基于近红外高光谱成像技术鉴别杂交稻品系

发布时间：2019-03-14 08:36

【摘要】：种子的筛选和鉴别是农业育种过程中的关键环节。该文基于近红外高光谱成像技术(874~1 734 nm)结合化学计量学方法以及图像处理技术实现杂交稻种的品系鉴别及可视化预测。采集了3类不同品系共2 700粒杂交水稻的高光谱图像,用SPXY算法,按照2∶1的比例划分建模集和预测集。基于水稻样本的光谱特征,采用主成分分析(PCA)方法初步探究3类样本的可分性。采用连续投影算法(SPA),提取出7个特征波长:985.08、1 106、1 203.55、1 399.04、1 463.19、1 601.81、1 645.82 nm。基于特征波长和全波段光谱,建立了偏最小二乘判别分析(PLS-DA)和支持向量机(SVM)模型。试验结果表明,所建模型判别效果较好,识别正确率均达到了90%以上,其中,SVM模型的判别效果优于PLS-DA模型,基于全谱的判别分析模型结果优于基于特征波长的判别模型。结合SPA-SVM校正模型和图像处理技术,生成样本预测伪彩图,可以直观的鉴别不同品系的水稻种子。结果表明,近红外高光谱成像技术可以实现杂交稻的品系识别及可视化预测,为农业育种过程中种子的快速筛选及鉴定提供了新思路。
[Abstract]:Seed selection and identification is the key link in the process of agricultural breeding. In this paper, the method of near infrared hyperspectral imaging (874x1734 nm) combined with chemometrics and image processing technique was used to realize line identification and visual prediction of hybrid rice varieties. The hyperspectral images of 2 700 hybrid rice varieties of three different lines were collected, and the model sets and prediction sets were divided according to the ratio of 2:1 by SPXY algorithm. Based on the spectral characteristics of rice samples, principal component analysis (PCA) was used to explore the separability of three types of samples. Seven characteristic wavelengths were extracted by continuous projection algorithm (SPA),: 985.08, 1,106, 1,203.5, 1,399.04, 1463.19, 1,601.81,1 645.82 nm. The partial least squares discriminant analysis (PLS-DA) and support vector machine (SVM) (SVM) model are established based on the characteristic wavelength and full-band spectrum. The experimental results show that the model has a good discriminant effect, and the correct rate of recognition is over 90%. Among them, the discriminant effect of SVM model is better than that of PLS-DA model. The discriminant analysis model based on full spectrum is better than the discriminant model based on characteristic wavelength. Combined with SPA-SVM correction model and image processing technology, sample prediction pseudo-color map can be generated, which can directly identify rice seeds of different lines. The results showed that the near infrared hyperspectral imaging technique could realize the line identification and visual prediction of hybrid rice, which provided a new idea for the rapid selection and identification of seeds in the process of agricultural breeding.
【作者单位】：浙江大学生物系统工程与食品科学学院;
【基金】：国家重大仪器设备开发专项(2014YQ470377) 国家十三五重点研发计划(2016YFD0200603)
【分类号】：S511;TP391.41

【相似文献】