摘要
选取db3小波函数,对分属7个门32个属的43种我国近海常见的优势藻种和赤潮藻种的3-D荧光光谱进行分解,提取小波分解后得到的第2,3层尺度分量作为荧光特征谱。再通过系统聚类的方法分别得到门类和属类层次上的标准谱。在此基础上,利用非负最小二乘法解析的多元线性回归方法建立了浮游植物荧光识别测定的技术。结果表明,利用db3小波函数分解并提取荧光光谱特征,从而在门类水平建立起来的标准谱库,对于单种藻样品的识别正确率平均为96.33%,对混合样品的识别正确率平均为84.27%。在属类水平上建立起来的标准谱库,对于单种藻样品的识别正确率平均为92.53%,对混合样品的识别正确率平均为69.60%。
The wavelet dh3 was used to decompose the 3-D fluorescence spectrum of 43 typical harmful algae bloom （HAB） species belong to 7 divisions 32 genera, and the 2nd and 3nd scale vectors were taken as characteristic spectrum. The standard fluorescence spectrum at division and genus level were obtained re- spectively by cluster analysis of these characteristic spectra. Based on this, a fluorimetric discriminating method for phytoplankton was established by multiple linear regression solved by the nonnegative least squares. Results indicated that the recognition methods based on the wavelet function of db3, the correct discriminating rates were more than 96. 33% at the division level and more than 92. 53 % at genus level for single algae samples; and a discriminating rate of 84. 27% at division level and of 69.60% at genus level for mixed samples were given out.
出处
《中国海洋大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第6期108-114,共7页
Periodical of Ocean University of China
基金
国家自然科学基金项目(40706036)
国家高技术研究发展计划(2006AA09Z178)资助
关键词
浮游植物
特征识别
小波分解
三维荧光光谱
phytoplankton
feature recognition
wavelet analysis
3-D fluorescence spectra
