基于光致发光光谱的水稻品种快速鉴别研究

Discrimination of Varieties of Paddy Based on Photoluminescence Spectroscopy Combined with Chemometrics

水稻是中国最重要的粮食作物之一,其种子品种鉴别是目前农业生产、作物育种和种子检验的重要问题。由于品种鉴别比较困难,每年因品种搞错和纯度差而造成的经济损失巨大。为此提出了一种应用光致发光光谱分析技术与化学计量学相结合的快速、无损鉴别稻谷品种的新方法。为了实现水稻品种的快速无损鉴别,采用北京卓立汉光仪器有限公司生产的光致发光光谱分析仪,对从长江大学农学院收集的5个不同品种的水稻共125个样本进行650～1000nm波段光致发光光谱采集,获取了每个品种水稻各25个样本数据。将光谱数据经主成分分析压缩成数目较少的新变量(主成分),其中前7个主成分能够解释99.696%的原始光谱信息。因此将前7个主成分作为BP神经网络的输入,不同水稻的品种值作为BP神经网络的输出,建立水稻品种的模式识别模型。样本被随机的分成包含100个样本的建模集和25个样本的预测集。结果表明,5次随机抽样建模的预测正确率均达到100%。同时也表明,主成分分析的数据压缩能力好,通过主成分分析和BP神经网络相结合建立模型进行不同品种水稻鉴别具有分析速度快、鉴别能力强的特点,为快速鉴别水稻品种提供了新的方法。

Rice is one of the most important food crops. Identification of rice seed variety is currently in agricultural production, crop breeding and seed testing of important issues. As species identification difficult, each of the species made a mistake and poor purity causing huge economic losses. This paper presented an application of photoluminescence spectroscopy combined with chemometrics rapid, non-destructive identification of a new method of rice varieties. In order to achieve the rapid discrimination of the varieties of rice seed, University of Agriculture collected from the Yangtze to 5 different varieties of rice were 125 samples of 650-1000 nm band photoluminescence spectra collected, for each species with 25 samples of rice data by Beijing Optical Instrument Co. Ltd. production Zhuoli Han photoluminescence spectrometer. The spectral data compressed by principal component analysis a smaller number of new variables （principal component）, of which the first seven principal components could explain 99.892% of the original spectral information. Therefore the first 7 principal components as BP neural network input, the value of different rice varieties as the output of BP neural networks, pattern recognition model for the establishment of rice varieties. Sampleswere randomly divided into model contains the set of 100 samples and 25 samples of the prediction set. The results showed that 5 random model prediction 100% correct rate, and principal component analysis has good capability of data compression. It indicated that the model set up by the combination of principal component analysis and BP neural network in the present study was rapid in analysis and precise in pattern discrimination. It concluded that a new approach to distinguishing rice seed was offered.