本文選題：拉曼光譜　切入點(diǎn)：光譜質(zhì)量評(píng)價(jià)　出處：《中國(guó)科學(xué)院研究生院(長(zhǎng)春光學(xué)精密機(jī)械與物理研究所)》2014年博士論文　論文類型：學(xué)位論文

【摘要】：拉曼光譜分析技術(shù)由于具有無(wú)損、信息豐富、無(wú)需樣品制備等優(yōu)點(diǎn)，，在食品、材料、環(huán)境監(jiān)測(cè)等眾多領(lǐng)域得到了越來(lái)越廣泛的應(yīng)用。手持拉曼光譜儀由于具有操作簡(jiǎn)單、小巧輕便等優(yōu)點(diǎn)被廣泛應(yīng)用于工業(yè)生產(chǎn)中的物料鑒定。目前，國(guó)外各大光譜儀生產(chǎn)商均已經(jīng)推出了各種型號(hào)的手持拉曼光譜儀，國(guó)內(nèi)市場(chǎng)已被其壟斷，因此研制我國(guó)擁有獨(dú)立自主知識(shí)產(chǎn)權(quán)的手持拉曼光譜儀具有重要意義。 本文針對(duì)手持拉曼光譜儀數(shù)據(jù)處理與定性分析技術(shù)展開(kāi)了相關(guān)研究。由于手持拉曼光譜儀主要應(yīng)用于工業(yè)生產(chǎn)中的定性判別問(wèn)題，生產(chǎn)線上操作員往往不具備專業(yè)的化學(xué)分析知識(shí)，因此減少拉曼光譜分析過(guò)程中的人工干預(yù)，實(shí)現(xiàn)拉曼光譜數(shù)據(jù)處理與定性分析的自動(dòng)化是手持拉曼光譜定性分析技術(shù)的關(guān)鍵。 本文系統(tǒng)地研究了手持拉曼光譜的數(shù)據(jù)處理與定性分析算法流程，主要研究工作如下： (1)研究了拉曼光譜質(zhì)量評(píng)價(jià)方法。實(shí)現(xiàn)了小尺度小波變換法、空域相關(guān)小波變化法、Donoho魯棒估計(jì)法及改進(jìn)的二階差分法等噪聲標(biāo)準(zhǔn)差估計(jì)方法，對(duì)比了其估計(jì)噪聲標(biāo)準(zhǔn)差的精度，結(jié)果表明改進(jìn)的二階差分法的估計(jì)精度最高；提出了一種新的信噪比計(jì)算方法，新的信噪比計(jì)算方法與傳統(tǒng)的信噪比定義相比能更好的表征信號(hào)質(zhì)量。 (2)研究了拉曼光譜數(shù)據(jù)處理技術(shù)。實(shí)現(xiàn)了各種常見(jiàn)的光譜預(yù)處理方法，重點(diǎn)研究了拉曼光譜的去尖峰、去基線、去噪聲方法。提出了一種無(wú)需設(shè)置任何參數(shù)的、可實(shí)現(xiàn)完全自動(dòng)化的去尖峰方法——改進(jìn)的循環(huán)消去法；實(shí)現(xiàn)了一種基于三點(diǎn)零階Savitzky-Golay濾波器的自動(dòng)化降噪方法，對(duì)比了其與傳統(tǒng)的滑動(dòng)窗口平均法、滑動(dòng)窗口中位值法、Savitzky-Golay濾波器法、小波閾值濾波法的降噪效果，數(shù)值實(shí)驗(yàn)表明該方法具有最優(yōu)的去噪效果，同時(shí)引起的譜峰退化程度也最小，能夠最大量的保留譜峰信息；提出了一種新的基線估計(jì)方法——改進(jìn)的小窗口滑動(dòng)平均法，該方法無(wú)需設(shè)置任何參數(shù)，可實(shí)現(xiàn)基線的自動(dòng)估計(jì)，估計(jì)精度良好。 (3)研究了拉曼光譜譜峰識(shí)別技術(shù)。實(shí)現(xiàn)了連續(xù)小波變換法識(shí)別拉曼譜峰，提出了兩種新的拉曼譜峰識(shí)別方法——雙尺度相關(guān)法及多尺度局部信噪比法，對(duì)比了其與連續(xù)小波變換法識(shí)別拉曼譜峰的能力。仿真實(shí)驗(yàn)表明，多尺度局部信噪比法具有最優(yōu)的譜峰識(shí)別能力。對(duì)于處于檢測(cè)限的單峰，仍有95.1%的識(shí)別準(zhǔn)確率，譜峰信噪比大于等于6時(shí)譜峰識(shí)別準(zhǔn)確率高達(dá)100%；對(duì)于重疊峰，譜峰信噪比大于等于7時(shí)達(dá)到100%識(shí)別。多尺度局部信噪比法具有最高的峰位估計(jì)準(zhǔn)確度。 (4)研究了拉曼光譜判別分析技術(shù)。實(shí)現(xiàn)了直接比較法和基于簇類的軟獨(dú)立模型法，對(duì)比了兩者的性能。直接比較法的不足是，當(dāng)參考譜庫(kù)中沒(méi)有待測(cè)樣品的參考光譜時(shí)，直接比較法仍然會(huì)給出一個(gè)最佳的匹配結(jié)果�；诖仡惖能洩�(dú)立模型法具有更加優(yōu)越的性能，該方法具有較高的識(shí)別準(zhǔn)確率，當(dāng)未知樣品不屬于類庫(kù)中的任何類時(shí)，該方法可識(shí)別出未知樣品屬于某一新的類型。利用基于簇類的軟獨(dú)立模型法還可以獲得兩類間的相似度、變量對(duì)樣品判別的重要性、樣品與某類的相關(guān)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.

