【作者】 吴海云； 舒韵涛； 卫勇； 杨仁杰； 单慧勇； 艾成龙；

【Author】 WU Hai-yun;SHU Yun-tao;WEI Yong;YANG Ren-jie;SHAN Hui-yong;AI Cheng-long;College of Engineering and Technology,Tianjin Agricultural University;Sinotech(Tianjin) Intelligent System Engineering Co.,Ltd.;

【机构】 天津农学院工程技术学院； 中科国技(天津)智能系统工程有限责任公司；

【摘要】 垩白是我国稻米外观品质中最重要的指标之一。高效、快速、准确的稻米单籽粒检测,可为水稻品质改良育种提供技术支持。提出并建立一种基于近红外光谱技术的单籽粒大米垩白的鉴别方法。从超市购置五常稻花香型大米,在实验室人工分离出垩白籽粒、非垩白籽粒各50粒,共100个样品。并在4 000～12 000 cm 1范围内采集所有样品的近红外漫反射光谱。对其光谱数据进行主成分（PCA）分析,结果发现PCA方法无法完全实现垩白籽粒的聚类鉴别。分别从第一主成分（PC-1）及第二主成分（PC-2）的载荷图中提取特征峰,并建立基于特征峰的偏最小二乘判别模型（PLS-DA）,结果发现,融合PC-1及PC-2的特征峰,即在波数为4 250,4 494,4 962,5 410,5 614,5 994,6 394,7 018及8 338 cm^-1处,建模结果最优,其对校正集和预测集进行判别,正确率均为100%。更多还原

【Abstract】 Chalk is one of the most important indexes of rice appearance quality in China.Efficient,rapid and accurate detection of single kernel chalky rice can provide technical support for rice quality improvement breeding.A new method based on Near Infrared Spectroscopy（NIRS） was proposed to identify single kernel chalky rice.Wuchang milled rice with flower flavor was purchased at supermarket.Fifty chalky kernels and fifty non-chalky kernels were manually separated in the laboratory.The near-infrared diffuse reflectance spectra of all samples were collected in the range of 4 000～12 000 cm^-1.Principal component analysis（PCA） was used to analyze the spectral data.It was found that PCA method could not be used to realize the cluster identification of chalky grains completely.Therefore,the feature peaks were extracted from the loading spectrum of the first principal component（PC-1） and the second principal component（PC-2） respectively.The partial least squares discriminant（PLS-DA） models based on the feature peaks were established.The results show that based on the combined peaks of PC-1 and PC-2,that were,at the wave numbers of 4 250,4 494,4 962,5 410,5 614,5 994,6 394,7 018 and 8 338 cm^-1,the correct classification rates of calibration set and predication set were both 100%.更多还原