基于PCA⁃LDA的车用保险杠显微激光拉曼光谱模式分类

为实现对案发现场车用保险杠物证快速、无损、准确的分类与识别，提出了一种显微激光拉曼光谱分析技术结合多元建模用于车用保险杠模式分类方法。选择自动基线校正、峰面积归一化、Savitzky-Golay平滑（3次多项式，7点平滑）作为预处理方法，借助主成分分析和线性判别分析构建分类模型。结果表明，前27个主成分下，除了奥迪品牌的2个样本被误判在了广汽品牌的样本当中，其他不同品牌的样本均实现了100.00 %的准确区分，总体分类准确率为95.24 %，分类效果较为理想；针对实际案件中的未知样本，借助该方法确定其属于别克品牌，这与实际案件中物证信息相吻合；利用显微激光拉曼光谱分析技术多元建模分析可实现对不同品牌保险杠样本准确的识别与分类，可为微量物证鉴定方面的相关研究提供一定的思路与参考。

Classification of Vehicle Bumpers Through Micro-laser Raman Spectroscopy Based on PCA-LDA

The method based on micro-laser Raman spectroscopy and multivariate modeling was proposed to realize the fast,non-destructive and accurate classification of vehicle bumpers in the forensic science.The automatic baseline correction,peak area normalization,and Savitzky-Golay smoothing(3rd degree polynomial,7-point smoothing)were selected for pre-processing.A classification model was constructed through principal component analysis and linear discriminant analysis.The results indicated that the overall classification accuracy rate was 95.24%as an ideal result.For the unknown sample in the actual case,this method was used to determine the brand as Buick,which was consistent with the real result.A micro-laser Raman spectroscopic analysis technology and multivariate modeling methods can realize the classification of different brands of vehicle bumpers exactly.This can provide a certain reference for the relevant research in the trace evidence identification.