基于THz-ATR技术的水稻品种快速无损识别研究

水稻的品质和产量与自身品种密切相关，本研究应用太赫兹衰减全反射式（THz-ATR）技术采集8个品种水稻种子0~359.97 cm-1光谱，并通过光学参数计算得到样本集的折射率谱和吸收谱。采用3点移动窗口平滑预处理和支持向量机分别优化及建立基于折射率和吸收系数的8个水稻品种识别模型，其中基于吸收系数的SVM模型预测准确率98.5 %，基于折射率谱的SVM模型预测准确率89 %；实验结果表明:将THz-ATR技术用于水稻品种识别具有初步可行性，有望为水稻品种快速鉴定提供参考。

Research on rapid and non-destructive identification of rice varieties based on THz-ATR technology

The quality and yield of rice are closely related to its own varieties. In this study, terahertz attenuated total reflection (THz-ATR) technology was used to collect the 0 ~ 359.97 cm-1 spectra of rice seeds of 8 varieties, and the refractive index spectrum and absorption spectrum of the sample set were calculated by optical parameters. Three point moving window smoothing preprocessing and support vector machine were used to optimize and establish eight rice variety recognition models based on refractive index and absorption coefficient respectively. The prediction accuracy of SVM model based on absorption coefficient was 98.5 %, and that of SVM model based on refractive index spectrum was 89 %; The experimental results show that it is preliminarily feasible to apply THz ATR technology to rice variety identification, which is expected to provide reference for rapid identification of rice varieties.