Determination of the Geographical Origin of Coffee Beans Using Terahertz Spectroscopy Combined With Machine Learning Methods

Abstract

Different geographical origins can lead to great variance in coffee quality, taste, and commercial value. Hence, controlling the authenticity of the origin of coffee beans is of great importance for producers and consumers worldwide. In this study, terahertz (THz) spectroscopy, combined with machine learning methods, was investigated as a fast and non-destructive method to classify the geographic origin of coffee beans, comparing it with the popular machine learning methods, including convolutional neural network (CNN), linear discriminant analysis (LDA), and support vector machine (SVM) to obtain the best model. The curse of dimensionality will cause some classification methods which are struggling to train effective models. Thus, principal component analysis (PCA) and genetic algorithm (GA) were applied for LDA and SVM to create a smaller set of features. The first nine principal components (PCs) with an accumulative contribution rate of 99.9% extracted by PCA and 21 variables selected by GA were the inputs of LDA and SVM models. The results demonstrate that the excellent classification (accuracy was 90% in a prediction set) could be achieved using a CNN method. The results also indicate variable selecting as an important step to create an accurate and robust discrimination model. The performances of LDA and SVM algorithms could be improved with spectral features extracted by PCA and GA. The GA-SVM has achieved 75% accuracy in a prediction set, while the SVM and PCA-SVM have achieved 50 and 65% accuracy, respectively. These results demonstrate that THz spectroscopy, together with machine learning methods, is an effective and satisfactory approach for classifying geographical origins of coffee beans, suggesting the techniques to tap the potential application of deep learning in the authenticity of agricultural products while expanding the application of THz spectroscopy.

Probing the Metabolic Landscape of Plant Vascular Bundles by Infrared Fingerprint Analysis, Imaging and Mass Spectrometry

“…Similar to fingerprint databases for forensic purposes, characteristic features are compared between a sample and a database, allowing for a correlative likelihood to be estimated insofar as the sample characteristics and database entry are identical. Currently, Fourier transform infrared (FTIR) spectroscopy has evolved into a valuable analytic tool with applications in various fields of science, including medicine [ 1 , 2 ], agriculture, food, and plant sciences [ 3 , 4 ]. Harnessing FTIR spectroscopy for the study of metabolite distribution inside plant tissues has opened new perspectives for high-resolution investigations into the vascular system of plants [ 5 , 6 , 7 ].…”

To Estimate Performance of Artificial Neural Network Model Based on Terahertz Spectrum: Gelatin Identification as an Example

“…Thus, we do not concern about extra scattering effect from different size and different distribution of two grains in a tablet. The THz curve in gelatin studies is representative of several foods that do not show peaks because of complex chemical constituents ( 16 , 32 ). The reason for optimizing an ANN with complex structure in similar studies is to balance fit and overfit.…”

Identification of Geographical Origin of Chinese Chestnuts Using Hyperspectral Imaging with 1D-CNN Algorithm

“…Since the network cannot recognize labels of type 0, 1, and 2, category labels need to be converted into one-hot encoding to store the n-bit state in the form of 0,1 in the n-bit register. Based on the initial size of the input dimension, we used a smaller convolution kernel and a deeper network [50]. Different from traditional CNN, the input of 1D-CNN was one-dimensional, and, accordingly, its convolution layer and pooling layer were also one-dimensional.…”

“…After the feature selection steps, the selected wavelengths were imported into traditional machine learning algorithms. As shown in Yang et al [50], the 1D-CNN model based on full spectra was also listed for easy comparison. More detailed classification results are provided in Table 7.…”