Loading repository data…
starfiee / repository
Automated satellite imagery pipeline using Sentinel-2 and Google Earth Engine for field-level NDVI analysis
A transparent discovery signal based on current public GitHub metadata.
This score does not audit code, security, maintainers, documentation quality, or suitability. Verify the repository and its current documentation before adoption.
Automated satellite imagery pipeline that accepts field polygon coordinates, retrieves clean Sentinel-2 imagery using field-level cloud scoring, computes NDVI statistics, stores per-pixel data to a database, and generates vegetation maps on demand.
Built for production use at a precision agriculture startup in Pakistan.
src/phase2/ndvi_pipeline.py
src/phase3/data_collector.py
src/phase3/png_generator.py
Major architectural upgrade over Phase 2:
Why pixel-level storage matters:
Tree pixel identification — pixels that stay consistently above 0.6 NDVI across seasons can be flagged as permanent vegetation and excluded from crop calculations improving accuracy
Time series per pixel — each pixel builds its own NDVI history over time enabling crop growth stage detection and yield prediction
ML and deep learning ready — pixel data with lat/lng coordinates and class labels is the exact format needed for training crop classification and disease detection models
Each candidate image is scored at two levels:
Only images with field-level cloud probability below 25% are accepted.
This prevents cases where a tile appears mostly clear but the specific field polygon is under cloud cover — a critical distinction for small agricultural fields in Pakistan.
See database/schema.sql for complete table definitions.
Two tables:
ndvi_pixels — one row per pixel per field per date field_id, image_date, pixel_row, pixel_col,
latitude, longitude, ndvi_value, cloud_prob, ndvi_class
ndvi_results — one row per field per date summary field_id, image_date, ndvi_min, ndvi_max, ndvi_mean,
cloud_prob, total_pixels, valid_pixels, crop_cover_pct, status
git clone https://github.com/starfiee/ndvi-pipeline.git
cd ndvi-pipeline
pip install -r requirements.txt
earthengine authenticate
cp .env.example .env
Edit .env and add your GEE project ID.
Import database/schema.sql into your MySQL database.
python src/phase2/ndvi_pipeline.py 14
31.181412 31.181004 31.180714 31.180635
31.180370 31.180285 31.180252 31.179164
31.179100 31.179600 31.179972 31.179927
31.180291 31.180273
70.872642 70.871164 70.871153 70.869695
70.869708 70.868820 70.868386 70.868454
70.869017 70.871185 70.871865 70.872600
70.872497 70.872891
12362 2026-05-06
python src/phase3/data_collector.py 14
31.181412 31.181004 31.180714 31.180635
31.180370 31.180285 31.180252 31.179164
31.179100 31.179600 31.179972 31.179927
31.180291 31.180273
70.872642 70.871164 70.871153 70.869695
70.869708 70.868820 70.868386 70.868454
70.869017 70.871185 70.871865 70.872600
70.872497 70.872891
12362 2026-05-06
All 3 maps python src/phase3/png_generator.py 12362 2026-05-07 all NDVI map only python src/phase3/png_generator.py 12362 2026-05-07 ndvi Crop cover only python src/phase3/png_generator.py 12362 2026-05-07 cropcover
| NDVI Map | Vegetation Classes | Crop Cover |
|---|---|---|
Sawie-ndvi-parameters
0.0868 0.7801 0.3396
[{"class":"Water","area_ha":2.9552},{"class":"Builtup Area","area_ha":3.7483},...]
Sawie-crop-cover
77.1
image_date:2026-05-07
cloud_prob:0.1433
status:clean
created_at:2026-06-15 09:54:18
ndvi-pipeline/
├── src/
│ ├── phase2/
│ │ └── ndvi_pipeline.py # Field NDVI with cloud filtering
│ └── phase3/
│ ├── data_collector.py # Pixel storage to database
│ └── png_generator.py # On-demand PNG generation
├── database/
│ └── schema.sql # MySQL table definitions
├── demo/
│ └── sample_output/ # Sample PNG outputs
├── docs/
│ └── how_it_works.md
├── .env.example
├── requirements.txt
└── README.md
Python · Google Earth Engine · Sentinel-2 L2A · GeoPandas · Shapely · Rasterio · NumPy · Matplotlib · MySQL · mysql-connector-python
Selected from shared topics, language and repository description—not editorial ratings.
Jarpula-Nirjala /
Automated correction of shifted Maharashtra cadastral plot boundaries using satellite imagery, UTM median shift, cross-correlation alignment, and multi-signal confidence scoring. Achieves median IoU 0.807 vs official baseline 0.612 (+0.203 improvement, 100% plots improved).
55/100 health