基于向量空间模型的岩屑LIBS光谱分类识别方法

向量空间模型最初用于文献检索，该模型是通过对文献内容进行特征文本提取后，将文献转换到文本向量空间，然后在文本向量空间中通过计算文献的特征文本向量与检索文本的特征文本向量的相似度，实现文献的检索，该方法基于模式识别中模板匹配的最近邻原则。针对光谱数据的特点和模式识别中模板匹配的基本原则，将向量空间模型引入基于样品光谱的分类识别。通过训练集中光谱数据获得各样品的光谱数据模板，提取训练集中各样品光谱数据模板特征峰的波长和相对强度信息，构建特征峰信息数据库，计算获得特征峰信息权值，将光谱数据转换到特征峰向量空间，获得各样品光谱数据模板的特征峰向量，构建样品特征峰向量数据库。同理获得预测集样品光谱的特征峰向量，在特征峰向量空间中通过计算预测集样品特征峰向量与样品特征峰向量数据库中各样品模板特征峰向量的余弦值，完成对预测集样品的分类识别。以岩屑样品的LIBS光谱为研究对象，将向量空间模型应用于LIBS光谱的分类识别。分类结果表明，该方法能够实现对岩屑样品LIBS全谱的快速分类识别，且在对预测集光谱数据进行平均处理后，分类准确率为100%。提出的基于向量空间模型的LIBS光谱分类方法可以拓展应用于其他光谱数据的分类识别。

Study on Identification Method Based on Vector Space Model for Geological Cuttings Using Laser-Induced Breakdown Spectroscopy

Vector Space Model (VSM)was originally used in document retrieval.It is characterized by extracting the texts from the document and converting the document into a text vector space.VSM compares the similarity between document text vectors and text retrieval text vectors.The document retrieval is accomplished according to the similarity based on the nearest template principle of template matching in pattern recognition.This article applied such principle into the identification of samples based on the characteristic of LIBS spectrum.To create the database of characteristic peaks of training dataset,we obtained the spectra template of all kinds of samples from the training dataset and extracted the wavelength and intensity of peaks from the spectra template.In another hand,to create the database of characteristic peak vectors for all the training samples,we calculated the spectral characteristic peak weight and converted the spectra to peak vector space.The characteristic peak vector of the test data-set was obtained by the same way.The cosine value between the characteristic peak vector of test data and every characteristic peak vector in the database were calculated and the maximum cosine was taken as the identification result.Geological cuttings, the research subj ect,were identified by the VSM in this paper.The result demonstrated that the VSM could rapidly identify the spectra from 4 kinds of geological cuttings'LIBS spectral and the correct identification rate was 100% after the spectra of the test dataset were averaged.The proposed LIBS spectral identification method based on VSM can be expanded to the identification of other spectral data.