Multi-indicator adaptive prediction model for mutton based on hyperspectral imaging technology

This study proposes a multi-indicator adaptive synchronous prediction model for mutton that is based on hyperspectral imaging. By integrating spectral features, texture features, wavelet features, and visible light region color features (L∗, a∗, b∗ specific), a multidimensional information space is constructed. An adaptive optimization strategy is integrated and adopted, including adaptive feature selection, adaptive model selection, and adaptive model hyperparameter optimization. Among these, the prediction results for pH and TVC were the best, with Rc2 values of 0.953 and 0.973, respectively, and Rp2 values of 0.937 and 0.951, respectively; the Rc2 value for TVB-N was 0.896, and the Rp2 value was 0.856; and the Rc2 values for L∗, a∗, and b∗ were 0.872, 0.837, and 0.938, respectively, with Rp2 values of 0.498, 0.771, and 0.888, respectively. This study provides valuable references and technical support for the application of spectral multi-indicator simultaneous prediction, benefiting related research and practical fields.