Zhang Ting-Lu, Du Xiang-Zhi, Xu Qing-Na, Qiu Guo-Qiang
Ocean Remote Sensing Laboratory of the Ministry of Education, Ocean University of China, Qingdao 266100, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2009 Oct;29(10):2743-7.
In the present study, the feasibility of using wavelet analysis to extract the eigen spectra from the absorption spectra of phytoplankton for species detection was investigated. Thirteen absorption spectra taken from single species cultures, representing four divisions (Dinophyta, Bacillariophyta, Haptophyta, and Chlorophyta), six genus (Gymnodinium, Prorocentrum, Skeletonema, Guinardia, Phaeocystis, and Prasinophyte) and seven species (Karenia mikimotoi, Karenia brevis, Prorocentrum minimum, Skeletonema costatuma, Guinardia delicatula, Phaeocystis globosa, and Pyramimonas parkeae), were used. First, the 1D wavelet analysis with five levels was applied to the thirteen absorption spectra, so each spectrum was decomposed with 5 levels. The 5th level component of low frequency corresponds to the background without information for species detection, and 1st and 2nd level component of high frequency is the random noise. The other levels (3rd to 5th) of high frequency are the useful information, and the sum of levels (3rd to 5th) of high frequency was retained as the eigen spectra for species detection. Second, the clustering analysis was used to the eigen spectra for examining the performance of the wavelet analysis in extracting species information. The clustering results were compared with the known species class information, and the results show that the 13 absorption spectra are correctly classified at the level of division, genus and species. This means that the wavelet analysis has good performance in extracting the eigen spectra for species detection. However, the above results were obtained with only limited species, and the more species data are required to identify the extensive validity of the conclusion.
在本研究中，探讨了使用小波分析从浮游植物吸收光谱中提取特征光谱以进行物种检测的可行性。使用了从单物种培养物中获取的13条吸收光谱，这些光谱代表了四个门类（甲藻门、硅藻门、定鞭藻门和绿藻门）、六个属（裸甲藻属、原甲藻属、骨条藻属、格氏藻属、棕囊藻属和原绿球藻属）和七个物种（米氏凯伦藻、短裸甲藻、微小原甲藻、中肋骨条藻、柔弱格氏藻、球形棕囊藻和帕克扁藻）。首先，对这13条吸收光谱应用五级一维小波分析，因此每条光谱都被分解为5级。低频的第5级分量对应于无物种检测信息的背景，高频的第1级和第2级分量是随机噪声。高频的其他级别（第3级到第5级）是有用信息，高频的第3级到第5级的总和被保留为用于物种检测的特征光谱。其次，对特征光谱进行聚类分析以检验小波分析在提取物种信息方面的性能。将聚类结果与已知的物种类别信息进行比较，结果表明这13条吸收光谱在门类、属和物种级别上被正确分类。这意味着小波分析在提取用于物种检测的特征光谱方面具有良好的性能。然而，上述结果仅在有限的物种中获得，需要更多的物种数据来确定该结论的广泛有效性。
