摘要
采用激光诱导击穿光谱技术,结合BP神经网络技术对5种水稻种子进行了品种鉴别研究。讨论了两种用于水稻品种鉴别的方法,第一种是"特征谱法",从样品的全谱光谱图中选取Mg、Si、Ca、Na、K等5种元素的谱线构成特征谱,再将此特征谱输入BP神经网络对水稻种子进行识别;第二种是"分段特征谱法",将样品的全谱分为12段光谱,在每一段光谱中,利用自动选谱法选择一些峰构成特征谱,将其输入BP神经网络对水稻种子进行识别。实验结果表明:利用BP神经网络进行水稻种子的品种鉴别时,"分段特征谱法"比"特征谱法"更加适用,且前者的BP神经网络识别率最高可达100%。
The LIBS (Laser Induced-Breakdown Spectroscopy) combined with BPNN (Back Propagation Neural Network) was applied for identifying 5 varieties rice seeds. In this paper, two methods were used to identify rice seed varieties, one of them is "The Characteristic Spectral Method" : Selected 5 elements from the spectral lines to constitute the characteristic spectra, they are Mg, Si, Ca, Na and K. Then put the characteristic spectra into BPNN for identification. Anoter is "The Subsection Characteristic Spectral Method : The spectrum of sample can be divided into 12 segments, some peaks will choose from each of them to constitute the characteristic spectrum according to the Automatic Selection Method. Then 12 segments characteristic spectrum will be input to BPNN for identification. The experimental results shown that The Subsection Characteristic Spectral Method was more suitable for the identification of rice seed varieties by BPNN, and the identification rate of it was up to 100%.
出处
《激光杂志》
北大核心
2016年第9期56-60,共5页
Laser Journal
基金
荆州市科技发展计划(2015AB35)
长江大学大学生创新创业训练计划项目(20150100)
关键词
光谱学
激光诱导击穿光谱
BP神经网络
水稻品种鉴别
特征谱法
分段特征谱法
spectroscopy
laser induced breakdown spectroscopy
back propagation neural network
identification of rice seed varieties
characteristic spectral method
subsection characteristic spectral method