停车场检测 | Parking Lot Detection
Detects parking lots by fusing spectral, textural, and geometric features, supporting urban facility surveys and land-use mapping.
Core algorithm: asphalt score = low-brightness factor × low-vegetation factor (absolute scale); marking density extracts high-frequency bright lines via Sobel gradient + brightness threshold; regularity characterizes row/column periodicity with the ratio of local to global variance; the composite score is a weighted sum clipped to [0,1], and thresholding segments out the parking lots.
Dependencies / 依赖
pip install 'numpy' 'rasterio' 'scipy'
Usage / 使用方法
Basic usage
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 [other options]
Examples
Example 1 (synthetic data (offline))
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
Example 2 (use case 2)
python geoskill-parking-lot-detection.py --input multispectral.tif --output-dir ./out
Example 3 (use case 3)
python geoskill-parking-lot-detection.py --bbox 121.0 31.0 122.0 32.0 --threshold 0.5 --output-dir ./out --quiet
Example 4 (use case 4)
python geoskill-parking-lot-detection.py --input ms.tif --regularity-block 24 --output-dir ./out
Example 5 (use case 5)
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 --synthetic --threshold 0.35 --output-dir ./out --quiet
Output / 输出
| File | Format | Description |
|---|---|---|
parking_score.tif |
GeoTIFF | Two bands: band1=parking score, band2=classification mask |
parking_stats.json |
JSON | Means of score/marking/regularity, parking-lot proportion |
output-manifest.json |
JSON | Run manifest |
Data Source / 数据源 / Source
Local multispectral GeoTIFF (Red, NIR); --synthetic mode simulates parking lots with regular markings plus vegetation/roof control areas.
Privacy / 隐私声明 / Privacy
- Runs offline by default;
--syntheticmode requires no network at all. - All processing is performed locally; user data is never uploaded.
License / License
MIT
name: geoskill-parking-lot-detection description: 'Detect parking lots using asphalt spectral signature, regular row and column texture, and painted marking density.'
停车场检测 | Parking Lot Detection
融合光谱、纹理与几何特征检测停车场，服务于城市设施调查与用地制图。
核心算法：沥青分数 = 低亮度因子 × 低植被因子（绝对标度）；标线密度由 Sobel 梯度 + 亮度阈值提取高频亮线；规则性用局部方差/全局方差之比刻画行列周期性；综合评分 = 加权和，裁剪到 [0,1]，阈值分割出停车场。
依赖
pip install 'numpy' 'rasterio' 'scipy'
使用方法
基本用法
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 [其他参数]
示例
示例 1（合成数据（离线））
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
示例 2（用法 2）
python geoskill-parking-lot-detection.py --input multispectral.tif --output-dir ./out
示例 3（用法 3）
python geoskill-parking-lot-detection.py --bbox 121.0 31.0 122.0 32.0 --threshold 0.5 --output-dir ./out --quiet
示例 4（用法 4）
python geoskill-parking-lot-detection.py --input ms.tif --regularity-block 24 --output-dir ./out
示例 5（用法 5）
python geoskill-parking-lot-detection.py --bbox 116.0 39.0 117.0 40.0 --synthetic --threshold 0.35 --output-dir ./out --quiet
输出
| 文件 | 格式 | 说明 |
|---|---|---|
parking_score.tif |
GeoTIFF | 双波段：band1=停车场评分，band2=分类掩膜 |
parking_stats.json |
JSON | 评分/标线/规则性均值、停车场比例 |
output-manifest.json |
JSON | 运行清单 |
数据源 / Source
本地多光谱 GeoTIFF（Red, NIR）；--synthetic 模式模拟含规则标线的停车场与植被/屋顶对照区。
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