Research On Spectral Data Preprocessing Algorithms Based On LIBS Technology

Laser-induced breakdown spectroscopy(LIBS)technology has been attracting attention since its inception and has become the first choice for on-line detection of material elements.It has been applied to the smelting,archeology,military,space exploration,chemical analysis and other fields,and achieved good application results.This thesis is based on "innovative multi-functional laser research and development application of spectrum analysis instrument",which is the major national instruments and equipment development projects.Based on the previous research of the research group,this thesis focuses on the adaptive denoising,baseline correction and spectral peak recognition of LIBS spectrum.The main contents of this thesis includes:(1)Most of the current LIBS spectral denoising algorithms require human intervention,aiming at the problem,an improved denoising algorithm based on three-point zero-order Savitzky-Golay is proposed.The algorithm based on the idea of the three-points zero-order Savitzky-Golay filter,the fitting order is improved as an adjustable parameter,and the LIBS spectral data is denoised with the minimum initial points and order iteration.At the same time,the eigenvalue extraction strategy is studied to realize the baseline correction of spectral data.The experimental results show that the improved three-point zero-order Savitzky-Golay filter algorithm can realize the adaptive denoising of LIBS spectrum and the effect is good.The eigenvalue extraction strategy also shows the feasibility when the spectral baseline is corrected.(2)For the problems that the spectral peak recognition algorithm for LIBS spectrum is less efficient and requires a lot of preprocessing,a bi-scale motion spectrum peak recognition algorithm based on Voigt line is proposed.The algorithm is used Voigt line as reference peak type,which can fit the reference peak shape of LIBS spectrum better,and realize the spectral peak recognition of LIBS spectrum.The experimental results show that the improved spectral peak recognition algorithm has good performance on LIBS spectral peak recognition,which reflects the effectiveness and feasibility of the algorithm.(3)The preprocessing software system of LIBS spectral data is designed and implemented as an integral part of metallurgical-specific LIBS analysis system,and the corresponding software authorization is obtained.