A Machine Learning Method for the Fine-Grained Classification of Green Tea with Geographical Indication Using a MOS-Based Electronic Nose

Abstract

Chinese green tea is known for its health-functional properties. There are many green tea categories, which have sub-categories with geographical indications (GTSGI). Several high-quality GTSGI planted in specific areas are labeled as famous GTSGI (FGTSGI) and are expensive. However, the subtle differences between the categories complicate the fine-grained classification of the GTSGI. This study proposes a novel framework consisting of a convolutional neural network backbone (CNN backbone) and a support vector machine classifier (SVM classifier), namely, CNN-SVM for the classification of Maofeng green tea categories (six sub-categories) and Maojian green tea categories (six sub-categories) using electronic nose data. A multi-channel input matrix was constructed for the CNN backbone to extract deep features from different sensor signals. An SVM classifier was employed to improve the classification performance due to its high discrimination ability for small sample sizes. The effectiveness of this framework was verified by comparing it with four other machine learning models (SVM, CNN-Shi, CNN-SVM-Shi, and CNN). The proposed framework had the best performance for classifying the GTSGI and identifying the FGTSGI. The high accuracy and strong robustness of the CNN-SVM show its potential for the fine-grained classification of multiple highly similar teas.

A Machine Learning Method for the Quantitative Detection of Adulterated Meat Using a MOS-Based E-Nose

“…After the establishment of the baseline, the volatile gases from the meat samples were pumped into the sensor chamber following Arrow 1 in Figure 1 at a constant flow rate of 10 mL/s where they contacted the MOS sensors. In this manner, the gas molecules were absorbed on the sensors’ surfaces where they changed the sensors’ conductivity through redox reactions with the sensors’ active elements [ 23 ]. The sensors’ conductivities stabilized when they were saturated.…”

A Transfer Learning Framework with a One-Dimensional Deep Subdomain Adaptation Network for Bearing Fault Diagnosis under Different Working Conditions

“…A CNN is a multi-stage neural network composed of convolutional blocks and FC layers [26]. Traditionally, a convolutional block is composed of a convolution layer and a pooling layer [27], as shown in Figure 1a. In general, a batch normalization (BN) layer [28] is added after the convolution layer to improve the network training speed, prevent overfitting, and control gradient explosion and gradient disappearance.…”

Application of Electronic Nose to Predict the Optimum Fermentation Time for Low-Country Sri Lankan Tea

“…e linear kernel can be used as a normal dot product of any two given observations, and the polynomial kernel is a more generalized form of the linear kernel. e radial basis function (RBF) kernel is a popular one and is commonly used in SVM classification problems as it can map an input space in infinite-dimensional space [44].…”