拉曼光谱数据处理与定性分析技术研究

发布时间：2018-03-22 17:44

本文选题：拉曼光谱　切入点：光谱质量评价　出处：《中国科学院研究生院(长春光学精密机械与物理研究所)》2014年博士论文　论文类型：学位论文

【摘要】：拉曼光谱分析技术由于具有无损、信息丰富、无需样品制备等优点，，在食品、材料、环境监测等众多领域得到了越来越广泛的应用。手持拉曼光谱仪由于具有操作简单、小巧轻便等优点被广泛应用于工业生产中的物料鉴定。目前，国外各大光谱仪生产商均已经推出了各种型号的手持拉曼光谱仪，国内市场已被其垄断，因此研制我国拥有独立自主知识产权的手持拉曼光谱仪具有重要意义。本文针对手持拉曼光谱仪数据处理与定性分析技术展开了相关研究。由于手持拉曼光谱仪主要应用于工业生产中的定性判别问题，生产线上操作员往往不具备专业的化学分析知识，因此减少拉曼光谱分析过程中的人工干预，实现拉曼光谱数据处理与定性分析的自动化是手持拉曼光谱定性分析技术的关键。本文系统地研究了手持拉曼光谱的数据处理与定性分析算法流程，主要研究工作如下： (1)研究了拉曼光谱质量评价方法。实现了小尺度小波变换法、空域相关小波变化法、Donoho鲁棒估计法及改进的二阶差分法等噪声标准差估计方法，对比了其估计噪声标准差的精度，结果表明改进的二阶差分法的估计精度最高；提出了一种新的信噪比计算方法，新的信噪比计算方法与传统的信噪比定义相比能更好的表征信号质量。 (2)研究了拉曼光谱数据处理技术。实现了各种常见的光谱预处理方法，重点研究了拉曼光谱的去尖峰、去基线、去噪声方法。提出了一种无需设置任何参数的、可实现完全自动化的去尖峰方法——改进的循环消去法；实现了一种基于三点零阶Savitzky-Golay滤波器的自动化降噪方法，对比了其与传统的滑动窗口平均法、滑动窗口中位值法、Savitzky-Golay滤波器法、小波阈值滤波法的降噪效果，数值实验表明该方法具有最优的去噪效果，同时引起的谱峰退化程度也最小，能够最大量的保留谱峰信息；提出了一种新的基线估计方法——改进的小窗口滑动平均法，该方法无需设置任何参数，可实现基线的自动估计，估计精度良好。 (3)研究了拉曼光谱谱峰识别技术。实现了连续小波变换法识别拉曼谱峰，提出了两种新的拉曼谱峰识别方法——双尺度相关法及多尺度局部信噪比法，对比了其与连续小波变换法识别拉曼谱峰的能力。仿真实验表明，多尺度局部信噪比法具有最优的谱峰识别能力。对于处于检测限的单峰，仍有95.1%的识别准确率，谱峰信噪比大于等于6时谱峰识别准确率高达100%；对于重叠峰，谱峰信噪比大于等于7时达到100%识别。多尺度局部信噪比法具有最高的峰位估计准确度。 (4)研究了拉曼光谱判别分析技术。实现了直接比较法和基于簇类的软独立模型法，对比了两者的性能。直接比较法的不足是，当参考谱库中没有待测样品的参考光谱时，直接比较法仍然会给出一个最佳的匹配结果。基于簇类的软独立模型法具有更加优越的性能，该方法具有较高的识别准确率，当未知样品不属于类库中的任何类时，该方法可识别出未知样品属于某一新的类型。利用基于簇类的软独立模型法还可以获得两类间的相似度、变量对样品判别的重要性、样品与某类的相关性等信息。
[Abstract]:Raman spectroscopy analysis technology with nondestructive, rich information, without sample preparation etc, in food, materials, environmental monitoring and other fields has been more and more widely used. The handheld Raman spectrometer has the advantages of simple operation, compact and lightweight materials are widely used in the identification of industrial production. At present, the major spectrometer manufacturers abroad have already launched the handheld Raman spectrometer of various types, the domestic market has been the monopoly, therefore the development of our country has significance of independent intellectual property rights of handheld Raman spectrometer.
According to the handheld Raman spectrometer data processing and qualitative analysis technology related research has been carried out. Because of discrimination is mainly applied to the qualitative handheld Raman spectrometer in the industrial production, the production line operators often do not have the professional knowledge of chemistry, thus reducing the Raman spectral analysis in the process of manual intervention, automatic data processing and analysis of Raman spectra qualitative is a handheld Raman spectra qualitative analysis of the key technology.
This paper systematically studies the data processing and qualitative analysis algorithm flow of handheld Raman spectra. The main research work is as follows:
(1) studied the evaluation method. The Raman spectral quality small scale wavelet transform method, spatial correlation of wavelet transform method, Donoho robust estimation method and modified two order difference method difference estimation method of noise standard, compared the standard deviation of the noise estimation accuracy, results show that modified two order difference method the estimation accuracy is highest; proposes a new SNR calculation method, characterization of signal quality of the signal-to-noise ratio and the calculation method of the traditional definition of signal-to-noise ratio compared to better.
(2) the study of Raman spectrum data processing technology. To achieve a variety of spectral preprocessing methods commonly, focuses on the study of Raman spectroscopy to peak to baseline denoising method. We propose a no need to set any parameters that can be fully automated to cycle peak elimination method - improvement; implement an automatic noise reduction method based on three order Savitzky-Golay filter, compared with the traditional sliding window averaging method, method of bit values in a sliding window, Savitzky-Golay filter method, the noise reduction effect of wavelet threshold filter method, the numerical results show that the method has the best denoising effect, at the same time caused by the peak degradation degree minimum, to retain the greatest amount of spectral information; propose a new baseline estimation method -- small window sliding average method improved, the method does not need to set any parameters, can be realized The automatic estimation of the baseline is of good accuracy.
(3) Raman spectra identification technology is researched. The continuous wavelet transform method to identify the Raman peaks, we proposed two new Raman peak identification method of dual scale correlation method and multi-scale local signal-to-noise ratio method, compared with the continuous wavelet transform method to identify the Raman peak of simulation capability. Experimental results show that the multi-scale local peak signal-to-noise ratio method. The optimal recognition ability in the detection limit of single peak, there are still 95.1% of the recognition accuracy, peak signal-to-noise ratio is greater than or equal to 6 peak recognition accuracy rate of up to 100%; for overlapping peaks, peak signal-to-noise ratio is greater than or equal to 100% recognition 7. Multi scale local signal-to-noise has the highest peak position estimation accuracy ratio method.
(4) studied by Raman spectroscopy technology. The discriminant analysis and direct comparison method based on soft independent model clusters, and their performances are compared. The lack of direct comparison method is that when the reference spectrum without reference spectrum sample in the library, the direct comparison method will still be given a best match results. Soft independent model based on clustering has more superior performance, this method has high recognition accuracy, when unknown samples do not belong to any class in the class, the method can identify the unknown samples belong to a new type. Using soft independent model method based on clustering can also get two similarity between classes, variables on the importance of sample discrimination, samples and certain types of correlation information.

【学位授予单位】：中国科学院研究生院(长春光学精密机械与物理研究所)
【学位级别】：博士
【学位授予年份】：2014
【分类号】：O433.4

【参考文献】