High Speed Spectrally Coded Illumination- and Machine Learning- based Hyperspectral Microscope imaging

This project presents a high-speed hyperspectral microscopy framework that integrates spectrally coded illumination with advanced machine learning algorithms to achieve rapid and accurate spectral imaging at microscopic scales. By projecting optimized spectral codes through a programmable illumination module, the system enables fast acquisition of spectrally enriched measurements without relying on slow wavelength scanning mechanisms. A learning-based reconstruction pipeline then transforms these encoded measurements into fast spectral contrast images, preserving fine spectral signatures essential for material identification. Leveraging these reconstructed spectra, machine learning models accurately classify materials used in semiconductor devices, even when their optical characteristics differ only subtly. This approach provides a compact, efficient, and high throughput solution for next generation nanomaterial inspection, semiconductor characterization, and computational microscopy applications.