Objective This paper aims to achieve automatic identification and classification of Chinese spirits, solve the production fluctuation of traditional "alcoholic strength determination based on foam watching" for Chinese spirits, and balance the accuracy, real-time performance, and universality of existing deep learning-based methods for foam classification of Chinese spirits.Methods An automatic foam classification method of Chinese spirit based on image acquisition optimization recognition was proposed. The foam images were collected through a self-built platform, and the data quality was improved by preprocessing via ENet. The foam images were classified by using the Vision Transformer (ViT) and ConvNeXt models.Results This method improved the automation level and accuracy of alcoholic strength determination for Chinese spirits and achieved a classification accuracy of 99.4% while ensuring real-time performance.Conclusion This method effectively optimizes the traditional alcoholic strength determination technology for Chinese spirits, enabling rapid and accurate real-time detection and classification of foams.