Multispectral Imaging

Overview

For my undergraduate thesis, I designed a complete multispectral imaging system to estimate fruit maturity. With no existing datasets available, I built the physical acquisition setup, collected and curated the data, and trained machine learning models to classify ripeness stages.

I emphasized interpretability, analyzing which spectral bands contributed most to prediction accuracy. This project exemplifies my approach of building end-to-end systems from raw signals rather than relying on curated inputs.

Key Contributions

• Designed custom illumination and image acquisition setup
• Built original multispectral and UV datasets
• Engineered spectral features and trained classification models
• Focused on interpretability and robustness of predictions

Technologies