【學(xué)位授予單位】：中國(guó)科學(xué)院研究生院(長(zhǎng)春光學(xué)精密機(jī)械與物理研究所)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2014
【分類號(hào)】：O433.4

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 白利濤;張麗萍;趙國(guó)文;;拉曼光譜的應(yīng)用及進(jìn)展[J];福建分析測(cè)試;2011年02期

2 丁松園;吳德印;楊志林;任斌;徐昕;田中群;;表面增強(qiáng)拉曼散射增強(qiáng)機(jī)理的部分研究進(jìn)展[J];高等學(xué)�；瘜W(xué)學(xué)報(bào);2008年12期

3 劉鵬;李吉祥;李文靜;;表面增強(qiáng)拉曼光譜在反恐中的應(yīng)用[J];光譜實(shí)驗(yàn)室;2010年01期

4 林逸群;LRZ-1型激光拉曼光譜儀計(jì)算機(jī)系統(tǒng)的研制[J];光譜學(xué)與光譜分析;1988年01期

5 李慶波,楊麗敏,凌曉鋒,王健生,周孝思,石景森,吳瑾光;SIMCA法在中紅外癌癥檢測(cè)技術(shù)中的應(yīng)用[J];光譜學(xué)與光譜分析;2004年04期

6 楊忠;江澤慧;費(fèi)本華;覃道春;;SIMCA法判別分析木材生物腐朽的研究[J];光譜學(xué)與光譜分析;2007年04期

7 陳珊;李曉寧;梁逸曾;張志敏;劉朝霞;張啟明;丁麗霞;葉菲;;拉曼光譜的熒光背景扣除及其用于藥物聚類分析[J];光譜學(xué)與光譜分析;2010年08期

8 劉燕德;劉濤;孫旭東;歐陽(yáng)愛(ài)國(guó);郝勇;;拉曼光譜技術(shù)在食品質(zhì)量安全檢測(cè)中的應(yīng)用[J];光譜學(xué)與光譜分析;2010年11期

9 李曉舟;楊天月;丁建華;;唾液表面增強(qiáng)拉曼光譜用于肺癌的診斷[J];光譜學(xué)與光譜分析;2012年02期

10 鄒文龍;蔡志堅(jiān);吳建宏;;便攜式拉曼光譜儀發(fā)展綜述[J];光學(xué)儀器;2011年06期

相關(guān)博士學(xué)位論文 前3條

1 紀(jì)偉;表面增強(qiáng)拉曼散射化學(xué)增強(qiáng)機(jī)理的研究及其應(yīng)用[D];吉林大學(xué);2013年

2 來(lái)永超;面向環(huán)境有機(jī)污染物快速分析的表面增強(qiáng)拉曼光譜方法[D];山東大學(xué);2013年

3 黃祖芳;基于拉曼光譜的人體精液質(zhì)量的初步研究[D];福建師范大學(xué);2013年

本文編號(hào)：1649709