Smartphone-based pH titration for liquid food applications

Yuhui Xiao,Yaqiu Huang,Junhong Qiu,Honghao Cai,Hui Ni

DOI: https://doi.org/10.1007/s11696-024-03715-9

IF: 2.146

2024-09-29

Chemical Papers

Abstract:The pH detection helps control food quality, prevent spoilage, determine storage methods, and monitor additive levels. In the previous studies, colorimetric pH detection involved manual capture of target regions and classification of acid–base categories, leading to time-consuming processes. Additionally, some researchers relied solely on R*G*B* or H*S*V* to build regression models, potentially limiting their generalizability and robustness. To address the limitations, this study proposed a colorimetric method that combines pH paper, smartphone, computer vision, and machine learning for fast and precise pH detection. Advantages of the computer vision model YOLOv5 include its ability to quickly capture the target region of the pH paper and automatically categorize it as either acidic or basic. Subsequently, recursive feature elimination was applied to filter out irrelevant features from the R*G*B* , H*S*V* , L*a*b* , Gray, X R , X G , and X B . Finally, the support vector regression was used to develop the regression model for pH value prediction. YOLOv5 demonstrated exceptional performance with mean average precision of 0.995, classification accuracy of 100%, and detection time of 4.9 ms. The pH prediction model achieved a mean absolute error (MAE) of 0.023 for acidity and 0.061 for alkalinity, signifying a notable advancement compared to the MAE range of 0.03–0.46 observed in the previous studies. The proposed approach shows potential in improving the dependability and effectiveness of pH detection, specifically in resource-constrained scenarios.

chemistry, multidisciplinary