Study On Soil Heavy Metal Detection Method Based On Near Infrared Spectroscopy

With the development of society,the pollutants produced by human activities lead to excessive heavy metal elements in soil,resulting in soil pollution.Heavy metal pollution in soil not only reduces the quality of groundwater,surface water and air,but also endangers human health through bio-enrichment.It is extremely urgent to carry out risk assessment,monitoring and prevention of soil heavy metal pollution,and the content of soil heavy metal is an important index of site pollution investigation.Traditional soil heavy metal detection methods need laboratory reagents and professional operation of laboratory personnel,and cannot be used for large-scale monitoring of soil heavy metal pollutants.Due to its rapid and efficient characteristics,near infrared（NIR）spectral analysis technology is gradually applied in the field of rapid detection.Nir spectroscopy can reflect the information of hydrogen groups in substance molecules.Because heavy metals in soil are easy to combine with organics containing hydrogen groups,NIR spectroscopy also has an important scientific basis for analyzing heavy metals in soil.This topic uses the portable near infrared optical fiber spectrometer built in the laboratory to study the detection method and experiment of heavy metal content in soil.The soil was collected from qingshan Site pollution remediation site in Wuhan,150 soil samples were collected and screened by natural air drying and screen mesh.The contents of Chromium（Cr）,zinc（Zn）and lead（Pb）in samples were determined by X-ray fluorescence spectrometry and the NIR data of soil samples were measured by portable NIR spectrometer.The main research contents and results are as follows:1.Spectral data preprocessing: SG convolution smoothing algorithm,standard Normal variable transformation（SNV）algorithm and SG convolution smoothing + standard Normal variable transformation（SNV）algorithm were used to process the spectrum of soil samples,and the optimal pretreatment algorithm was selected.2.Spectral quantitative prediction modeling: For the spectral data after the pretreatment processing with competitive adaptive reweighted sampling（CARS）and successive projections algorithm（SPA）these two kinds of characteristics of band extraction algorithm for extracting feature band BP neural network,which are respectively,extreme learning machine（ELM）and random forest（RF）,a total of three kinds of modeling algorithms quantitative model to predict the soil heavy metal content,The best modeling algorithm for each heavy metal element was determined by comparative analysis.The experimental results show that the contents of zinc（Zn）,Chromium（Cr）and Lead（Pb）in the soil of qingshan District,Wuhan can be predicted and analyzed by portable near-infrared fiber optic spectrometer.After spectral data preprocessing,feature band extraction and machine learning algorithm integrated optimization modeling,zinc（Zn）and Lead（Pb）have achieved better prediction accuracy.Compared with the traditional laboratory bench NIR spectrometer,the portable NIR fiber optic spectrometer is more convenient and convenient to carry,and more suitable for the investigation of large-area site pollution remediation.