Classification of Crops at Different Nitrogen Levels Using Drone-Based Hyperspectral Imaging: A Machine Learning Approach

Efficient nitrogen (N) management is essential for sustainable agriculture, as improper N fertilizer use can cause environmental issues like water pollution and greenhouse gas emissions. Hyperspectral imaging (HSI) offers a high-resolution, non-invasive approach to assess N levels. Unlike previous studies focusing on broader landscapes, this research targets plant-level nitrogen mapping with drone-based hyperspectral data. 139 spectral channels with 4 nm resolution in visible to Near Infrared region were used in the study. Machine learning models, such as Random Forest, Support Vector Machine (SVM), and Gradient Boosting, were utilized for crop variety classification and nitrogen level prediction, with SVM achieving a classification accuracy of 95%. This study demonstrates the potential of hyperspectral imaging combined with machine learning in precision agriculture, providing insights into optimizing nitrogen application and reducing environmental impact.