StaMPS-PyRefactor

StaMPS 持久散射体 InSAR 处理链的 Python 移植与重构

GPL-3.0

本项目基于 StaMPS（Stanford Method for Persistent Scatterers）翻译重构，遵循 GPL-3.0。使用或分发时请保留许可证并引用适当的 StaMPS 文献。本项目非官方 StaMPS 发行版。

概述

StaMPS-PyRefactor 将 MATLAB StaMPS 处理链翻译为 Python 实现，聚焦 ISCE/ISCE2 栈输出的 PS-InSAR 与 SLC 原生 Small Baseline 工作流。中间产物基于 HDF5，参数文件使用 JSON，支持 snaphu 解缠与 GeoPackage/Shapefile 导出。本项目为独立重构，非官方 StaMPS 发行版。

功能特性

ISCE2 直接预处理

prep_isce.py 可直接读取 ISCE2 merged 目录，无需 MATLAB parms.mat。

Steps 1-8 完整实现

从初始加载、gamma 估计、PS 选取、weeding、相位校正、snaphu 解缠到 SCLA/SCN 滤波。

PS 与 SB 双工作流

支持 PS-InSAR 与 SLC 原生 Small Baseline 流程，含 define_sb_pairs 与 build_sb_phase。

HDF5 + JSON 参数

parms.json / localparms.json 替代 MATLAB .mat，优先级 localparms > parms > 代码默认值。

矢量导出

export_results.py 导出速度（mm/yr）与位移时序（mm），默认 GeoPackage，可选 Shapefile。

性能优化

Step 3 gamma 重估计、Step 4/5 合并等关键路径已针对大规模 HDF5 读写优化。

系统架构

ISCE2 merged

prep_isce.py

stamps_main.py (Steps 1-8)

export_results.py

典型 PS 流程：prep_isce 准备补丁与候选点 → stamps_main 执行 Steps 1-8 → export_results 导出 GeoPackage。SB 流程额外使用 define_sb_pairs 与 build_sb_phase 构建 SB 相位。

工作流

prep_isce.py 读取 ISCE2 merged 栈
stamps_main.py Steps 1-8
export_results.py 导出 GeoPackage

python stamps_python/prep_isce.py path/to/merged \
  --output runs/prep_ps \
  --reference-date 20200101 \
  --bootstrap-metadata \
  --range-patches 8 --azimuth-patches 2 \
  --write-step1

python stamps_python/stamps_main.py --start 1 --end 8 --config runs/prep_ps

已完成功能

ISCE2 merged 直接预处理与 Python 元数据引导
StaMPS Steps 1-8 Python 实现与 snaphu 集成
真实 ISCE2 栈全链 Step 8 验证
PS 与 SB 双工作流与 GeoPackage 导出
Step 3/4/5 性能优化与 incidence 角提取

已知限制

SB 工作流仍需更多跨数据集验证
MATLAB 绘图 GUI 未移植
TRAIN 对流层校正未端到端移植
尚未正式打包为可安装 Python 包
自动化测试覆盖仍有限

系统要求

Python 3.10+（已在 Windows Python 3.11 验证）
snaphu 已安装并在 PATH 中可用（Step 6 解缠）
ISCE2 merged 栈目录（含 SLC/、geom_reference/、baselines/）
pip install -r requirements.txt

快速开始

1. 安装依赖
git clone 仓库后执行 pip install -r requirements.txt
2. PS 预处理
python stamps_python/prep_isce.py path/to/merged --output runs/prep --reference-date YYYYMMDD --bootstrap-metadata --write-step1
3. 运行 Steps 1-8
python stamps_python/stamps_main.py --start 1 --end 8 --config runs/prep
4. 导出结果
python stamps_python/export_results.py --input-path runs/prep --output-dir runs/prep/export --format gpkg --correction v-dso

下载

StaMPS-PyRefactor v0.20

2026年6月26日

源码发布

当前版本仅提供源码包，请 clone 仓库后 pip install -r requirements.txt

ZIP (源码)tar.gz (源码)GitHub

更新说明

StaMPS-PyRefactor v0.20 adds the first validated Small Baseline (SBAS) workflow support on top of the existing PS-InSAR Python refactor.

Highlights

Added SLC-native Small Baseline pair definition with define_sb_pairs.py.
Added split SB preprocessing support in prep_isce.py, including SB candidate preparation without immediate phase extraction.
Added build_sb_phase.py to build wrapped SB phase samples from co-registered ISCE2 SLC pairs and write Python-native Step-1 HDF5 products.
Extended Step 1-8 processing paths for SB projects, including SB-aware merge, unwrapping, SCLA/SCN handling, and export behavior.
Updated GeoPackage/Shapefile export so SB time series can be exported through the Python pipeline.
Expanded README and operation manual with the recommended Small Baseline workflow and command examples.
Added a synthetic regression check for split SB preparation plus SLC-native SB phase building.

Validation

Syntax checks passed for the modified Python modules and new SB scripts.
stamps_python/tests/prep_isce_sb_split_check.py passed on the synthetic ISCE2 SB fixture.
git diff --check passed, with only local LF/CRLF conversion warnings from Git on Windows.

Known limitations

SBAS support is now available, but broader numerical equivalence against every MATLAB StaMPS SB branch still needs more datasets.
TRAIN/tropospheric correction integrations remain outside this release.
The project is still a development refactor, not a packaged Python distribution.

This project is an independent GPL-3.0 refactor and is not the official StaMPS distribution.

StaMPS-PyRefactor v0.10

2026年6月21日