FarmOptima AI

Inspiration Farmers often struggle to identify crop health and types over large areas. Governments and researchers need accurate crop maps for food security and planning. Satellite data + AI + decision-making can solve this effectively.

What it does Collects satellite imagery (Sentinel-2, Landsat). Extracts vegetation indices (NDVI, EVI, SAVI, NDWI) from spectral bands. Uses AI models (Random Forest, CNN, or SVM) for crop classification. Enhances decision-making with MCDM (AHP, TOPSIS, ELECTRE) to improve crop identification confidence. Provides visual maps and reports for end-users.

How we built it Data Collection – Sentinel-2, Landsat imagery, ground truth crop datasets (Kaggle, FAO). Preprocessing – cloud masking, normalization, band selection, vegetation index calculation. AI Models – CNN/Random Forest for supervised classification. MCDM – AHP for feature weighting, TOPSIS/ELECTRE for ranking crop candidates. Deployment – Streamlit dashboard + Overleaf/PDF report generation.

Challenges we ran into Handling large satellite imagery (big data). Aligning ground-truth crop labels with satellite images. Balancing AI model accuracy vs. interpretability. Combining AI predictions with MCDM outputs.

Accomplishments that we’re proud of Successfully merged AI with MCDM for crop classification. Built a working prototype that classifies crops at regional scale. Automated report generation for easy stakeholder communication.

What we learned How to preprocess and analyze remote sensing data. Combining AI classification + mathematical decision-making improves accuracy. Deploying geospatial projects on limited resources.

What’s next Scale to multi-season crop monitoring. Integrate with real-time farmer advisory systems. Add yield prediction and climate impact analysis.

Built With

• csv-apis:-copernicus-open-access-hub
• gdal
• google-earth
• google-earth-engine-cloud-services:-aws/gcp-(for-satellite-data-+-model-training)-databases:-geotiff
• languages:-python-(numpy
• pandas
• postgis
• rasterio
• scikit-learn
• tensorflow/pytorch)-frameworks:-streamlit
• usgs