geoskill-polarimetric-decomposition

极化SAR分解：Cloude-Pottier H/A/α 特征分解与 Freeman 三分量分解，输出散射熵、各向异性、散射角与表面/二面角/体散射功率

极化SAR分解 | Polarimetric SAR Decomposition

Performs polarimetric target decomposition on fully polarimetric SAR data to extract the physical scattering mechanisms of surface features. Two mainstream methods are implemented:

• Cloude-Pottier (H/A/α) eigen decomposition: performs a Hermitian eigen decomposition of the 3×3 coherency matrix T3 for each pixel, yielding the entropy H (scattering randomness; 0 = a single mechanism, 1 = fully random), the anisotropy A and the scattering angle α (surface scattering α≈0–30°, volume scattering α≈40–60°, dihedral scattering α≈90°).
• Freeman-Durden three-component decomposition (simplified): solves for the surface scattering Ps, double-bounce scattering Pd and volume scattering Pv from |Shh|², |Svv|² and |Shv|².

Application Scenarios / 应用场景

• Land-cover classification and scattering mechanism identification (urban buildings, forest, bare soil, water)
• Extraction of forest structural parameters, crop growth monitoring
• Physical feature engineering for polarimetric SAR data

Dependencies / 依赖

bash

pip install 'numpy' 'rasterio' 'scipy'

Input Conventions / 输入约定

• Cloude / H/A/α: the input GeoTIFF must contain 9 bands encoding the upper triangle of T3 as [T11, T22, T33, Re T12, Im T12, Re T13, Im T13, Re T23, Im T23].
• Freeman: the input GeoTIFF must contain 3 bands [|Shh|², |Svv|², |Shv|²].

Usage / 使用方法

Example 1 (synthetic data, offline)

bash

python geoskill-polarimetric-decomposition.py --bbox 116.0 39.0 117.0 40.0 --method cloude --output-dir ./out

Example 2 (Freeman three-component synthetic)

bash

python geoskill-polarimetric-decomposition.py --bbox 116.0 39.0 117.0 40.0 --method freeman --synthetic --output-dir ./out

Example 3 (real T3 data)

bash

python geoskill-polarimetric-decomposition.py --input T3_9band.tif --method ha_alpha --output-dir ./out

Example 4 (real C3 three-component)

bash

python geoskill-polarimetric-decomposition.py --input C3_3band.tif --method freeman --output-dir ./out

Output / 输出

File	Format	Description
cloude_H_A_alpha.tif	GeoTIFF (3 band)	Entropy H, anisotropy A, scattering angle α (degrees)
freeman_three_component.tif	GeoTIFF (3 band)	Ps / Pd / Pv scattering powers
decomposition_stats.json	JSON	Statistics and synthetic ground truth
output-manifest.json	JSON	Run manifest

Data Source / 数据源 / Source

Local polarimetric GeoTIFF (T3 9-band or C3 3-band), or physically simulated scenes via --synthetic.

Privacy / 隐私声明 / Privacy

• Runs offline by default; --synthetic mode requires no network at all.
• All processing is performed locally; user data is never uploaded.

License / License

MIT

---

---

name: geoskill-polarimetric-decomposition description: '极化SAR分解：Cloude-Pottier H/A/α 特征分解与 Freeman 三分量分解，输出散射熵、各向异性、散射角与表面/二面角/体散射功率'

极化SAR分解 | Polarimetric SAR Decomposition

对全极化 SAR 数据执行极化目标分解，提取地物的散射物理机制。实现两类主流方法：

• Cloude-Pottier（H/A/α）特征分解：对每个像元的 3×3 相干矩阵 T3 做 Hermitian 特征分解，得到熵 H（散射随机性，0=单一机制、1=完全随机）、各向异性 A 与散射角 α （表面散射 α≈0-30°、体散射 α≈40-60°、二面角散射 α≈90°）。
• Freeman-Durden 三分量分解（简化）：从 |Shh|²、|Svv|²、|Shv|² 求解表面散射 Ps、 二面角散射 Pd 与体散射 Pv。

应用场景

• 地物分类与散射机制识别（城市建筑、森林、裸土、水体）
• 森林结构参数提取、农作物长势监测
• 极化 SAR 数据的物理特征工程

依赖

bash

pip install 'numpy' 'rasterio' 'scipy'

输入约定

• Cloude / H/A/α：输入 GeoTIFF 需含 9 个波段，按 T3 上三角编码 [T11, T22, T33, Re T12, Im T12, Re T13, Im T13, Re T23, Im T23]。
• Freeman：输入 GeoTIFF 需含 3 个波段 [|Shh|², |Svv|², |Shv|²]。

使用方法

示例 1（合成数据，离线）

bash

python geoskill-polarimetric-decomposition.py --bbox 116.0 39.0 117.0 40.0 --method cloude --output-dir ./out

示例 2（Freeman 三分量合成）

bash

python geoskill-polarimetric-decomposition.py --bbox 116.0 39.0 117.0 40.0 --method freeman --synthetic --output-dir ./out

示例 3（真实 T3 数据）

bash

python geoskill-polarimetric-decomposition.py --input T3_9band.tif --method ha_alpha --output-dir ./out

示例 4（真实 C3 三分量）

bash

python geoskill-polarimetric-decomposition.py --input C3_3band.tif --method freeman --output-dir ./out

输出

文件	格式	说明
cloude_H_A_alpha.tif	GeoTIFF (3 band)	熵 H、各向异性 A、散射角 α（度）
freeman_three_component.tif	GeoTIFF (3 band)	Ps / Pd / Pv 散射功率
decomposition_stats.json	JSON	统计量与合成真值
output-manifest.json	JSON	运行清单

数据源 / Source

本地极化 GeoTIFF（T3 9 波段或 C3 3 波段），或 --synthetic 物理模拟场景。

隐私声明 / Privacy

• 默认离线运行，--synthetic 模式完全无网络。
• 所有处理在本地完成，不上传用户数据。

License

MIT