Real-time monitoring and quality evaluation of medium and high temperature Daqu fermentation process by computer vision technology

Daqu, as a traditional solid-state fermentation product, plays an important role in brewing, seasoning and other industries, and its fermentation process has an important impact on the flavor, quality, production efficiency and cost control of the product. Traditional monitoring methods rely on manual observation and laboratory detection, which are time-consuming and labor-consuming, and it is difficult to realize real-time monitoring and accurate evaluation. In this study, an image acquisition system for the fermentation process of medium and high temperature Daqu was designed and implemented. The industrial camera was used, combined with wide dynamic range technology and high temperature resistance to ensure high quality and real-time image acquisition. Then, through image preprocessing technology, including denoising, enhancement and normalization, the image quality is improved, which is convenient for subsequent feature extraction and recognition. In the aspect of feature extraction and selection, this study combined the characteristics of Daqu fermentation process, extracted key features such as color and texture, and used HSV color space and gray level co-occurrence matrix (GLCM) to quantify the features. Based on these characteristics, this study constructs a fermentation state identification and quality evaluation model that integrates deep learning technology and traditional machine learning algorithm. The model can automatically extract useful features from image data and make accurate classification and prediction. The experimental results show that the model designed in this study has achieved high accuracy and recall in real-time monitoring and quality evaluation of medium-high temperature Daqu fermentation process. The experimental results show that the model performs best in the middle stage of fermentation, but it also shows good performance in the early and late stage of fermentation. The successful application of this model not only brought technological innovation to Daqu production industry, but also laid a solid foundation for the wider application of computer vision technology in traditional solid-state fermentation industry.