Study On Separation Of Wheat Scab By Photoelectric Separation Technology

Wheat fusarium head blight (FHB) brings a decrease in food production andproducts mycotoxin which is dangerous to human health. So, during the processing,study on separation of wheat scab by photoelectric separation technology hassignificant importance to the food security.This dissertation was written on the background of FHB, and the samples were gotfrom Huang huai and Jiang huai area. They were harvested in summer,2012, morethan20copies. Spectra of wheat collected by Near Infrared Spectroscopy(NIRS),and the Smoothing/MSC/Derivation/Standardized, and Chemicalmetrology was used to the spectral preprocessing. Then the pattern identificationmodel of NIR was established for the multi-kernel, ten-kernel and single wheat scab.Finally, according to the models, we designed the equipment model of photoelectricsorting equipment.Research results were as follows:(a) A model was developed to discriminate infection rates of FHB by means of nearinfrared spectroscopy. First, we got the model threshold of0.96%based on therelationship between the incidence of FHB and the content of DON; Within thesensitive wavelength range, The prediction accuracy of the PLS model was86.49%.(b) For the experiment of multi-kernel sample, The SIMCA model was establishedwith the spectral information of985nm,1130nm,1160nm,1190nm,1235nm,1320nm,1385nm, and1410nm. The identification rates and rejection rates of theunknown samples are all100%. With the threshold of±0.2,the PLS-DA modelwas established, with the spectral information of any point of1200nm,1430nm-1585nm, and the model prediction accuracy was100%. The LDA modelwas established with the spectral information of any point of1000nm,1200nm,1430nm-1585nm, with the model back substitution rate and discrimination accuracy are all100%.(c) For the experiment of ten-kernel sample, the SIMCA model was establishedbased on the spectral information of985nm,1130nm,1200nm,1385nm, and1415nm, with the model identification rates and rejection rates of the unknownsamples are all over85%. With the threshold of±0.2, the PLS-DA model wasestablished with the spectral information of any point of1200nm,1430nm-1650nm, and the model prediction accuracy was over80%. The LDAmodel was established with the spectral information of any point of1000nm,1200nm,1420nm-1600nm, with the model back substitution rate anddiscrimination accuracy are all over85%;(d) For the experiment of single-kernel sample, the SIMCA model was established inthe wave number range of950nm-1650nm. The recognition rate and therejection rate of diseased kernel are96.52%and85.90%, and those of ordinarykernel are96.43%and78.34%. With the threshold of±0.5,the PLS-DA modelwas established with the spectral information of1140nm-1165nm and1330nm-1415nm. The model prediction accuracy of unknown samples is97.62%.The LDA model was established with the spectral information of1000nm,1200nm,1420nm-1600nm, with the model back substitution rate is96.82%anddiscrimination accuracy96.38%.(e) Based on the building of discriminate model aforementioned, the equipmentmodel of the diseased kernel infection degree and the discriminate-sortingsystem of scab kernels were proposed.The NIR pattern recognition models of the multi-kernel, ten-kernel and single–kernel wheat were built based on the near-infrared spectroscopy (NIR).Furthermore,the photoelectric sorting equipment model of FHB wheat was designed.A theory and technology reference for the separation of FHB was offered in thispaper.