Hyperspectrum Imaging and Processing in Forensic Document Analysis.
Hyper-spectrum image is 4D data array that consist of optical spectra for each
point of the image. Technically the Hyperspectrum can be measured either by
scanning each point of an object with spectrometer or by recording series of
the images taken at narrow spectral band. In case of forensic document examination
the hyper-spectrum is reflection (or/and luminescence) spectral data set measured
at each point of a questioned document. The hyperspectrum processed with our
advanced software reveals much more information about the image as compared
to visual spectral comparison (VSC) method or single point spectrum measurement.
For example, if one is to compare (impartially detect) different inks in one
document, then processing is done such as to reveal unique spectral signatures
in each point of the document. Each ink has unique spectral signature. This
signature is different even for same ink but for old drawn and fresh drawn lines,
because inks spectral features are subject of slow temporal changes. Dedicated
hyperspectral data processor can uncover those signatures and bring the result
into easy understandable form of the 2D / 3D image. Several patent pending spectral
signature unwrapping algorithms were developed as a result of cooperation between
ScienceGL and UEKAE. This is basic operation principle of ForensicXP and MST-2
hyperspectral instruments. Hyperspectral imaging was found to be the most powefull
tool to view obliterated writings such as graphite, printer, ink and mixed obliterations.
One might ask "When obliteration detected is it possible to distinguish
which line was drawn the first or, in other words, is it possible to solve classical
forensic "line sequence" problem with hyper-spectral approach. This
more complicated part of the forensic examiner job stars from this point. As
our research has shown in recent few years the answer is "yes", provided
that high-resolution and sensitive enough data acquisition hardware is used
for hyper-spectrum measuring and that the hyperspectral data is processed with
adequate smart computing spectral signature unwrap algorithm. Independently
similar research results were reported by Oak Ridge Institute for Science and
Education- FBI Laboratory Research Division. See citation and link below [1].
Advantages
1. Nondestructive new principle in document examination that utilizes physically
measurable (unprejudiced, i.e. true) spectral data analysis principles
2. Effective 16 bits hyper-spectral data acquisition per , that is 16 bit per
hyper-spectrum point, and effective 64 bits per pixel.
3. High spatial resolution with 1.5 Mega pixel camera
4. Highest commercially available sensitivity / resolution of the hyper-spectral
imaging
5. Automatic hand free hyper-spectrum data recording from questioned document
6. Automatic data processing with patent pending algorithms of spectral signatures
reveal
7. Patent pending ink spectral signature unwrap algorithm of highest possible
sensitivity
8. 2D and 3D result visualization and reporting for maximum human perception
9. Line sequence drawing problem solution in about 95% of the cases
10.
Forensic 3D digital image that resulted from questioned document measured with
ForensicXP and processed with 4D hyper-spectrum processor. Obliterated writing
is unambiguously detected.
Reference
[1] Hina Ayub, Diane Williams
"The Role of Hyperspectral Imaging in the
Visualization of Obliterated Writings"
2006 APS March Meeting, March 1317, 2006;
Baltimore, MD, USA
Citation "We report the use of hyperspectral
imaging to successfully view
obliterated
writings in which a ``true black'' ink obliterated graphite
as well as graphite/graphite and ink/ink obliterations"
http://meetings.aps.org/Meeting/MAR06/Event/43696.
3D
Graphics Home
3D
Forensic Image Tutor
3D art ActiveX GIS map Forensic 4D volume Laser Multilayer CCD IR images Resources 1 2 3 4 Bio Imaging Hybrid AFM | |
Europe FringeCatch AFM Products | |
Address: 140 North Main St., Suite 2A Attleboro, MA, 02703 / Phone: (401)-323-0603 (voice) |
-----------------------------------------------
Copyright © 2003-2013 ScienceGL,
Inc.