Camera-Based Measurement of Glucose Concentration in Aqueous Solution: Achieving High Precision and Accuracy with Frame Accumulation and Pixel Stacking

Abstract

Accurate measurement of glucose concentration in solution is imperative for diverse scientific applications. In this study, conducted at Tianjin University in 2024, the optical measurement accuracy of glucose in aqueous solution was substantially improved using a novel approach of pixel stacking and frame accumulation. Glucose solutions (n = 28) of different concentrations (range, 0.0–5.0%) were prepared, and the transmitted light intensity from these solutions was recorded using a polarization-sensitive camera. To improve the signal-to-noise ratio (SNR), we used frame accumulation and signal superposition of multiple pixels. The results showed a linear decrease in optical intensity captured by the camera with increasing glucose concentration in the solution. Following frame accumulation and pixel stacking, a substantial enhancement in precision and accuracy up to a factor of 1000 was achieved. Moreover, three algorithms including partial least squares (PLS) model, support vector machine (SVM) model, and random forest (RF) model were constructed and trained on one set of the samples and then used to blindly predict another set of the samples. The prediction of glucose concentration by the PLS model showed a minimum root mean square error (RMSE) of 0.2653, mean square error (MSE) of 0.0704, and a correlation coefficient of 0.9879 between the measured optical intensity and the glucose concentration. The use of frame accumulation and pixel stacking methods substantially improved glucose measurement accuracy.