Nondestructive Detection Of Transgenic Rice Based On Spectroscopy

The feasibility of using Raman technology, near-infrared technology and UV/VIS/N1R technology to discriminate transgenic rice "Huahui-1" from their parents was investigated, and Nexus FT-NIR spectrometer accompanied with Si detector was applied to determine the chlorophyll content in rice leaves. Besides, a dynamic detection platform was designed based on instrument theory and laboratory condition.The main results and conclusions were summarized as followings:Firstly, the influence of different analysis techniques was compared, Raman technology and near infrared technology were employed to discriminate transgenic rice from their parents. The results indicated that their overall classification accuracy could reach 100%. However, Raman spectrum's acquisition time is longer than that of NIR's and the Raman spectrum of rice seed would be seriously interfered by fluorescence.Secondly, Zeiss spectrometer and Nexus FT-NIR spectrometer (InGaAs) were employed to discriminate transgenic rice seeds from their parents, and the influence of different spectrometers with different wavelengths was compared. The results indicated that the performance of Nexus models was better than that of Zeiss ones except some results of discriminant analysis, what's more, an overall classification accuracy of 100% was achieved in all Nexus PLSDA models.Thirdly, Nexus FT-NIR spectrometer with InGaAs detector was employed to discriminate transgenic rice seeds from their parents in different years. The results indicated that their overall accuracy of discrimination could reach 100%, and the performance of nonlinear pattern recognition models was better than those of linear ones.Fourthly, the performance of Nexus FT-NIR spectrometer with different detectors (InGaAs and Si) was investigated, and both of them could reach a discrimination accuracy of 95%. For the quantification models, performance was improved after orthogonal signal correction (OSC), in the meantime, the modeling time was reduced effectively and the performance was improved using the effective wavelengths selected by successive projections algorithm (SPA). The optimal model was obtained by SPA-PLS-SVM combined with OSC. Its correlation coefficient (r) and root mean square error of prediction (RMSEP) was 0.9022 and 1.3121 respectively.Finally, an on-line system was established after the design of optical system, mechanical system, electrical system and the optimization of the collecting software. The intermittent way was adopted for the on-line system testing. After comparison, we found that 94.48% recognition ration could be achieved by PLSDA after SNV preprocessing, which means that our on-line system could be used to differentiate transgenic rice seeds from their parents. This made a well foundation to develop our equipment with independent intellectual property rights for fast nondestructive discrimination of transgenic rice seeds from their parents.