Identification and detection of frozen-thawed muscle foods based on spectroscopy and machine learning: A review

Abstract

Background

The use of frozen and thawed muscle food labeled as fresh foods is one of the most common frauds and has attracted widespread attention from consumers, government regulators and retailers. The combination of spectroscopy and machine learning has revolutionized the detection of frozen-thawed muscle foods, making it possible to develop more complex and automated solutions.

Scope and approach

This paper comprehensively reviews the latest findings of various studies on the potential characteristics of spectroscopy in frozen-thawed muscle foods. In addition, this paper also discusses the contribution of machine learning in the process of spectral detection and identification of frozen-thawed muscle foods, such as feature engineering, model complexity and model evaluation. The ultimate goal of this review is to highlight the contribution of machine learning and its integration with spectral methods in the identification and detection of frozen-thawed muscle foods.

Key findings and conclusions

The combination of spectroscopic techniques and machine learning has successfully achieved the prediction of the quality of frozen-thawed muscle foods and the identification between fresh and frozen-thawed muscle foods. By diminishing reliance on manual feature engineering, machine learning can systematically analyze spectral features and refine models to accurately identify frozen-thawed muscle foods. Concurrently, deep learning and data augmentation techniques effectively tackle challenges related to data variability and quality. Furthermore, advanced technologies such as multimodal machine learning, lifelong learning, ensemble learning and reinforcement learning are expected to play a key role in the future.