摘要
随着葵花籽发芽,其脂肪和蛋白质等营养物质的含量会减少,影响油脂产品的产量及品质。利用太赫兹时域光谱技术,分别结合支持向量机(Support vector machine,SVM)算法和极限学习机(Extreme learning machine,ELM)算法建立葵花籽发芽粒的定性识别模型,保障葵花籽的品质安全。实验过程首先对60粒葵花籽进行萌芽培养,之后从中随机选择30颗作为发芽粒样本,另选择30粒正常葵花籽样本,共计60粒实验样本。之后利用太赫兹衰减全反射技术采集0.3~3.6 THz范围内的实验样本光谱数据,经过快速傅里叶变换与加窗操作转换到频域中,通过提取光学常数得到样本的吸收系数光谱和折射率光谱,选取10~40 cm^(–1)和60~80 cm^(–1)2个特征波段的折射率进行归一化预处理,分别结合SVM算法和ELM算法建立定性识别模型。实验结果表明,基于联合特征波段-SVM算法的定性模型与基于联合波段-ELM建立的定性识别模型对发芽粒识别正确率、正常粒识别准确率均为100%。相比ELM模型,SVM模型识别速度更快。研究结果表明,利用太赫兹时域光谱技术分别结合SVM算法与ELM算法对葵花籽发芽粒进行识别检测均具有可行性,所建立模型可靠性很强,为贮藏、加工期间葵花籽发芽粒的监控监测提供方法参考。
With the germination of sunflower seeds,the content of nutrients such as fat and protein will decrease,affecting the yield and quality of oil products.Using terahertz time domain spectroscopy technology combined with support vector machine(SVM)algorithm and extreme learning machine(ELM)algorithm,the qualitative identification model of sunflower seed germination was established to ensure the quality and safety of sunflower seed.During the experiment,60 sunflower seeds were firstly incubated,and then 30 sunflower seeds were randomly selected as germination samples,and another 30 normal sunflower seeds were selected,totaling 60 experimental samples.Then,the spectral data of the experimental samples in the range of 0.3~3.6 THz were collected by terahertz attenuated total reflection technology,and converted to the frequency domain through fast Fourier transform and windowing operation.The absorption coefficient spectrum and refractive index spectrum of the samples were obtained by extracting optical constants.The refractive indexes of 10~40 cm^(–1)and 60~80 cm^(–1)feature bands were selected for normalized preprocessing,and the qualitative recognition models were established by SVM algorithm and ELM algorithm,respectively.The experimental results showed that the recognition accuracy of germinated seeds and normal seeds was 100%by the qualitative model based on the joint feature band-SVM algorithm and the qualitative recognition model based on the joint feature band-ELM.Compared with ELM model,SVM model has faster recognition speed.The results show that it is feasible to use terahertz time domain spectrum technology combined with SVM algorithm and ELM algorithm to identify sunflower seed germination seeds,and the established model has strong reliability,providing a method reference for monitoring sunflower seed germination seeds during storage and processing.
作者
田密
王少敏
孙晓荣
刘翠玲
吴静珠
TIAN Mi;WANG Shaomin;SUN Xiaorong;LIU Cuiling;WU Jingzhu(School of Artificial Intelligence,Beijing Technology and Business University,Beijing 100048,China;Beijing Key Laboratory of Big Data Technology for Food Safety,Beijing Technology and Business University,Beijing 100048,China;Market Supervision Administration of Shijingshan District,Beijing 100049,China)
出处
《食品科技》
CAS
北大核心
2022年第9期238-243,共6页
Food Science and Technology
基金
北京市自然科学基金项目(4222043)
2021年教育部高教司产学合作协同育人项目(202102341023)。
关键词
葵花籽
THz-ATR技术
支持向量机算法
极限学习机算法
定性识别
sunflower seed
THz-ATR technology
support vector machine algorithm
extreme learning machine algorithm
qualitative identification
