Rapid identification of foodborne pathogenic bacteria using hyperspectral imaging combined with convolutional neural networks(高光谱结合卷积神经网络对食源性致病菌的快速识别)
This paper presents a new classification model combined with hyperspectral imaging for rapid identification of foodborne pathogens. It adopts hyperspectral analysis to detect Shigella, Salmonella, Clostridium perfringens, and Streptococcus suis, and collects hyperspectral data. The region of interes...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | Chinese |
| Published: |
Zhejiang University Press
2025-07-01
|
| Series: | Zhejiang Daxue xuebao. Lixue ban |
| Online Access: | https://doi.org/10.3785/j.issn.1008-9497.2025.04.008 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper presents a new classification model combined with hyperspectral imaging for rapid identification of foodborne pathogens. It adopts hyperspectral analysis to detect Shigella, Salmonella, Clostridium perfringens, and Streptococcus suis, and collects hyperspectral data. The region of interest in the hyperspectral image using MATLAB software is selected and the average reflectance within that region is calculated. PCA dimensionality reduction on one-dimensional spectral data and hyperspectral images are then preformed, a 1D-CNN classification model for spectral data using Python software, and a 2D-CNN classification model for hyperspectral images are established. We build a feature fusion convolutional neural network based on spectra and images. At the same time, establish random forest (RF), K-nearest neighbor (KNN), and support vector machine (SVM) classification models for the reduced spectral data. The performance of the model using Precision, Recall, and F1-score metrics are evaluated. Through the analysis of the 1D-CNN, 2D-CNN, and feature fusion neural network classification models established on one-dimensional spectral data and hyperspectral images, it is show that the accuracy of the three models was 89.0%, 71.6%, and 93.3%, respectively. The feature fusion CNN model based on spectra and images can achieve good classification of the four strains, which outperforms the other three traditional machine learning algorithm models. In summarg, the combination of hyperspectral and convolutional neural networks is simple and easy to operate, and the established feature fusion neural network classification model can achieve rapid classification of foodborne pathogens.利用高光谱成像技术，采集了志贺氏菌、沙门氏菌、产气荚膜梭菌以及猪链球菌的高光谱图像和光谱数据。用MATLAB软件选取高光谱图像中的感兴趣区域（ROI），计算了该区域的平均反射率。通过Python软件采用主成分分析（PCA）方法分别对一维光谱数据和高光谱图像进行降维，构建了一维卷积神经网络（one dimensional convolutional neural network，1D-CNN）模型和二维卷积神经网络（2D-CNN）模型，将二者相结合，提出了一种快速识别食源性致病菌的特征融合CNN模型，同时对降维后的光谱数据建立随机森林（RF）、K-最近邻（KNN）及支持向量机（SVM）模型，用Precision，Recall及F1-score指标对模型性能进行评价。结果表明，1D-CNN，2D-CNN和特征融合CNN模型的分类准确率分别为89.0%，71.6%和93.3%，且特征融合CNN模型优于其他3种传统机器学习算法模型。将高光谱与CNN相结合的特征融合CNN模型可对食源性致病菌进行快速分类。 |
|---|---|
| ISSN: | 1008-9497 |