Computer-aided fiber analysis for crime scene forensics

The forensic analysis of fibers is currently completely manual and therefore time consuming. The automation of analysis steps can significantly support forensic experts and reduce the time, required for the investigation. Moreover, a subjective expert belief is extended by objective machine estimation. This work proposes the pattern recognition pipeline containing the digital acquisition of a fiber media, the pre-processing for fiber segmentation, and the extraction of the distinctive characteristics of fibers. Currently, basic geometrical features like width, height, area of optically dominant fibers are investigated. In order to support the automatic classification of fibers, supervised machine learning algorithms are evaluated. The experimental setup includes a car seat and two pieces clothing of a different fabric. As preliminary work, acrylic as synthetic and sheep wool as natural fiber are chosen to be classified. While sitting on the seat, a test person leaves textile fibers. The test aims at automatic distinguishing of clothes through the fiber traces gained from the seat with the help of adhesive tape. The digitalization of fiber samples is provided by a contactless chromatic white light sensor. First test results showed, that two optically very different fibers can be properly assigned to their corresponding fiber type. The best classifier achieves an accuracy of 75 percent correctly classified samples for our suggested features.

[1]  K G Wiggins,et al.  The European Fibres Group (EFG) 1993–2002: "Understanding and improving the evidential value of fibres" , 2003, Analytical and bioanalytical chemistry.

[2]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[3]  Edward G. Bartick,et al.  Forensic Discrimination of Dyed Textile Fibers using UV-VIS and Fluorescence Microspectrophotometry , 2004 .

[4]  K W Smalldon,et al.  The transfer of fibres between clothing materials during simulated contacts and their persistence during wear. Part I--Fibre transference. , 1975, Journal - Forensic Science Society.

[5]  A. J. Wilkinson,et al.  Forensic Textile Fibre Matching by Multiresolution Planning and Colour Space Pattern Recognition , 1988, Alvey Vision Conference.

[6]  Yan Ma,et al.  New Algorithms for Complex Fiber Image Recognition , 2009, AICI.

[7]  Ian H. Witten,et al.  The WEKA data mining software: an update , 2009, SKDD.

[8]  Ashraf Mozayani,et al.  The forensic laboratory handbook : procedures and practice , 2011 .

[9]  Jorge Herbert de Lira,et al.  Two-Dimensional Signal and Image Processing , 1989 .

[10]  David W. Schiering,et al.  The identification of fibers by infrared and Raman microspectroscopy , 1986 .

[11]  Claus Vielhauer.  Biometric User Authentication for it Security - From Fundamentals to Handwriting , 2006, Advances in Information Security.

[12]  N. Otsu.  A threshold selection method from gray level histograms , 1979 .

[13]  Scott F. Stoeffler.  A Flowchart System for the Identification of Common Synthetic Fibers by Polarized Light Microscopy , 1996 .