Study On The Variety Identification And Quality Detection Of Transgenic Soybean Based On Spectroscopy And Spectral Imaging

Genetically modified technique is increasingly applied in crops, due to the deterioration of the envierment, reduction of arable land and population growth. All over the world, China importes the most soybean production every year so that GM-soybeans are often traded into China through illegal ways. The conventional detection methods of GMO are time-consuming, non-professionals not to do and don't apply to real-time online rapid detection of non-GMO and GMO. My paper based on hyperspectral imaging technology and mid-infrared spectroscopy study on the variety identification and quality detection of transgenic soybeans, which are important to ensure quality, breeding culture and management. The main results are as follows:(1)Varieties identification of non-GMO and GMO soybeans were studied by Vis/NIR and NIR hyperspectral imaging techniques. After removing the noise and process of MA-7, a complete comparison was performed between full and sensitive spectra. Sensitive spectra was extracted by Bw, x-loading weights, PCA-loadings, SPA and CARS. Discriminant models are PLS-DA, BPNN, SVM and ELM. For non-GMO soybeans, the discriminant effects of HC6, JACK and TL1 in NIRS region, which are 100%,100% and 92.50% of identification rates, are better significantly than in Vis/NIRS. For GMO soybeans, based on full spectra, the discriminant results of BPNN in both Vis/NIR and NIR region are better and reached 99.12% and 98.67% in total identification rate. Based on sensitive wavelengths, the discriminant results of CARS's models in both region are better and superior slightly to SPA. But the number of SPA's sensitive wavelengths are fewer significantly than CARS. The total results performed hyperspectral imaging technique is feasible to study on variety identification of transgenetic soybean and NIRS performed better.(2)The protein content's prediction methods of GMO soybeans were established by mid-infrared spectroscopy and NIR hyperspectral imaging technique. After the removal of noise, PLS prediction model of GMO soybeans' protein content with mid-IR data by WT and NIR-HIS data by MA-7, which achieved better prediction results. In prediction results of NIRS region, Rp of HC6, JACK and TL1 are 0.7842,0.9198 and 0.9371, RMSEP are 0.6860,0.7240 and 0.6335, respectively, which are better than mid-infrared. Based on sensitive wavelengths selected by five methods, CARS-PLS in mid-infrared and SPA-PLS in NIRS region are better. But the prediction effects of NIRS region are better than mid-infrared totally. The total results indicated mid-infrared spectroscopy and NIRS hyperspectral imaging technique is feasible to predict protein content of transgenetic soybean and NIRS-HSI performed better.(3)The fat content's prediction methods of GMO soybeans were established by mid-infrared spectroscopy and NIR hyperspectral imaging technique. After the removal of noise, PLS prediction model of GMO soybeans'fat content with mid-IR data by WT and NIR-HIS data by MA-7, which achieved better prediction results. In prediction results of NIRS region, Rp of HC6, JACK and TL1 are 0.8178,0.9309 and 0.9452, RMSEP are 1.0072,1.0336 and 0.8960, respectively, which are better than mid-infrared. Based on sensitive wavelengths selected by five methods, CARS-PLS in mid-infrared and SPA-PLS in NIRS region are better. But the prediction effects of NIRS region are better than mid-infrared totally. The total results indicated mid-infrared spectroscopy and NIRS hyperspectral imaging technique is feasible to predict fat content of transgenetic soybean and NIRS-HSI performed better.