Image-Based Glucose Concentration Detection in Liquids Using Refractometer and Grayscale-RGB Processing

Abstract

Excessive sugar consumption is a major driver of the global diabetes epidemic, underscoring the need for low-cost, accurate, and non-invasive sugar detection technologies. This study introduces an image-processing-based prototype that integrates a handheld analog refractometer with a digital microscope camera and Python-based processing to estimate glucose concentrations in liquid samples. Refractometer scale images were captured and analyzed using both grayscale and RGB transformations to enhance boundary clarity and interpretability. The methodology was validated on 30 laboratory-prepared glucose solutions and 15 commercial beverages. Laboratory results showed that raw refractometer readings systematically overestimated glucose mass fractions by ~4 percentage points, but regression-based calibration reduced error to below 1 percentage point (MAE = 0.77 pp, RMSE = 1.02 pp, R²=0.93.

Grayscale consistently provided sharper boundary detection compared to individual RGB channels, confirming its robustness as the preferred preprocessing mode. Commercial beverage testing revealed residual discrepancies (~3 pp on average) relative to label-derived sugar values, attributed to non-sugar solutes influencing refractive index. The proposed prototype demonstrates strong potential for semiautomated glucose quantification in low-resource environments. While not intended for clinical diagnostics, it provides a portable and reproducible tool for food safety, nutrition monitoring, and public health applications, with future extensions toward mobile integration and real-time quality control.