Image Processing using Color Space Models for Forensic Fiber Detection

摘要

The purpose of this study is to investigate the feasibility of automating fiber analysis in forensic science applications. In order to make self-directed collection of spectral data possible the number of measuring locations needs to be restricted to fibers only. Full scans of the samples would result in very large amounts of data of which only a small part carries actual information about the objects of interest. Images obtained by optical microscopes are used for preprocessing to find suitable candidates for measuring locations that subsequently may be used to control the microscope stage for spectroscopic measurements. This paper presents a method based on a nonlinear transform known to enhance the contrast in a way that makes segmentation on grayscale images possible. It further introduces an approach using the differences between the color channels of RGB images combined with common morphological operators to segment color images. A third application is presented that enables the search for fibers matching a query object in color based attributes, for which it is necessary to consider multiple color models. This approach reduces time and efforts significantly when trying to match one specific fiber to other samples.