Problem Statement

The project aims to enhance precision agriculture by accurately estimating sugarcane yields using satellite imagery. It addresses challenges such as identifying water and nitrogen stress factors, estimating crop heights, and optimizing farming practices across diverse farm sizes for better resource management and sustainability.

Approach Used

The project integrates geospatial analysis and remote sensing to calculate indices like NDVI for stress assessment. A Support Vector Regression (SVR) model is used for yield prediction, while CNNs classify farm plots based on stress levels. Geospatial tools and QGIS are employed to extract and analyze plots, and the methodology is validated through collaborations with local farms and test plots.

Results

• Achieved a classification accuracy of 72% for stress detection using CNNs.
• Predicted sugarcane yields with promising results across diverse farm sizes, identifying yield variations due to stress levels.
• Developed insights for water and nutrient stress management to support sustainable agricultural practices.

Technology Used

• Models: Support Vector Regression (SVR) for yield prediction and CNNs for stress classification.
• Indices: Normalized Difference Vegetation Index (NDVI) for analyzing stress levels.
• Tools: Sentinel-2 satellite imagery for remote sensing, QGIS for geospatial mapping, and Python libraries for data processing and modeling.
• Techniques: Data harmonization, temporal normalization, and visualization of farm plots for improved accuracy.