Abstract:

Animal-derived foods are nutrient-rich and have a high biological value, but they are highly susceptible to contamination by foodborne pathogens. Traditional freshness detection methods are destructive, time-consuming, require skilled personnel to operate, and involve complex sample preparation procedures. In contrast, new rapid and non-destructive detection technologies exhibit high sensitivity, short response time, ease of operation, and non-invasiveness, although their data accuracy still needs improvement. In recent years, rapid advancements in deep learning have provided more accurate identification results for food freshness detection and enabled high-throughput detection. This paper systematically analyzes the application of deep learning in the rapid and non-destructive detection for freshness of animal-derived foods, including commonly used techniques for rapid and non-destructive detection of freshness and their drawbacks, the application of deep learning in the rapid and non-destructive detection of major animal-derived foods such as eggs, livestock and poultry meat, and aquatic products as well as the adaptability and improvement effects of different deep learning models on these technologies. It also summarizes and prospects the limitations and development directions of deep learning, aiming to provide a reference for further research and application of deep learning in the rapid and non-destructive detection for freshness of animal-derived foods.

Key words: deep learning, animal-derived foods, rapid non-destructive detection, spectroscopy, computer vision