小波分析在活体浮游植物离散三维荧光光谱特征提取及识别中的应用研究

Research on Wavelet Analysis in the Characteristics Extracting and Identification of Discrete 3D Fluorescence Spectra of Phytoplankton

为了区分和识别不同属的浮游植物,选择了Daubechies-3小波的二阶低频分量对10种浮游植物的离散三维光谱进行了特征提取.Bayes判别分析结果表明此类特征谱对不同属间浮游植物的正确判别率可达96.75%.利用非负最小二乘法,依据此类特征谱建立的标准谱库可对加入不同噪声的某些藻进行100%的定性识别.可对绝大多数混合样中优势种进行定性识别;并可使某些优势种的识别量达到真实量的75%以上.小波分析可对浮游植物在属的层次上进行有效的特征提取.

In order to discriminate and identify phytoplankton of different genuses, the second scaling function of Daubechies-3 was selected to extract the characteristics of 10 phytoplankton species. The results of Bayes discriminant analysis show that these characteristic spectra show a discriminating rate of 96.75% at genus level. Standard spectra were obtained from these characteristic spectra. The results of the nonnegative least square show that they can identify qualitatively some single species added with different ratios of random noise with a ratio of 100%. The dominant genus (species) of most phytoplankton mixtures can be identified by them qualitatively. The identifying amount of some species can be more than 75% of the real one. Wavelet analysis was proven to be an effective method in extracting the characteristic features of phytoplankton on the level of genus.