Olive oil classification with Laser-induced fluorescence (LIF) spectra using 1-dimensional convolutional neural...

Area of Science:

• Spectroscopy
• Machine Learning
• Food Analysis

Background:

• Laser-induced fluorescence (LIF) spectroscopy is a key technique for olive oil analysis.
• Accurate and efficient classification of olive oil is crucial for quality control and authenticity verification.

Purpose of the Study:

• To develop and validate a novel 1-dimensional convolutional neural network (1D-CNN) with a dual convolution structure (Dual-conv) for classifying olive oil based on LIF spectra.
• To assess the performance of the proposed Dual-conv model against a standard 1D-CNN and a Support Vector Machine (SVM) approach.

Main Methods:

• Utilized a dataset of 72,000 LIF spectra from olive oil samples.
• Implemented a 1D-CNN model enhanced with a Dual-conv structure, designed for direct application to spectral data without pre-processing.
• Compared classification accuracy and convergence speed with a standard 1D-CNN and an SVM classifier.

Main Results:

• The Dual-conv model achieved a high classification accuracy of approximately 99.69% for olive oil.
• The Dual-conv model demonstrated superior performance, including faster convergence and better evaluation parameters, compared to the standard 1D-CNN.
• Neural network models, particularly the Dual-conv variant, outperformed the SVM in olive oil classification.

Conclusions:

• The proposed Dual-conv 1D-CNN model offers a highly accurate and efficient method for olive oil classification using LIF spectroscopy.
• This approach eliminates the need for data pre-processing, making it suitable for large-scale, real-world applications.
• The Dual-conv architecture represents a significant advancement in applying deep learning to spectral data analysis for food authentication.