荧光光谱结合主成分分析对赤潮藻的识别测定

Discrimination of Red Tide Algae by Fluorescence Spectra and Principle Component Analysis

利用主成分分析和非负最小二乘法,建立了我国近海常见的11种赤潮藻在属水平上的活体荧光识别测定技术.利用Delaunay三维插值法,消除浮游藻三维荧光光谱的瑞利散射和拉曼散射峰,根据Fisher判别分析方法对各主成分及其组合的分辨能力的判别结果,确定赤潮藻三维荧光光谱的第一主成分得分谱和第二主成分得分谱作为赤潮藻识别特征谱,建立赤潮藻荧光特征谱库,在此基础上,利用非负最小二乘法对11种赤潮藻在属水平上进行识别测定,赤潮藻的识别正确率≥85%,其中9种赤潮藻的识别正确率≥90%,特别是对于近几年东海发生赤潮频率极高的东海原甲藻(Prorocentrum donghaiense)、中肋骨条藻(Skeletonema costatum)和裸甲藻(Gymnodinium sp.),识别正确率≥95%.本研究表明,利用主成分分析提取赤潮藻三维荧光光谱特征以实现赤潮藻在属水平上的识别测定是可行的.

Fluorescence discrimination technology for 11 species of the Red Tide algae at genus level was constructed by principle component analysis and non-negatlve least squares. Rayleigh and Raman scattering peaks of 3D fluorescence spectra were eliminated by Delaunay triangulation method. According to the results of Fisher linear discrimination, the first principle component score and the second component score of 3D fluorescence spectra were chosen as discriminant feature and the feature base was established. The 11 algae species were tested, and more than 85 % samples were accurately determinated, especially for Prorocentrum donghaiense, Skeletonema costatum, Gymnodinium sp., which have frequently brought Red tide in the East China Sea. More than 95% samples were right discriminated. The results showed that the genus discriminant feature of 3D fluorescence spectra of Red Tide algae given by principle component analysis could work well.