The Study On Rapid Measurement And Evaluation Technology Of Wheat Quality

Evaluation indexes of wheat grain quality mainly include water, protein, volume weight,hardness, falling number, unsound kernels,color and odour, color and smell which can beused for assessment of existing national standards, by regular observation or using simpleapparatus can be quickly learned that. In order to establish a set of scientific and reasonablefast detection method of wheat quality, using wheat grain as the experiment material, usingnear-infrared spectroscopy to establish forecasting model that moisture, protein, volumeweight, hardness, falling number, unsound kernels of wheat grain. Study the effect of differentderivative and scattering methods on calibration models, and separately usinglinear(PLS,Modified PLS,PCR) and nonlinear regression method of BP neural network tomodel calibrations, finally to evaluation the models. By comparing the chemical analysis andnear-infrared predictive value of indexes, testing reliability of near-infrared models. theresult were summarized as follows:(1)2441was the best derivative processing to establishment calibration model of water,protein, falling number, volume weight.1441was the best derivative processing toestablishment calibration model of hardness. SNV only was the best scattering processing toestablishment calibration model of water. SNV+Detrend was the best scattering processing toestablishment calibration model of protein, hardness.Inverse MSC was the best scatteringprocessing to establishment calibration model of falling number,volume weight.(2) BP neural network was the best regression to establishment calibration model ofwater, protein, volume weight, hardness. PLS was the best regression to establishmentcalibration model of falling number.(3) Nnder the conditions of the optimal model, respectively verify the predicted values ofwater, protein, volume weight, hardness, falling number, the correlation coefficients of Nearinfrared predictive values and the chemical analysis values R respectively were0.9893,0.9659,0.7870,0.9059,0.8925. The coefficient of determination R2respectively were0.9786,0.9329,0.6193,0.8207,0.7966， using F-test and T-test for the predicted results of theNear-infrared, an F-test results F value respectively were0.799(p=0.630>0.05),0.934(p=0.884>0.05),0.718(p=0.478>0.05),1.248(p=0.634>0.05),1.307(p=0.564>0.05); AnT-test results p value respectively were0.733>0.05,0.840>0.05,0.292>0.05,0.880>0.05, 0.249>0.05. Within5%the significance level, variance and mean of near infrared predictionvalues and chemical analysis values were not significant.(4) Application near-infrared to discrimination wheat grain. The results shows, imperfectkernels and unsound kernels were classified. Handling method select SNV+DETREND2411, the correct rate of the classify was100%, The results was best. Classification andidentification of adding different proportions of unsound kernels, handling method selectSNV+DETREND2411, the results was best, the correct rate of the classify was70%.From these results we can see that, applications of near infrared spectroscopy technologyto rapid measurement and evaluation of wheat quality is possible.