Dubai Police General Headquarters, Dubai 1492, United Arab Emirates.
Research and Graduate Department, Rochester Institute of Technology, Dubai 1492, United Arab Emirates.
Molecules. 2022 Jul 3;27(13):4281. doi: 10.3390/molecules27134281.
Synthetic fibers are one of the most valuable trace lines of evidence that can be found in crime scenes. When textile fibers are analyzed properly, they can help in finding a linkage between suspect, victim, and the scene of the crime. Various analytical techniques are used in the examination of samples to determine relationships between different fabric fragments. In this exploratory study, multivariate statistical methods were investigated in combination with machine learning classification models as a method for classifying 138 synthetic textile fibers using Fourier transform infrared spectroscopy, FT-IR. The data were first subjected to preprocessing techniques including the Savitzky-Golay first derivative method and Standard Normal Variate (SNV) method to smooth the spectra and minimize the scattering effects. Principal Component Analysis (PCA) was built to observe unique patterns and to cluster the samples. The classification model in this study, Soft Independent Modeling by Class Analogy (SIMCA), showed correct classification and separation distances between the analyzed synthetic fiber types. At a significance level of 5%, 97.1% of test samples were correctly classified.
合成纤维是犯罪现场最有价值的微量物证线索之一。当对纺织纤维进行适当分析时，它们可以帮助在嫌疑犯、受害者和犯罪现场之间建立联系。在对样本进行检查时，使用了各种分析技术来确定不同织物碎片之间的关系。在这项探索性研究中，研究人员结合了多元统计方法和机器学习分类模型，使用傅里叶变换红外光谱（FT-IR）对 138 种合成纺织纤维进行分类。首先对数据进行预处理技术，包括 Savitzky-Golay 一阶导数法和标准正态变量（SNV）法，以平滑光谱并最小化散射效应。然后构建主成分分析（PCA）以观察独特的模式并对样品进行聚类。本研究中的分类模型——软独立建模分类类比（SIMCA），显示出对分析的合成纤维类型的正确分类和分离距离。在 5%的显著水平下，97.1%的测试样本被正确分类。
