A Study On Data Preprocessing And Material Analysis Algorithm Of Portable Fourier Transform Spectroscopy System

Safety of food’s quality is a basic guarantee of people’s life. A variety of foodadditives and residues are double-edged swords. While they safeguard the quality offood, they will cause serious environmental pollution and threat consumers’ health. Asthe issues of safety of food’s quality will directly affect the development of China’sfood trade, so the research about how fast and accurately detect harmful residues infood are of great importance with good prospects. Fourier transform infraredspectroscopy with high detection precision and speed can achieve fast and accuratedetection of harmful substances and offer a guarantee for the quality of life ofconsumers.The appearance of infrared spectroscopy makes food safety testing technologymore quickly and accurately and has the advantages that traditional methods do nothave, such as: fast and high precision in the material testing, non-destructive, nosample pre-processing, high efficiency of analysis and etc. With the development ofinfrared spectroscopy instruments, chemometrics and computer technology, theyprovide the hardware and software infrastructure to achieve fast and accurate portablespectrum analyzer. However, it remains difficult to distinguish components, samples,spectrum analysis in the infrared spectrum analysis technology. This paper iscommitted to study a convenient algorithm of substance identification to achieve fastand accurate on-site spectroscopy.In order to realize material analysis algorithms, it needs to establish a chemometrics method in substances analysis based on MEMS micro-mirrors miniaturizedFourier transform infrared spectroscopy detection system according to macroscopicfingerprint characteristics of infrared spectra of the substance to be tested. This paperpresents a method by comparing the Euclidean distance of the characteristic peak ofspectral lines between the mixture ratio and the standard spectrum to determine thecomponents contained in the sample. By extracting the sample absorption spectracharacteristic peaks, it calculates the consistent rate between the samples absorptionspectral characteristics peak and that of spectral database to determine what kind ofpure substances contained in the mixture in sample, and the weight of respectivecharacteristic peak. Then, it calculates the weighted Euclidean distance between thecharacteristic curves and finds that the shorter the weighted Euclidean distance is, the higher the spectral similarity is, namely, the higher levels of the substance. So the levelof various pure substances in the mixture in sample is obtained and achieves a fast andaccurate analysis of the material.This paper presents a method by comparing the Euclidean distance of thecharacteristic peak of spectral lines between the mixture ratio and the standardspectrum to determine the components contained in the sample. By extracting thesample absorption spectra characteristic peaks, it calculates the consistent rate betweenthe samples absorption spectral characteristics peak and that of spectral database todetermine what kind of pure substances contained in the mixture in sample, and theweight of respective characteristic peak. Then, it calculates the weighted Euclideandistance between the characteristic curves and finds that the shorter the weightedEuclidean distance is, the higher the spectral similarity is, namely, the higher levels ofthe substance. So the level of various pure substances in the mixture in sample isobtained and achieves a fast and accurate analysis of the material.This paper accomplishes the theoretical analysis of the method, the design processflow diagram and function block diagram, improves the algorithm process. And it alsodevelops software of Linux-based intelligent substance analysis system to be applied toa portable spectrum analyzer for rapid on-site analysis of the material. Meanwhile, theexperiment proves the feasibility of this method and determines the accuracy of themethod.The experimental results show that: the algorithm can programmatically achieverapid detection of substances in the field.When applied to the test mixture spectroscopy,it meets the demand for routine testing with more than96percent accuracy. Beingsystematic, standardized and programmed, this method makes up the traditionalchemometric methods in substance analysis, achieves rapid on-site materialidentification and enhances the efficiency of material analysis.