    2.浙江省测绘科学技术研究院,浙江 杭州 310030
  • 收稿日期:2026-03-26 修回日期:2026-09-14 出版日期:2026-11-16 发布日期:2026-11-16
  • 通讯作者: 范洪冬 E-mail:ts24160020a31ld@cumt.edu.cn;cumtfanhd@163.com
  • 作者简介:李坚(2002—),男,博士生,研究方向为InSAR数据处理及其在地表形变监测领域的应用。E-mail:ts24160020a31ld@cumt.edu.cn
  • 基金资助:
    国家自然科学基金(42274054)

A circular variance iterative interferometric phase filtering algorithm with frequency-domain quantization error compensation

Jian Li1(), Hongdong Fan1(), Xiaoxiong Gao2, Jun Wang1   

  1. 1.School of Environment and Spatial Informatics, China University of Mining and Technology, Xuzhou 221116, China
    2.Zhejiang Academy of Surveying and Mapping, Hangzhou 310030, China
  • Received:2026-03-26 Revised:2026-09-14 Online:2026-11-16 Published:2026-11-16
  • Contact: Hongdong Fan E-mail:ts24160020a31ld@cumt.edu.cn;cumtfanhd@163.com
  • About author:Li Jian (2002—), male, PhD candidate, majors in InSAR data processing and its application in surface deformation monitoring. E-mail: ts24160020a31ld@cumt.edu.cn
  • Supported by:
    The National Natural Science Foundation of China(42274054)

摘要:

合成孔径雷达干涉测量(InSAR)在复杂地形起伏与低相干环境下易受强烈的散斑噪声干扰,导致干涉条纹拓扑结构破坏并诱发相位解缠误差传递。现有滤波算法在处理该类区域时,难以兼顾噪声抑制与干涉条纹恢复。针对上述问题,本文提出一种顾及频域量化误差补偿的圆方差迭代干涉相位滤波算法(QEC-CVIF)。该算法构建了一种干涉相位的非线性迭代演化模型:在每次迭代中,通过频域自适应阈值对频谱进行非线性加权以提取候选相位,并引入方向统计学的圆方差作为演化方程中迭代步长的非线性控制权重,驱动相位自适应更新,使噪声区域获得较大更新幅度而结构区域趋于保留,并在迭代完成后利用瞬时频率补偿修正像素级相位阶跃。通过多组典型拓扑特征的仿真数据,以及玛多地震、Etna火山和Aykol地震等真实InSAR数据开展滤波与解缠试验。定性与定量分析结果表明:在面临强噪声与复杂地形叠加的干涉场景下,与经典的非局部均值及三维协同滤波等基准算法相比,本文算法在结构相似度、均方根误差等指标上均取得较优结果;该算法通过包容局部微小自然纹理的适度弱化,换取了全局干涉相位的拓扑连续性,最大限度地消除了相位残差点,有效抑制了解缠误差的空间传递,为高精度的地形高程反演与地表形变提取提供了更为可靠的包裹相位基础。

关键词: InSAR, 相位滤波, 方向统计学, 圆方差, 频率补偿, 相位解缠

Abstract:

Interferometric synthetic aperture radar (InSAR) is susceptible to severe speckle noise interference in environments with complex topographical variations and low coherence, which destroys the topological structure of interference fringes and induces the propagation of phase unwrapping errors. When processing such regions, existing filtering algorithms struggle to balance denoising intensity with the complete restoration of interference fringes. To address these issues, this paper proposes a quantization error compensated circular variance iterative filter (QEC-CVIF) for interferometric phase. The algorithm constructs a nonlinear iterative evolution model for the interferometric phase: in each iteration, it extracts candidate phases by applying a self-constructed frequency-domain adaptive threshold for nonlinear weighting, and introduces the circular variance in directional statistics as a nonlinear control weight for the iterative step size within the evolution equation. This mechanism drives the phase to update adaptively, ensuring that noisy regions receive larger update amplitudes while structured regions tend to be preserved. It corrects pixel-level phase steps via instantaneous frequency compensation after the iteration terminates. Filtering and unwrapping experiments were conducted using multiple sets of simulated data with typical topological features, as well as real InSAR datasets from the Maduo earthquake, Mount Etna, and Aykol earthquake. Qualitative and quantitative analysis results demonstrate that, under interferometric scenarios superimposed with strong noise and complex terrain, compared with baseline algorithms such as classic non-local means and 3D collaborative filtering, the proposed algorithm achieves better performance in indicators such as structural similarity and root mean square error; by tolerating a moderate attenuation of local minor natural textures, the algorithm exchanges for strict topological continuity of the global interferometric phase, eliminates phase residue points to the greatest extent, and effectively suppresses the spatial propagation of unwrapping errors. This provides a more reliable wrapped phase foundation for high-precision terrain elevation inversion and surface deformation extraction.

Key words: InSAR, phase filtering, directional statistics, circular variance, frequency compensation, phase unwrapping

中图分类号: