时间序列三维荧光光谱的异常值检测

异常值的存在往往干扰着时间序列三维荧光光谱的定性和定量分析。充分利用时间维和光谱维的内在特性, 提出了一种有效的异常值检测方法。在时间维结合方差提取异常值可能性最大的波长点;通过对异常值存在方式的分析，在任意两个三维荧光光谱的相似度基础上给出了光谱维上的累积相似度;最后利用时间维的校正矩阵对所有三维荧光光谱的每个波长点荧光强度进行修正并计算对应的累积相似度，从而根据累积相似度对异常值进行判断。时间维校正矩阵的采用不仅提高了算法的有效性而且其特征区域的选择大大减少了光谱维相似度的计算量。相关的数值试验表明光谱维选取50%的波长点仍然能有效对异常值进行检测。

Outlier Detection of Time Series Three-Dimensional Fluorescence Spectroscopy

The qualitative and quantitative analysis are often interfered by the outliers in time series three-dimensional fluorescence spectroscopy. In this work, an efficient outlier detection method is proposed by taking advantage of the characteristics in time dimension and the spectral dimension. Firstly, the wavelength points that are mostly the outliers are extracted by the variance in time dimension. Secondly, by the analysis of the existence styles of outliers and similarity score of any two samples, the cumulative similarity is introduced in spectral dimension. At last, fluorescence intensity at each wavelength of all samples is modified by the correction matrix in time dimension and the outlier detection is completed according the to cumulative similarity scores. The application of the correction matrix in time dimension not only improves the validity of the method but also reduces the computation by the choice of characteristics region in correction matrix. Numerical experiments show that the outliers can still be detected by the 50 percent of all points in spectral dimension.