Problem Statement

The project tackles the issue of stubble burning in Punjab, India, a significant contributor to air pollution and public health hazards. It aims to predict stubble burning hotspots, identify sustainable alternatives to crop residue burning, and create practical frameworks to incentivize environmentally friendly farming practices.

Approach Used

The project develops a machine learning model to predict stubble burning incidents with 53% accuracy by analyzing geospatial and historical data. On-ground validation ensures real-world applicability, while a Django-based web application enables real-time data logging and model refinement. It also explores alternatives for crop residue reuse, including industrial applications, and proposes logistical and economic solutions for implementation.

Results

• Achieved 53% accuracy in identifying hotspots, focusing on reducing false positives in non-burned areas (Future goal to improve this accuracy).
• Developed a Django-based app to improve data collection, enhance validation accuracy, and support field researchers.
• Identified viable crop residue reuse methods, offering economic incentives to farmers while reducing environmental harm.

Technology Used

• Models: Random Forest model for prediction and geospatial analysis.
• Tools: Django-based web application for data logging and validation.
• Datasets: Sentinel and AVIRIS satellite imagery for high-resolution monitoring.
• Indices: Environmental indices like Normalized Burn Ratio (NBR) and Burned Area Index (BAI) to detect burn scars and assess vegetation health.