文章摘要
潘玉露.多源荧光光谱数据融合下的淡水浮游植物分类识别方法[J].林业调查规划,2024,49(3):7-12
多源荧光光谱数据融合下的淡水浮游植物分类识别方法
Classification and Recognition of Freshwater Phytoplankton Based onMulti-source Fluorescence Spectrum Data Fusion
  
DOI：
中文关键词: 淡水浮游植物  分类识别方法  多源荧光光谱  数据融合  特征提取  小波分解
英文关键词: freshwater phytoplankton  classification and recognition methods  multi-source fluorescence spectrum  data fusion  feature extraction  wavelet decomposition
基金项目:
作者单位
潘玉露 诸暨市水务集团 
摘要点击次数: 517
全文下载次数: 441
中文摘要:
      淡水浮游植物分类识别过程中,主要采用单一的荧光光谱数据进行特征提取,所得特征信息较 为片面,分类识别结果的F1 分数偏低。为此,以多源荧光光谱数据融合为前提,提出一种新型淡水浮 游植物分类识别方法。采用局部线性嵌入算法对多源荧光光谱数据进行降维处理,再通过小波分解 算法提取光谱特征信息。运用独立成分分析算法标记出有效的特征信息,依托于多源荧光光谱数据融 合原理结合有效特征得到全面的光谱特征信息。将光谱特征输入可解决多分类问题的支持向量机模型, 生成淡水浮游植物分类识别结果。实验结果显示,在噪声比例为40%时,文中设计的分类识别方法的分 类识别结果F1 分数依旧为0.95,与其他两种方法相比提高了14.74%和18.95%,分类结果更加准确。
英文摘要:
      In the process of classification and recognition of freshwater phytoplankton, single fluorescence spectrum data was mainly used to extract features, and the feature information obtained was relatively onesided, which made the F1 score of classification and recognition results low. Therefore, on the premise of multi-source fluorescence spectrum data fusion, a new classification and recognition method of freshwater phytoplankton was proposed. The local linear embedding algorithm was used to reduce the dimension of multisource fluorescence spectrum data, and then the wavelet decomposition algorithm was used to extract the spectral feature information. The independent component analysis algorithm was used to mark the effective feature information. Based on the multi-source fluorescence spectrum data fusion principle, the multi-source effective spectral features were fused to output comprehensive phytoplankton spectral feature information. The spectral features was input into the support vector machine model to solve the multi classification problem, and generate the classification and recognition results of freshwater phytoplankton. The experimental results showed that the F1 score of the proposed method was 0. 95 when the noise ratio was 40%, which was 14. 74% and 18. 95% higher than the other two methods, and the classification results were more accurate.
查看全文   查看/发表评论  下载PDF阅读器
关闭