几种图像矩特征在分析中的应用研究

发布时间：2018-03-08 06:33

本文选题：复杂体系　切入点：化学计量学　出处：《兰州大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着分析研究的深入,环境样本、中药以及生物样本等复杂多组分体系的定性定量分析成为化学分析中重要的研究课题之一。现代分析仪器的发展在为复杂体系的检测提供有力的分析手段的同时,获得的样本信息越来越全面。如何从这些分析仪器给出数据中快速、有效地提取出有用信息,实现待测组分的精准分析,成为分析化学工作者处理复杂体系分析的一大挑战。化学计量学的发展为复杂体系分析研究提供了较大的便利并取得了理想的结果。本学位论文将几种图像矩方法应用于分析领域中,利用矩方法的旋转、平移、缩放不变性以及多分辨等特性,实现水样、中药以及生物序列等体系中感兴趣组分的分析研究。本学位论文主要包括内容如下:第一章绪论部分。主要介绍了分析化学研究体系现状、分析仪器在复杂体系中的分析研究与应用、化学计量学在复杂体系中的分析研究与应用、图像矩特征。最后对本学位论文研究工作做了简单概括。第二章基于三维荧光光谱,介绍了Zernike矩方法用于水中腐殖酸含量测定的一种有效方法。水环境中腐殖酸含量的多少会影响水的质量,从而对人及动植物带来影响。荧光分光光度计给出的腐殖酸水样三维图谱,由于干扰物质以及瑞丽和拉曼散射等信号的影响,很难得到腐殖酸的纯组分峰,从而给传统定量分析方法带来困难。为了获得腐殖酸的含量信息,本章提出具有多分辨特征的Zernike矩方法提取三维荧光光谱的有效特征信息,之后借助逐步回归建立定量分析模型,以实现腐殖酸含量测定。本研究方法与多维偏最小二乘和交替三线性分解方法进行了对比,结果表明,我们提出的方法建立的模型更可靠、结果更准确,可以实现水中目标组分的定量分析。第三章基于高效液相色谱-二极管阵列检测器(HPLC-DAD),使用Tchebichef图像矩方法实现了对中成药的质量评估。中药成分复杂而且其原材料易受到外界环境影响,对中成药进行质量分类或判别可以实现中药产品质量监控,避免中药市场混乱。使用HPLC-DAD测定了14个不同批次的蒲地蓝消炎片的三维指纹图谱,采用Tchebichef矩方法对指纹图谱提取特征值,并借助层次聚类实现样本分类。虽然获得的指纹图谱存在噪音干扰以及峰漂移的影响,但由于矩方法多分辨性和不变性的特征,分析结果不受影响。通过与主成分分析得到的结果做对比,进一步说明我们方法分类结果的准确性。因此,本研究表明我们的方法能够为药品的质量监控提供一个快速准确的方法。第四章基于数值化的蛋白质序列,结合Krawtchouk图像矩方法我们提出了一种蛋白质磷酸化位点预测新方法。蛋白质磷酸化是细胞内重要的调节机制,磷酸化位点的有效辨别对生命活动研究有重要意义。本章提出了使用Krawtchouk矩方法对拟南芥蛋白质磷酸化位点进行预测。首先借助氨基酸描述符将蛋白质序列数值化,之后将数值矩阵转换成灰度图后使用Krawtchouk矩提取局部特征信息,最后所得矩值用于LibSVM建立预测模型。本工作中分别采用10-折交叉验证和独立测试集评价所建立的预测模型,并且选取4个已有的预测工具做了对比。结果表明,我们方法所建的模型有较好的准确性和可靠性,可以为蛋白质磷酸化位点实验测定提供较好的参考信息。第五章简单总结了本学位论文研究工作,并针对本实验室的研究和发展简单陈述了自己的观点。
[Abstract]:With the analysis of the in-depth study, environmental samples, biological samples and analysis of traditional Chinese medicine complex multicomponent system qualitative and quantitative become one of the important research topics in chemical analysis. The development of modern analytical instruments and powerful analysis method in detection of complex system at the same time, the sample information obtains more and more comprehensive. How fast from these analytical instrument data are given, and effectively extract the useful information, accurate analysis of analytes, has become a big challenge for chemical analysis of workers in dealing with the complex system analysis. The development of chemometrics provides great convenience and achieve satisfactory results in complex system analysis. This dissertation several methods of image moments in the analysis of the field, by using the moment method of translation, rotation, zoom invariance and water multiresolution characteristics, traditional Chinese medicine and biological sequence etc. Analysis on the interest component sense system. This dissertation mainly includes the contents as follows: the first chapter is the introduction part. Mainly introduces and analyzes the present situation of chemical research system, the research and application of analytical instruments in the complex system, and the application of Chemometrics in complex system research, the characteristics of image moments. Finally, do a simple summary of the research work of this dissertation. The second chapter is based on the three-dimensional fluorescence spectra, introduces the Zernike moment method is an effective method for the determination of humic acid content. How much will the content of humic acid in the water environment and affect water quality, so as to bring effects to people and animals and plants. The three-dimensional map of humic acid samples fluorescence spectrophotometry this meter, due to interfering substances and Ruili and the Raman scattering signal, it is difficult to set the peak purity of humic acid, which gives the traditional quantitative analysis method is difficult. The information content of humic acid, this chapter puts forward the effective feature extraction of 3D fluorescence spectra with Zernike moment method of multi-resolution feature, and then with the help of stepwise regression to establish the quantitative analysis model, in order to realize the determination of humic acid content. And this research method of multidimensional partial least square and three alternate linear decomposition methods were compared, results that, we present a method for building the model is more reliable, more accurate results can be achieved in quantitative objective component analysis. The third chapter is based on the high performance liquid chromatography with diode array detector (HPLC-DAD), using Tchebichef image moment method to realize quality evaluation of traditional Chinese medicine. Chinese medicine complex and its raw materials susceptible to outside environmental impact, quality classification of Chinese medicine or discrimination can achieve the quality monitoring of traditional Chinese medicine, Chinese medicine to avoid confusion in the market. The use of HPLC-DAD measured 14 The three-dimensional fingerprint of different batches of Pudilanxiaoyan tablets, the fingerprint feature extraction value using Tchebichef moment method, and with the implementation of classification hierarchical clustering. Although the fingerprint obtained in presence of noise interference and influence of peak drift, but due to the characteristics of multi-resolution and invariant moment method. The analysis result is not affected by the contrast. The results obtained with principal component analysis, the accuracy of classification results further illustrate our method. Therefore, this study shows that our method can provide a rapid and accurate method for monitoring the quality of medicines. The fourth chapter is based on the numerical value of the protein sequence, combined with Krawtchouk image moment method we propose a new method for protein phosphorylation site prediction. Protein phosphorylation is an important mechanism of regulating cells, effectively identify phosphorylation sites are important for life activities research Meaning. This chapter proposes to use the Krawtchouk moment method to predict the Arabidopsis protein phosphorylation sites. The protein sequence of amino acids by means of numerical descriptors, then numerical matrix is converted into gray image using Krawtchouk moments to extract the local feature information, finally the moment for the LibSVM prediction model is established. The prediction model adopted respectively 10- fold cross validation and independent test set of the evaluation, compares the forecasting tool and select 4 existing. The results show that the method we model has better accuracy and reliability, can provide reference information for protein phosphorylation sites measured. The fifth chapter briefly summarizes the research work of the thesis, according to the research and development of the laboratory a simple statement of his views.

【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O652

【相似文献】