Deep learning-assisted portable fluorescence device for dynamic monitoring of mercury absorption in rice.

Rice is a major pathway for mercury exposure in inland China, making dynamic monitoring of Hg2+ absorption essential. In this study, nitrogen-doped carbon dots (N-CDs) were synthesized via a one-step hydrothermal method using melamine and ammonium citrate, achieving a quantum yield of 20.92%. Leveraging their high sensitivity to Hg2+, a fluorescent detection method with a detection limit (DL) of 85 nM was established. Building upon this, a portable device combining a 3D-printed device and smartphone was developed to capture fluorescence images under varying Hg2+ concentrations. A convolutional neural network extracted RGB and HSV features to quantitatively map Hg2+ levels (R2 = 0.998) across 1-50 μM, with a DL of 0.2 μM. The device showed excellent recovery (96.33%-113.67%) in tap water, river water and wastewater samples. Application to rice demonstrated enhanced Hg2+ absorption after mercury adaptation. This study provides a simple, effective strategy for field detection of Hg2+ in the environment.