The purity of hybrid rice seeds reflects the typical consistency of seed varieties in characteristics.The accuracy and reliability of seed purity detecting are of great significance to ensure the quality of seeds.In t...The purity of hybrid rice seeds reflects the typical consistency of seed varieties in characteristics.The accuracy and reliability of seed purity detecting are of great significance to ensure the quality of seeds.In this study,the feasibility of identifying the purity of hybrid rice seeds,Xinong 1A/89,by terahertz(THz)time-domain spectroscopy system combined with chemometrics was explored.Three quantitative identification models for testing the purity of Xinong 1A/89 hybrid rice seed were developed and compared by THz absorption spectroscopy with extreme learning algorithm(ELM),Principal cComponent Regression(PCR)and Partial Least Squares Regression(PLSR).Experimental results showed that comparing with classical PLSR and PCR models,ELM presents a better feasibility and stability.For the testing set,the quantitative prediction result of ELM(ELoo=2.005×10^(-5),R^(2)=96.75%)is significantly better than those of PCR(ELoo=7.346×10^(-5),R^(2)=88.10%)and PLSR(ELoo=8.007×10^(-5),R^(2)=87.03%).The results highlight the feasibility of THz spectroscopy combined with ELM as an efficient and reliable method for verification of hybrid rice seeds.展开更多
基金This work is supported by Application Development Programs of Chongqing Science and Technology Commission(Grant No.cstc2014yykfA80006)Fundamental Research Funds for the Central Universities(Grant No.SWU117029)National Natural Science Foundation of China(Grant No.61401373 and 31771670).
文摘The purity of hybrid rice seeds reflects the typical consistency of seed varieties in characteristics.The accuracy and reliability of seed purity detecting are of great significance to ensure the quality of seeds.In this study,the feasibility of identifying the purity of hybrid rice seeds,Xinong 1A/89,by terahertz(THz)time-domain spectroscopy system combined with chemometrics was explored.Three quantitative identification models for testing the purity of Xinong 1A/89 hybrid rice seed were developed and compared by THz absorption spectroscopy with extreme learning algorithm(ELM),Principal cComponent Regression(PCR)and Partial Least Squares Regression(PLSR).Experimental results showed that comparing with classical PLSR and PCR models,ELM presents a better feasibility and stability.For the testing set,the quantitative prediction result of ELM(ELoo=2.005×10^(-5),R^(2)=96.75%)is significantly better than those of PCR(ELoo=7.346×10^(-5),R^(2)=88.10%)and PLSR(ELoo=8.007×10^(-5),R^(2)=87.03%).The results highlight the feasibility of THz spectroscopy combined with ELM as an efficient and reliable method for verification of hybrid rice seeds.
