Non-Destructive Quality Monitoring of Shanxi Vinegar Production During The Fumigation Stage Using Computer Vision, Electronic Nose and Near-Infrared Spectroscopy Assisted by Machine Learning

Abstract

The aim of this study is to develop a non-destructive quality inspection method for the fumigation stages of Shanxi vinegar (SV). To overcome the limitations of high subjectivity in traditional sensory evaluation, artificial odor analysis, and visual detection in actual production, electronic nose (E-nose) and computer vision (CV) technologies were used, respectively. Additionally, near-infrared spectroscopy (NIRS) was employed to construct a detection model for key chemical compositions, addressing the time-consuming and destructive nature of conventional chemical analysis. The results showed that, compared to single-technique discrimination, the fusion of CV and E-nose data with the support vector machine classification (SVC) algorithm achieved 100 % accurate identification of different fumigation stages. For predicting key chemical compositions, the combination of NIRS and partial least squares regression (PLSR) yielded the best results, with correlation coefficients for moisture, total acidity, amino acid nitrogen, and reducing sugar being 0.9711, 0.9596, 0.8390, and 0.9515, respectively. This study provides an effective non-destructive method for quality detection and evaluation of SV during production and plays an important role in promoting the development of related industries.

Keywords: Aged vinegar, Production process monitoring, Intelligent sensing, Machine learning