Quantitatively Detecting Camellia Oil Products Adulterated by Rice Bran Oil and Corn Oil Using Raman Spectroscopy: A...

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Quantitative detection of camellia oil is vital for quality control and authenticity.
• Adulteration with similar oils like rice bran oil and corn oil poses challenges due to spectral similarities.

Purpose of the Study:

• To evaluate the effectiveness of different feature extraction methods and regression algorithms for detecting camellia oil adulteration.
• To compare the performance of Back Propagation Neural Network (BPNN) with Random Forest (RF) and Partial Least Squares Regression (PLSR).

Main Methods:

• Utilized Independent Component Analysis (ICA) and Competitive Adaptive Reweighing Sampling (CARS) for spectral feature extraction.
• Developed regression models using BPNN, RF, and PLSR to predict oil concentrations.
• Tested models on camellia oil blended with rice bran oil and a mixture of rice bran and corn oils.

Main Results:

• For camellia oil adulterated with rice bran oil, ICA-BPNN and ICA-PLSR models showed satisfactory performance.
• When both rice bran and corn oils were present, BPNN models performed poorly.
• The combination of PLSR with CARS-ICA yielded the best prediction accuracy for complex adulteration scenarios.

Conclusions:

• The choice of feature extraction method and regression algorithm significantly impacts the accuracy of quantitative detection for camellia oil adulteration.
• Advanced methods like PLSR combined with CARS-ICA are essential for reliable detection in complex adulteration cases.