Deep Learning-Enabled Image Classification for the Determination of Aluminum Ions

Abstract In this work, an image classification based on deep learning for quantitative field determination of aluminum ions (Al 3+ ) was developed. Carbon quantum dots with yellow fluorescence were synthesized by a one-pot hydrothermal method which could specifically recognize Al 3+ and produce enhanced green fluorescence. Using the convolutional neural network model in deep learning, an image classification was constructed to classify Al 3+ samples at different concentrations. Then, a fitting method for classification information was proposed for the first time which could convert discontinuous, semi-quantitative concentration classification information into continuous, quantitative, and accurate concentration information. Recoveries of 92.0–110.3% in the concentration range of 0.3–320 μM were obtained with a lower limit of detection of 0.3 μM, exhibiting excellent accuracy and sensitivity. It could be completed in 2 min simply without requiring large equipment. Thus, the deep learning-enabled image classification paves a new way for the determination of metal ions.