Broadening the Scope of Food Quality Control: a Calibration Transfer Strategy for Minimizing External Influences on Spectral Analysis
Penghui Liu,Yihang Qian,Changqing An,Yang Yao,Yingjie Zheng,Huirong Xu,Lijuan Xie
DOI: https://doi.org/10.1016/j.foodcont.2024.110899
IF: 6
2025-01-01
Food Control
Abstract:Calibration transfer (CT) is an effective tool to address spectral inconsistencies and the decline in model prediction capabilities caused by various influences. This study introduced a modified semi-supervised parameterfree calibration enhancement (MSS-PFCE) strategy to enable calibration models extensively applicable across varying temperatures, instruments and orientations. Investigations conducted on three datasets of watermelon juice, corn, and apples were utilized to validate the efficacy of MSS-PFCE. The results showed that MSS-PFCE decreased the average RMSEP for slave spectrum predictions by 96.51%, 75.96%, and 23.04%, respectively. MSS-PFCE significantly outperformed the original SS-PFCE approach and three classical standardization methods (direct standardization (DS), piecewise direct standardization (PDS) and slope and bias correction (SBC)). Furthermore, MSS-PFCE required no standards and needed only 5%-10% of slave samples for CT, maintaining robust reliability across varying sample sizes and cost thresholds (cth). The conclusions were further examined in three additional datasets with these variations. Notably, MSS-PFCE exhibited high accuracy, fitting stability, low sample dependency, and versatile transferability, positioning it as a promising candidate for widespread implementation in visible/near-infrared (Vis/NIR) spectroscopy. This study has significant potential to enhance the applicability of quality analysis, broaden the scope of food quality control, and promote the sound development of the food industry.