Application of machine learning for quantitative analysis of industrial fermentation using image processing

• Published in: Food science and biotechnology Vol. 34; no. 2; pp. 373 - 381
• Main Authors: Jeong, Jieun, Kim, Sangoh
• Format: Journal Article
• Language: English
• Published: Korea (South) Springer Nature B.V 01.01.2025; 한국식품과학회
• Subjects: Artificial neural networks; continuous fermentation; Fermentation; Fermented food; fermented foods; Food industry; Food processing; Gas pressure; glucose; Image processing; Machine learning; Neural networks; Quantitative analysis; Real time; Time measurement; Yeast; yeasts; 식품과학; AI; Fermentation quantification sensor; Vision system; Fermentation; Machine learning
• ISSN: 1226-7708; 2092-6456; 2092-6456
• DOI: 10.1007/s10068-024-01744-4

The Real-time Fermentation Quantification Sensor (RFQS) was developed to quantitatively assess fermentation by detecting airlock bubbles created by fermentation gas pressure. The Convolutional Neural Network-based Fermentation Measurement Model was integrated into the RFQS to analyze and classify these bubble images, enabling continuous fermentation monitoring and real-time fermentation degree measurement. Validation experiments revealed that varying the quantities of dry yeast and glucose significantly impacted fermentation duration and degree. Upon fermentation completion, the total degree was calculated using real-time data. These results confirmed that AI-based image processing technology can effectively serve as a quantitative measurement tool in the fermentation food industry.