Application Research Of Classifier Ensemble On LIBS Spectral Data

Laser induced breakdown spectroscopy (LIBS) is a kind of frontier material and element analysis technology with the advantages of rapid, real-time analysis, multi-element detection at the same time, and as a result it can obtain the information of composition and content of the unknown sample. Consequently LIBS technique has a wide range of applications in materials, metallurgy, archaeology, space exploration, military and other fields; it has become a hot research topic. How to play advantages of LIBS technology and realize its application value becomes a key link.In this paper, the research based on national major equipment development projects---"innovative multi-functional laser research and development and application of spectrum analysis instrument", the thesis aimed at the LIBS spectral data classification algorithm with extensive, in-depth and systematic research, and developed LIBS analysis software of metallurgy with independent intellectual property rights.On the basis of previous research studies in our team, it is how to improve the accuracy and stability of classification algorithm that is the key of this article, the main research work includes:(1) In the light of the problem of the accuracy and stability of the classification algorithm, and the limitations of single algorithm on the multiple classification problems, Rotation Forest(ROF) classification algorithm was introduced to build the combination classification model, improving the classification accuracy, and the robust principal component analysis(RPCA) algorithm of spectral data was introduced to accelerate the algorithm modeling process; (2) For the high dimension characteristic of the LIBS spectra, a new ensemble method is proposed that combines two different classifier models via a transformation of rotation forest, named by isomerous multiple classifier ensemble algorithm. The classifier ensemble model based on partial least squares discriminant analysis (PLS-DA) and soft independent modeling of class analogy (SIMCA) to generate the base classifier, the proposed method achieves the improvement of accuracy by combining two different base classifier; (3) Designed and developed LIBS analysis software of metallurgy, implementing the metallurgical samples' fast discriminant analysis.Experiment verifies the feasibility and effectiveness of the algorithm in the paper.