Research On The Application Of Several Image Moment Characteristics In Analysis

With the deepening of the analysis research,the qualitative and quantitative analysis for complex multi-component samples such as environmental samples,traditional Chinese medicine and biological samples has become one of the research focuses in chemical analysis.The rapid development of modern analytical instrumentation has provided a powerful tool for the detection of complex system,and the obtained information was more and more comprehensive.While the problem of quickly and efficiently extracting the useful chemical information from the instrumental analysis data for accurate analysis the interest component has become a major challenge to analytical chemists.The development of Chemometrics has provided a great convenience in the study of complex system,and obtained the ideal result.In this dissertation,by using the characteristic of rotation,translation and scaling invariance as well as multi-resolution,several image moment method has applied to the analysis of water sample,traditional Chinese medicine and biological sequence,so as to realize the quantitative and qualitative analysis of the interest component.The main researches of this dissertation are as follows:Chapter 1 The introduction mainly included the parts that the present situation of the analytical chemistry research system,the research and application of the analysis instrument in the complex system,the research and application of Chemometrics in the complex system,and the characteristic of the image moment.At last,the research work of this dissertation is summarized briefly.Chapter 2 Based on three-dimensional(3D)fluorescence spectrum,we have proposed an effective method that using Zernike moment method for the quantitative analysis of humic acid(HA)in water.The content of HA will affect the quality of water and even influence the health of people and other creatures.In this chapter,based on fluorescence spectrophotometer,the 3D fluorescence spectrum of HA were obtained.However,because of the presence of interference substance as well as Rayleigh and Raman scatterings,it is hard obtained the pure peak of HA and difficult realized the quantitative analysis based on traditional method.Thus,we have proposed Zernike moment method that with multi-resolution feature to extract the effective information from 3D fluorescence spectra,then building quantitative analysis model by means of stepwise regression.Multiway partial least squares and alternative trilinear decomposition methods were used for comparison,and the results from our model were more accurate and reliable,which illustrated that our proposed method could apply in the quantitative analysis of target components in water.Chapter 3 Based on high performance liquid chromatography with diode array detector(HPLC-DAD),we have realized the quality assessment for traditional Chinese medicine(TCM)using Tchebichef image moment.Usually,the composition of TCM is complex and its raw materials are easily affected by external environment.To avoid the TCM market chaos,it is necessary to monitor the TCM quality by classifying or discrimination samples.In the chapter,the HPLC-DAD was used for obtaining the 3D fingerprint spectra of 14 batches of Pudilan Xiaoyan tablets.Based on the spectra,Tchebichef image moment was applied for extracting the features,and then hierarchical cluster analysis was used for classification.Although the influence of noise interference and peak drift,ideal effect was still obtained,which owing to the characteristics of invariance and multi-resolution.Compared with the traditional principal component analysis method,our method has obtained more accurate and reliable results.Thus,our method can provide a fast and accurate method for pharmaceutical quality control.Chapter 4 Based on the numerical protein sequence,a novel method for protein phosphorylation sites predicting was proposed using Krawtchouk image moment method.Protein phosphorylation is an important regulatory mechanism in intracellular,and the effective identification of phosphorylation sites is of great significance in the study of life activities.In the chapter,we have proposed to use Krawtchouk moment method for predicting the Arabidopsis thaliana protein phosphorylation sites.At first,the protein sequence was numerical based on amino acid descriptors;and then Krawtchouk moment was used for extracting features after the numerical matrix turned into grayscale image;lastly,LibSVM was applied for building model based on the extracted features.In the work,10-fold cross validation and independent test set were used for evaluating our model.And four other specialized prediction tools were used for comparison.The study indicated that our proposed method can provide accurate and reliable results,and also provide good reference information to phosphorylation sites prediction experiments.Chapter 5 The research works of the dissertation have been simply summarized.And the view of the research and development in future was simple stated.