Interferometric Phase Map Segmentation Algorithm Based on an Adaptive Optimization Tolerance Granularity Space Model

doi:10.1109/jstars.2026.3709326

Interferometric Phase Map Segmentation Algorithm Based on an Adaptive Optimization Tolerance Granularity Space Model

2026-01-01

0

PRE

AI

DOI：10.1109/jstars.2026.3709326

Original

Original request for help

Save

Abstract

En 中文

In large-scale and complex environments, phase unwrapping using a unified method over the entire interferometric phase map (IPM) of interferometric synthetic aperture radar (InSAR) often suffers from error propagation, where noise in low-quality regions spreads across the IPM and degrades the overall unwrapping accuracy. To address this issue, segmenting IPMs based on phase quality and noise distribution, followed by region-specific processing, is a reasonable and feasible strategy. An IPM segmentation algorithm based on adaptively optimized tolerance granular space model (TGSM) is proposed. The IPMs are first uniformly divided into subblocks, each treated as an original object of TGSM. Five-dimensional features, including amplitude variance of interferogram and phase gradient standard deviation, among others, are used for segmentation. A weighted squared Euclidean distance is introduced serves as the similarity measure within the tolerance relation, and recursive merging of tolerance granules is controlled by searching for an optimal similarity measure threshold. In this way, a coarse-to-fine multilevel TGSM is constructed to partition complex InSAR IPMs into multiple regions. To enhance performance, an adaptive genetic algorithm is used to determine feature weights; an adaptive similarity measure threshold searching and threshold interval boundary extension strategy is adopted to improve merging reliability; small subregion merging is employed to avoid postunwrapping stitching errors that may outweigh the benefits of preunwrapping segmentation. Experiments with Sentinel-1A data show that the proposed algorithm can effectively segments the varying quality regions of IPMs. Its results outperform the other three comparison methods on all four used evaluation metrics.

Keywords:

Algorithms

Optimization

Merging

Modeling

Distance measurement

Educational institutions

Indexes

Indexing

Art

Noise

Interferometric phase map (IPM) segmentation

tolerant granular space

complex geographical environment

genetic algorithm

adaptive threshold interval expansion

Interferometric Phase Map Segmentation Algorithm Based on an Adaptive Optimization Tolerance Granularity Space Model

Abstract

Journal

Organization

Cited Papers