拉曼光谱结合机器学习对面巾纸类物证的分类研究
Research on The Classification of Facial Tissue Evidence Based on Raman Spectroscopy and Machine Learning
-
摘要: 为实现面巾纸物证的快速检验,采集了 60 个不同品牌和产地的样品的拉曼光谱数据,并对其进行分析.对60 个样品的拉曼光谱进行基线校准等方法处理后,对光谱数据进行两步聚类法进行分类和特征提取,再使用K-近邻算法和径向基函数进行建模分析,K-近邻算法的最终判别正确率为 94.4%,而径向基函数的最终判别正确率为100%,取得了较好的分类结果.通过对样品的分类判别,可以依据样品的有机和无机成分含量差异进行更细化的分类和检验,较好地契合了数据集中不同样品的性质差异,能够为面巾纸类物证的检验提供一定的参考.Abstract: In order to realize the rapid examination of facial tissue evidence,the Raman spectral data of 60 different brands and origin samples were collected and analyzed.After the Raman spectra of 60 samples were processed by baseline calibration and other methods,the spectral data were classified and feature extracted by two-step clustering method,and then the K-nearest neighbor algorithm and radial basis function were used for modeling analysis.The final discrimination accuracy of the K-nearest neighbor algorithm was 94.4%,while the final discrimination accuracy of the radial basis function was 100%,and a good classification result was obtained.Through the classification and discrimination of the samples,more detailed classification and testing can be carried out according to the difference in the content of organic and inorganic components of the samples,which better fits the difference in the nature of different samples in the data set,and can provide a certain reference for the inspection of facial tissue material evidence.