小波分析在湖泊常见浮游藻荧光识别测定中的应用

利用Daubechies7（db7）小波函数对27种分属于6个门,22个属的主要湖泊常见浮游藻的三维荧光光谱进行分解,得到各尺度分量及小波分量,应用Bayesian判别分析选择第三层尺度分量（Ca3）作为最佳特征谱,通过系统聚类分析得到浮游藻的标准谱库,结合非负最小二乘法解析的多元线性回归建立湖泊常见浮游藻的荧光识别测定技术.将该技术用于单种藻样品、模拟混合样品及实际混合样品的识别分析,单种藻样品在门类水平上的识别正确率平均为98.6%,识别测定的平均相对含量为90.8%.在实验过程中加入一定比例的噪声,考察了标准谱库的抗噪能力.对于104个实验室混合样品,优势门的藻在门类水平上的平均识别正确率为97.0%,识别的平均相对含量为67.7%,次优势门的藻在门类水平上的平均识别正确率为90.7%,识别的平均相对含量为32.3%.结果表明,所建立的识别测定技术具有可行性.

Fluorescence Discrimination Technique for Phytoplankton Based on the Wavelet Analysis

Daubechies7 (db7) wavelet was selected to decompose the 3-D fluorescence spectra of 27 species of phytoplankton belonging to 22 generas of 6 divisions found in major lakes, then the scale vectors and time-series vectors were obtained as candidates for feature spectra. The third scale vector (Ca3) of db7 was chosen as feature spectra by Bayesian discriminant analysis, and the reference spectra were obtained via hierarchical cluster analysis to feature spectra. Based on the above data, a fluorescence discrimination technique was developed by multiple linear regression resolved by non-negative least squares. For single species algae cultures, the average correct discrimination ratio (CDR) was 98.6%, with the average relative content of 90.8% at division level. Furthermore, the noise immunity of reference spectra was tested by adding noise at different proportions. For the dominant division of laboratory mixed samples, the average CDR was 97.0%, with the average relative content of 67.7% at division level, and the average CDR of subdominant division was 90.7%, with the average relative content of 32.3%. The results showed that the technique is feasible to some extent.