The Rapid Detection Of Heavy Metal Pollution In Tegillarca With One-class Classificationand Infrared Spectrum Detection Technology

Cadmium,copper,zinc,lead and other heavy metal pollution elements lead global environmental pollution.If the aquatic organisms caused by heavy metal pollution enters the human's body,as food,it will pose a major threat to life,health and safety,and the use of spectral analysis which is fast,safe,low cost,no damage to the sample detection is a very good way.But the characteristic of different sources of pollution and small pollution sample in reality determines the lack of abnormal samples.Therefore,using one class classification algorithm for normal Tegillarca samples training,to build a model,and realize the classification prediction set including normal and abnormal samples.For recognition,the three one-class classification methods are adopted to realize the detection of Tegillarca in heavy metals.They are the support vector data description algorithm(SVDD),one class classification of partial least squares(ocpls)and one-class Gauss process(ocgp).The main contents of this paper are as follows:1.First of all,from the grim situation of heavy metal pollution,we take Tegillarca granosa,the aquatic organism as the research object,which makes sense in detection of heavy metal pollution.Then,it introduces the spectroscopic study of the history,basic principle and application status,this is the advantage of utilizing spectral analysis as a means of detecting.Then,the one-class classification approach is introduced and resumed the principle and research status,which is the necessary foundation for spectral detection of heavy metals.2.For normal and heavy metal contamination of sample data,they are firstly preprocessed,which is based on the preparation of the application of three kinds of one-class classification methods.Including the Savitzky-Golay filter(S-G filter)and the sample consistency detection.Among them,the savitzky Golay filter take sample data in smooth processing,making the noise reduced.The consistency checking of sample verifies the reliability of experimental data in statistical method.3.From the perspective of different one-class classification methods,we analyze the difference between the four heavy metals samples and normal samples.Secondly,from the perspective of different heavy metal elements in samples,we analyze the accuracy of different classification methods,and focuses on the research and analysis of the support vector data description algorithm(SVDD),one-class of partial least squares classification(ocpls)and one-class Gauss process(ocgp)in predicting the actual samples.For the cadmium sample,the classification method of ocgp is the best,but dependence on the number of training samples is relatively strong in the algorithm.For Lead and zinc samples,the best method is ocpls,the classification and recognition rate is about 95%,and the dependence on the number of training samples is not strong in the algorithm.For svdd,regardless of what kind of sample the classification accuracy is not high.