Autoencoder-CatBoost Model for Accurate Hyperspectral Quality Assessment of Yunling Snowflake Beef

Snowflake beef is highly valued for its nutritional benefits and exquisite taste, yet the inconsistent quality in the market poses challenges for consumers in distinguishing genuine products, leading to economic losses and trust issues. This study aims to develop a rapid, accurate, and non-destructive method for evaluating the quality of snowflake beef to enhance consumer satisfaction and confidence. We propose a novel Autoencoder-Catboost model based on an autoencoder and CatBoost classifier, utilizing hyperspectral imaging (HSI) technology to assess the quality grades of Yunling Cattle snowflake beef. A total of 250 beef samples, scanned in the 900-2500 nm wavelength range, were processed using seven data preprocessing techniques and two feature extraction methods. The autoencoder, comprising three layers of Transformer units, effectively extracted features from the hyperspectral data, which were then classified by the CatBoost classifier. The model demonstrated superior accuracy, precision, recall, and F1-score compared to traditional machine learning methods, achieving an average accuracy of 82.42%. This research introduces an innovative approach by integrating Transformer-based autoencoders with CatBoost for hyperspectral data analysis, providing a new methodology for non-destructive snowflake beef quality evaluation and offering valuable insights for future research in food quality assessment.