Study on robust model construction method of multi-batch fruit online sorting by near-infrared spectroscopy
Yong Hao1, Yuanhang Lu1, Xiyan Li1
1School of Mechatronics and Vehicle Engineering, East China Jiaotong University, Nanchang 330013, China.
Summary
Robust regression improves online fruit analysis using near-infrared spectroscopy (NIRS). This method enhances prediction accuracy for soluble solids content (SSC) across different sample batches and collection times, outperforming traditional partial least squares regression (PLSR).
Area of Science:
- Agricultural Science
- Analytical Chemistry
- Spectroscopy
Background:
- Online fruit analysis using near-infrared spectroscopy (NIRS) faces challenges with sample and instrument variability.
- Predicting soluble solids content (SSC) across different batches and times is crucial for quality control.
Purpose of the Study:
- To develop a robust modeling method for accurate online SSC prediction in tomatoes.
- To address the variability issues affecting traditional NIRS models in multi-batch fruit analysis.
Main Methods:
- Utilized multivariate statistical process control (MSPC) for stability monitoring.
- Applied robustness regression (Rob-Reg) and partial least squares regression (PLSR) for mixed-batch modeling.
- Collected spectral data and SSC for 440 tomato samples across four batches.
Main Results:
- MSPC effectively monitored sample consistency across different batches and times.
- Robust regression (Rob-Reg) significantly improved SSC prediction compared to PLSR.
- Rob-Reg achieved a higher correlation coefficient (0.66 vs. 0.61) and lower RMSEP (0.44 vs. 0.55).
- An RPD of 3.85 indicated excellent model performance.
Conclusions:
- Robust modeling methods are well-suited for fruit NIRS online detection systems.
- Rob-Reg offers superior adaptability and prediction accuracy for variable sample and instrument conditions.
- The developed approach enhances the reliability of online NIRS for fruit quality assessment.
