Deep learning assisted fluorescent gas sensing system based on MAPbBr3@PMMA nanofibers for camera detection of ammonia

In recent years, with the widespread deployment of surveillance cameras and the advancement of visual perception technologies, optoelectronic gas sensors, especially colorimetric and fluorometric sensors, have garnered growing attention in academic and industrial circles. Meanwhile, high-performance fluorescent sensing materials and advanced intelligent data processing algorithms are recognized as core factors driving the development of optoelectronic gas sensing technology. As a fluorescent material with high photoluminescence conversion efficiency and ambient sensitivity, metal halide perovskite quantum dots (MHPQDs) show remarkable potential for gas sensing if the stability and selectivity are effectively improved. In this study, we propose an ammonia gas sensing system integrating an economic CMOS camera and perovskite nanoparticle-based sensing units with high sensitivity, excellent stability, and remarkable prediction accuracy. The MAPbBr 3 @PMMA sensing probe exhibits outstanding luminescence stability in air up to 120 days. Meanwhile, a camera detection algorithm integrated with YOLO-based smart ROI targeting and polynomial ridge regression enables rapid and facile adaptation to repeatably responsive perovskite materials, achieving a Mean Absolute Percentage Error (MAPE) of only 2.69% for quantitative ammonia monitoring and a practical detection limit of 20 ppm NH 3 .