摘要
汽车灯罩碎片是交通肇事案件现场经常出现的物证。为了实现对汽车灯罩物证的准确检验,该文提出一种将原始光谱与导数光谱相结合的光谱融合技术。收集不同类别和多种品牌的汽车灯罩共计44个,采用傅里叶变换红外光谱技术对样本进行分析,提取其原始光谱数据和一阶导数光谱数据,并结合化学计量学构建分类模型。在对汽车灯罩类别进行分类的Fisher判别分析模型中,单独的原始光谱数据和一阶导数光谱数据的分类准确率分别为86.40%和84.10%,融合后的光谱数据分类准确率达到93.20%,分类准确率明显提高。通过主成分分析优化模型后,融合光谱的分类准确率达到97.70%,且在进一步对汽车灯罩品牌进行分类时,分类准确率达到100.00%,实验结果理想。而在K近邻算法模型中,由于受到样本不均匀的影响,分类准确率较低。结果表明,基于原始光谱与导数光谱的光谱融合技术能够实现对汽车灯罩样本的准确分类,可以为光谱融合技术在分析检测领域的应用提供借鉴和参考。
The car lampshade fragment is the physical evidence which often appears at the traffic accident case scene.In order to realize the accurate examination on the physical evidence of automobile lampshade,a spectral fusion technique combining the original spectrum and derivative spectrum was proposed.A total of 44 lampshades of different categories and brands were collected.The samples were analyzed by Fourier transform infrared spectroscopy(FTIR)to extract the original spectral data and first-order derivative spectral data,which were used to construct the classification model by combining with chemometrics.In Fisher discriminant analysis model,the classification accuracies for single original spectral data and first-derivative spectral data were 86.40%and 84.10%,respectively,while the classification accuracy for fused spectral data reached up to 93.20%,which was significantly higher than those for original spectral data and first-derivative spectral data.After the model optimization of principal component analysis,the classification accuracy for fusion spectrum reached up to 97.70%.In addition,when further classifying the lamp shade brands,the classification accuracy reached to 100.00%.The result of the experiment was ideal.However,in the K-nearest neighbor algorithm model,the classification accuracy was low due to the influence of uneven samples.Results showed that the spectral fusion technique based on the original spectrum and derivative spectrum could realize the accurate classification of automobile lampshade samples,which could provide a reference for the application of spectral fusion technique in the field of public security.
作者
卫辰洁
王继芬
曾啸虎
WEI Chen-jie;WANG Ji-fen;ZENG Xiao-hu(School of Criminal Investigation,People's Public Security University of China,Beijing 102600,China;Jiuquan Satellite Launch Center,Jiuquan 735000,China)
出处
《分析测试学报》
CAS
CSCD
北大核心
2021年第7期1043-1048,共6页
Journal of Instrumental Analysis
基金
中央高校基本科研业务费专项资金资助(2021JKF208)。
关键词
汽车灯罩
傅里叶变换红外光谱
光谱融合
K近邻算法
FISHER判别分析
automobile lampshade
Fourier transform infrared spectroscopy
spectral fusion
K-nearest neighbor algorithm
Fisher discriminant analysis
