茶叶夹杂昆虫异物THz光谱检测研究

中国红茶因其悠远的文化底蕴和多种保健功效而广受推崇。红茶经杀青、揉捻、晒干、筛选、干燥制成成品。但复杂的加工工艺也增加了物理污染的风险，在加工过程中难免会夹杂非茶类异物。物理污染属于随机事件，占食品安全投诉事件的19.8%。降低食品污染是未来造成厂家与消费者、进出口贸易双方争议的关键。昆虫等有机异物尚属于X射线不可检测范畴的食品异物，但昆虫异物混入频率较高，容易造成消费投诉，迫切需要开发相应的无损检测方法。太赫兹时域光谱（THz-TDS）技术作为一种新兴的检测技术，在农产品、食品以及药品无损检测方面均具有良好的应用潜力。THz具有良好的低能透射和指纹光谱特性、且无电离辐射损伤，能透过食品基质获取潜藏异物的光谱和图像信息，是农产品、食品无损检测的较佳选择。为了实现茶叶夹杂低密度有机异物的高效检测，本文基于THz光谱技术，探索红茶夹杂昆虫异物的无损检测新方法。在0.2～3.0 THz范围内，采集了红茶基质、昆虫异物以及夹杂昆虫异物红茶的THz光谱。分析了茶叶基质和昆虫异物的THz吸收系数和介电损耗响应特性，从频谱图可以看出茶叶基质与昆虫异物的THz吸收系数和介电损耗存在显著的差异，主要由昆虫异物的蛋白质和脂肪成分引起，这奠定了红茶夹杂昆虫异物的THz光谱检测基础。但茶叶与昆虫异物吸收系数均无明显的特征峰，且在2.0～3.0 THz高频波段内存在较为明显的噪声。采用主成分分析方法分别对吸收系数和介电损耗进行降维处理，通过得分图可以判断出红茶基质与夹杂昆虫异物的红茶之间存在明显的区别，且吸收系数的聚类效果优于介电损耗系数。分别选取0.5～1.0 THz范围的THz吸收系数和介电损耗为输入向量，建立了支持向量机（SVM）和线性判别分析（LDA）判别模型。实验结果表明，基于THz吸收系数的LDA判别模型精度最高，新样品的正确识别率为73.68%。这说明应用THz时域光谱无损检测红茶夹杂的昆虫异物具有可行性，同时THz光谱结合模式识别算法为茶叶夹杂昆虫异物无损检测提供了新方法，也可为其他农产品和食品检测提供参考。

THz Spectroscopy Detection of Insect Foreign Body Hidden in Tea Products

Chinese black tea is widely respected for its long-term cultural heritage and multiple health benefits. Black tea uses fresh leaves of tea trees as raw materials and is finished by finishing, kneading, drying, screening, and drying. These traditional techniques create a mellow, varied, and rich taste of the famous tea. However, the complex processing technology also increases the risk of physical pollution, and non-tea and foreign tea matter will inevitably be mixed during the processing. Physical contamination is a random event, accounting for 19.8% of food safety complaints. Physical pollution can only be reduced, and it is difficult to eliminate it. It is the key to future disputes between manufacturers and consumers and import and export trade. X-rays use the density difference between the food matrix and foreign objects to detect foreign metal objects effectively and extend to high-density plastics. Organic foreign bodies such as insects are still food foreign bodies that X-rays cannot detect. Insect foreign bodies are mixed with a high frequency, which causes sensory discomfort and easily introduces pathogenic bacteria. There is an urgent need to develop corresponding non-destructive testing methods. As an emerging detection technology, Terahertz time-domain spectroscopy (THz-TDS) technology has good application potential in non-destructive testing of agricultural products, food and medicines. THz has good low-energy transmission and fingerprint spectrum characteristics and has no ionizing radiation damage. It can obtain the spectrum and image information of hidden foreign objects through the food matrix. It is a better choice for non-destructive testing of agricultural products and food. In order to realize the high-efficiency detection of low-density organic foreign matter in tea, this paper explores a new method of the non-destructive detection of black tea-infested insect foreign matter based on THz spectroscopy. In the range of 0.2~3.0 THz, the THz spectra of black tea matrix, insect foreign matter, and black tea mixed with foreign insect matter were collected. The THz absorption coefficient and dielectric loss response characteristics of the tea matrix and insect foreign bodies are analyzed. From the spectrogram, it can be seen that there are significant differences between the THz absorption coefficient and dielectric loss of the tea matrix and the foreign insect bodies, mainly caused by the protein and fat of the foreign insect bodies. The ingredients are caused, laying the foundation for the THz spectrum detection of black tea mixed with insects and foreign bodies. However, the absorption coefficients of tea and insects have no obvious characteristic peaks, and there is more obvious noise in the high-frequency band of 2.0~3.0 THz. The principal component analysis method is used to reduce the dimensionality of the absorption coefficient and the dielectric loss. The score map shows that there is a clear difference between the black tea matrix and the black tea with insect foreign bodies, and the clustering effect of the absorption coefficient is better than that of the dielectric loss factor. The THz absorption coefficient and dielectric loss in the range of 0.5~1.0 THz were selected as input vectors, and support vector machine (SVM) and linear discriminant analysis (LDA) discriminant models were established. The experimental results show that the LDA discrimination model based on the THz absorption coefficient has the highest accuracy, and the correct recognition rate of new samples is 73.68%. It shows that applying THz time-domain spectroscopy for non-destructive detection of black tea inclusions of foreign insect bodies is feasible. At the same time, THz spectrum combined with pattern recognition algorithm provides a new method for non-destructive detection of tea inclusions of foreign insect bodies and can also provide a reference for other agricultural products and food detection.