Advanced Human-Centered Food Quality Assessment: Second Derivative Linear Prediction of Raw Broiler Shear Force Using Near-Infrared Spectroscopy

Abstract

This study investigates the application of linear predictive models and a low-cost Near-Infrared Spectroscopy (NIRS) system for non-invasive texture measurement of raw broiler meat, emphasizing its impact on food safety, consumer health, and industrial efficiency. Traditional meat quality assessment methods are often destructive, time-consuming, and labor-intensive, highlighting the need for more efficient, non-invasive alternatives. This research evaluated the effectiveness of Principal Component Regression (PCR) and Partial Least Squares (PLS) in predicting meat texture, demonstrating that the PLS outperformed the PCR with fewer latent variables and higher predictive accuracy. The results showed that the prediction accuracy of the PLS model for shear force estimation reached 64.01% for breast meat and 64.94% for drumstick samples, surpassing the performance of the PCR model. The application of second-order Savitzky-Golay derivative filtering and optimized spectral pre-processing further enhanced the model performance. By eliminating the need for invasive testing, this approach advances smart food technology, providing a fast, cost-effective, and non-destructive solution for automated meat quality assessment. These findings contribute to the development of AI-driven, real-time monitoring systems, improving food processing, supply chain efficiency, and ultimately ensuring better food quality and safety for consumers.