摘要
为实现对现场车用保险杠物证快速无损准确检验鉴定,提出了一种结合牛顿插值多项式、Savitzky—Golay平滑和Fisher判别分析的车用保险杠鉴别方法。实验采集并获取奥迪等5种品牌样本的红外光谱数据,考察了牛顿插值多项式、Savitzky—Golay平滑两种预处理方法对样本分类结果的影响,同时借助判别分析构建分类模型。结果表明,4次牛顿插值多项式结合Savitzky—Golay平滑3次多项式和33点平滑点数为最佳预处理方法,其能够有效削弱噪声和干扰信息的负面影响,同时提升模型的分类准确率。在Fisher判别分析模型中,5类样本均实现了100%的准确区分和归类,实验结果较为理想。基于Fisher判别分析的识别模型具有较高的预测精度,可作为一种可靠的保险杠物证鉴别方法,该方法快速、无损、准确,具有一定的普适性和借鉴意义。
In order to realize fast,non-destructive and accurate identification,a method,which combined Newton interpolation polynomial,Savitzky-golay smoothing and discriminant analysis,was proposed.The infrared spectrum data of 5 brands including Audi were collected and obtained,Newton interpolation polynomial and Savitzky-golay smoothing were adopted in the pretreatment and discriminant analysis was used to construct the classification model.The results showed that the 4 Newton interpolation polynomials,the 3 Savitzky-golay smoothing polynomials and the 33 Savitzky-golay smoothing points could effectively reduce the negative effects of noise and interference information and improve the classification accuracy of the model.In discriminant analysis,all samples were identified exactly,which is fast,lossless,accurate and has certain universality and reference.
作者
朱密
朱洪建
ZHU Mi;ZHU Hong-jian(Department of Forensic Science and Technology,Hunan Police College,Changsha 410138,China;Yuelu branch of Changsha Public Security Bureau,Changsha 410000,China)
出处
《化学研究与应用》
CAS
CSCD
北大核心
2020年第9期1541-1546,共6页
Chemical Research and Application
关键词
红外光谱
化学模式
保险杠
鉴别
infrared spectroscopy
chemical pattern recognition
bumper
identification
