New AI Sensor 'Sniffs' Out Spectral Targets (newscenter.lbl.gov)

Researchers at the Lawrence Berkeley National Laboratory have developed an innovative AI-enhanced sensor that significantly improves the capabilities of spectral imaging tools. This groundbreaking sensor can "sniff and seek" spectral targets in real-time by integrating machine learning directly into its design, thus eliminating the traditional data-processing bottleneck associated with spectral imaging technology. The new approach allows for rapid object identification and material characterization, enhancing applications in semiconductor fabrication, pollutant tracking, and crop monitoring. The key innovation lies in the sensor's ability to perform AI computations during the image capture process itself. Instead of relying on separate sensor and computational modules, this sensor utilizes automatic photodetection to map light intensity directly to electrical signals, facilitating immediate spectral analysis. Researchers demonstrated its capability by training the sensor to distinguish between birds and backgrounds in images, showcasing its potential for intelligent identification in diverse applications, from detecting oxide layer thicknesses in semiconductors to monitoring plant hydration states. This development paves the way for advanced optical sensing technologies that can integrate intelligence directly into hardware, expanding possibilities across multiple fields.