太赫兹衰减全反射技术对板栗果仁霉变程度判别研究

Identification of moldy degree of Chinese chestnut kernel by terahertz attenuated total reflection technique

中文关键词:  板栗  太赫兹衰减全反射  遗传算法  粒子群算法  支持向量机算法  定性识别

英文关键词:Chinese chestnut  terahertz attenuated total reflection  genetic algorithm  particle swarm optimization  support vector machine algorithm  qualitative recognition

中文摘要:

      目的 建立基于太赫兹衰减全反射光谱法(terahertz attenuated total reflection spectroscopy, THz-ATR)快速检测板栗果仁霉变程度判别方法。方法 实验选取迁西板栗、沂蒙短枝、怀柔板栗3个品种的60颗饱满果仁进行霉变培养, 并依据GB/T 22346—2008《板栗质量等级》将板栗果仁分为正常、轻度霉变、重度霉变3类, 采集板栗果仁样本太赫兹时域光谱(波段0.3~3.6 THz)后进行光学常数提取, 从而得到样本的吸收系数谱图和折射率谱图, 并结合基于遗传算法(genetic algorithm, GA)寻优和基于粒子群算法(particle swarm optimization, PSO)寻优的支持向量机算法(support vector machine algorithm, SVM)建立定性判别模型。结果 PSO-SVM算法模型对板栗果仁霉变程度的预测集识别正确率为91.6667%, GA-SVM算法模型对板栗果仁霉变程度的预测集识别正确率为100%。结论 本研究所建立的定性判别模型准确率高, 利用太赫兹时域光谱技术可以实现对板栗果仁霉变程度的区分识别, 为太赫兹技术在食品检测领域的应用提供了现实基础。

英文摘要:

      Objective To establish a method for the rapid detection of moldy degree of chestnut kernel based on terahertz attenuated total reflection spectroscopy (THz-ATR). Methods Sixty plump kernels of Qianxi Chestnut, Yimeng short branch and Huairou District Chestnut were randomly selected for mouldy culture, according to GB/T 22346—2008 Quality grade of Chinese chestnut, the Chinese chestnut kernel was divided into 3 categories: Normal, mild moldy and severe moldy. Terahertz spectrum (band 0.3-3.6 THz) was collected and optical constants were extracted. The absorption coefficient spectrum and refractive index spectrum of the sample were obtained, and the qualitative discrimination model was established by combining the support vector machine algorithm (SVM) based on genetic algorithm (GA) and particle swarm optimization (PSO). Results The experimental results showed that the recognition accuracy of PSO-SVM algorithm model on the prediction set of chestnut kernel mildew degree was 91.6667%, and the recognition accuracy of GA-SVM algorithm model on the prediction set of chestnut kernel mildew degree was 100%. Conclusion The established model is accurate and terahertz time-domain spectroscopy can be used to distinguish the degree of moldy in chestnut kernel, which provides a theoretical basis for the application of terahertz technology in food detection.