| Pre-harvest Sprouting is a phenomenon that the grain germinate on the ears of wheat before harvest because of the raining weather and the highly humidity of air, which can seriously affects the yield and quality of wheat and then bring a threat to national food security, so observing the related factors which can influence the properties of wheat germination and achieving the fast detecting of wheat germination which can help fostering the varieties which can resist the pre-harvest sprouting is one of the important issue. However, the traditional detection methods have many disadvantages such as long cycle, heavy workload and complicated process, the near infrared spectroscopy technology which developed rapidly in recent years also makes poor stability results because of the heterogeneity of sample, so to establish an efficient, accurate and reliable system has become an urgent need for the current practice. Surface application of hyperspectral imaging technology with high resolution in both spatial and spectral providing possibility, the hyperspectral imaging can effectively obtain the spectrum of the object region and then make up the deficiency of point acquisition based on near infrared spectroscopy.In view of the above, experiment started from the factors of both varieties and years which have great influence in germination, taking both single and group grain of wheat as testing object, the near-infrared hyperspectral imaging technology, chemical pattern recognition and data analysis also have been used to analyze the differences of spectral of wheat with different varieties and years, thus to achieve the purpose of accurate identification in varieties and years. The end, to complete the observe by the forecasting of germination on single and group of wheat. The main research contents and results are as follows:(1) Difference between varieties of wheat was analyzed by the hyperspectral imaging technology. hyperspectral images of single seed of three types including six varieties has been obtained by the EVK spectrograph which comes from German, the characteristics of both image and spectral has been analyzed by imaging processing, pattern recognition and information fusion and used to be constructed the identification model, so did the fused information. The results of image model showed know that the classification accuracy between strong gluten wheat and weak gluten wheat could achieve 100%, it indicated that hyperspectral images could reflect the differences of varieties; the results of spectral model showed that the classification effect of endosperm is slightly better than the embryo, it demonstrated that the grain shape couldinfluence the classification accuracy; the fusion model show better performance than the image model and spectral model, the classification accuracy raise from 95.56% to 98.89%, it showed that digging the morphological and spectral characteristics of the hyperspectral image could effectively improve the classification effect.(2) The characteristic changes of single wheat seed during storage procedure were measured through hyperspectral imaging technology. Firstly, hyperspectral imaging data of wheat grain including six years which from year 2007 to 2012 had been obtained. The original spectral showed clear difference in the band from 1400 nm to 1600 nm, which may be caused by the decreasing of moisture and protein content during storage; Principal component analysis(PCA) was applied to analyze the spectral data of wheat grain including six years, the clustering chart of the principal components indicated that the grain between same or similar year have an clustering characteristic, and the characteristic would become obviously with the increasing of storage year; SIMCA was applied to classify the grain of different years, results showed that the classification accuracy of the dichotomy between adjacent years could reach 97.44%, and the accuracy of the mixed classification of six years could also reach 82.5%. These results declared that hyperspectral imaging technology could be used to recognize and divide the quality change of wheat seed during storage.(3) First started the forecasting of germination on single and group of wheat, the difference of spectral between sprouted grains and not sprouted has been analyzed, results showed that the spectral reflectance of sprouted grains was lower than not, LSSVM was applied to established the forecasting model, results show that the average currency could reach 72.5%, which declared that the forecasting of single based on hyperspectral imaging technology is useful. Then the average spectral of 145 group of wheat including eleven years from 2003 to2013 was obtained, so did the percentage of germination; the PLS model also has been constructed, it showed that the correlation coefficient of calibration set reaches 0.6657 and the prediction set reaches 0.6296, both of them were little lower, the error of related analysis could only reach 1.2614, which declared that the stability of model need to be verified. |
