Abstract
Chemical imaging is a new latent fingerprint examination
technique that combines molecular spectroscopy and digital
imaging technology. Chemical imaging, employing luminescence
and visible absorbance, has been successfully applied
to various treated and untreated fingerprint samples,
demonstrating the usefulness of this technology to aid
routine forensic latent fingerprint development. This
validated technique exhibits improved detection limits
over conventional approaches. Chemical imaging has also
been used to demonstrate increased contrast of fingerprints
developed on difficult backgrounds such as fluorescent,
dark, and rough substrates and multicolored surfaces.
Chemical imaging is a viable strategy for detecting
the most challenging latent fingerprints when standard
development methods fail.
Introduction
Chemical imaging combines molecular,
spectroscopic, and digital imaging information by recording
images of the sample as a function of wavelength through
the use of an efficient electro-optic imaging spectrometer.
The electro-optic imaging spectrometer combines an efficient
electronic filter system with no moveable parts and
a slow-scan CCD detector. Samples are illuminated using
a standard variable wavelength light excitation source
(SPEX Forensics, Edison, New Jersey) followed by image
data collection through the imaging spectrometer at
preselected wavelength increments. The resulting images
are combined to create a multidimensional image data
set. A fully resolved spectrum is recorded for each
pixel location in the image where each spectrum can
provide information to better describe the sample of
interest (Morris et al. 1994; Morris et al. 1996). Contrast
in the resulting chemical images arises from the varying
amounts of absorption, emission, or scatter that occur
in the measured spectrum at each image pixel. As a result,
chemical images provide molecular, compositional, structural,
or quantitative information about the sample of interest
in addition to generating enhanced image contrast.
Conventional fingerprint imaging systems use standard
variable wavelength light excitation followed by data
collection at one specific color, often employing a
single-barrier optical filter configuration. As a result,
fingerprint detection on complex substances, including
paper, curved surfaces, or dark objects can be challenging.
Chemical imaging separates an image into its component
colors in a quantitative manner at many different wavelengths.
Many more resulting wavelengths are recorded than conventional
red, green, and blue color imaging. Typically, hundreds
to thousands of colors can be accessed using a novel
electronic imaging spectrometer. This data enables the
examiner to discern usable information from a background
on a pixel-by-pixel basis. Unwanted background including
fluorescence, texture, and colors can be efficiently
minimized, effectively revealing the detail of the fingerprint
pattern. The enhanced sensitivity of chemical imaging
has been demonstrated for fingerprint examination that
exceeds the capabilities of conventional imaging techniques
(Exline et al. in press; Wallace 2001).
To collect chemical images, the imaging spectrometer
is computer-controlled; hence, the parameters for a
particular series of experiments need only be set up
once and can be automated, which is a distinct advantage.
Also, conventional luminescence imaging systems require
the use of suitable barrier filters that block the reflected
excitation light and only transmit the weak fingerprint
emission. With chemical imaging, the imaging spectrometer
acts as the barrier filter that eliminates the need
for additional filter optics. Another advantage of chemical
imaging is that, with limited knowledge of a particular
fingerprints absorption or luminescence and despite
the presence of background interference from the substrate,
the system can be configured to analyze the fingerprint
emission over a wide spectral range. Software that locates
and isolates the maximum absorbance or emission of a
treated fingerprint, thereby optimizing image contrast
can then be used. Increased contrast in the imaging
data is also enhanced by reducing background signal
and revealing the fingerprint signal through the use
of robust, well-tested, and validated multivariate statistical
analysis tools.
Methodology
and Results
The CONDOR Macroscopic Chemical
Imaging System (ChemImage Corporation, Pittsburgh, Pennsylvania)
was equipped with a visible wavelength range electro-optic
imaging spectrometer and a front illuminated 1024x1024
slow-scan CCD detector on a macroscopic imaging platform.
The system provides 1,048,576 spatially resolved spectra
for each data set collected. The 16:1 visible macro
optics enabled images to be collected at fields of view
ranging from 1.68 to 108mm with 25-400µm spatial
resolutions, respectively, in a 16-bit digital image
format. The excitation source was a halide arc lamp
used in combination with a range of excitation filters.
A major difference between chemical imaging and conventional
methods of latent fingerprint detection is the use of
a liquid crystal-based electro-optic imaging spectrometer.
Wavelength-resolved images could be collected from 400
to 720nm at sub-nm tuning increments, which provides
flexible sampling of the optical wavelengths for generating
fingerprint contrast. Typical operation of the system
does not require sampling at sub-nm increments.
Figure
1. |
|
|
|
Figure
1A.
Untreated latent fingerprint on a paper surface
using conventional 35mm photography. |
|
Figure
1B.
The same fingerprint developed using visible absorption
chemical imaging followed by substrate division
to correct for background effects. |
Figure
1 shows reflectance chemical images of an untreated
latent fingerprint on white paper captured over 420
to 720nm at 10nm increments. Figure 1A shows a color
image of the latent fingerprint. Figure 1B shows a visible
reflectance chemical image of the fingerprint. To produce
Figure 1B, ChemAcquire 6.0 software (ChemImage
Corporation, Pittsburgh, Pennsylvania) was used to collect
the chemical image and a background image in a clear
region of the paper using identical collection parameters.
The original chemical image was divided by the background
chemical image to correct for several background effects,
including illumination light source variation, substrate
reflectance, and instrument response.
Chemical image analysis is provided by ChemAnalyze
6.0 software (ChemImage Corporation, Pittsburgh, Pennsylvania).
Contrast is generated in the images based on the relative
amounts of light that are produced by the different
species located throughout the sample. Since a spectrum
is generated for each pixel location, chemometric analysis
tools such as principal component analysis (Wold et
al. 1987) and multivariate curve resolution (Andrew
and Hancewicz 1998) can be applied to the image data
to extract pertinent information otherwise missed by
ordinary univariate (single wavelength) measures.
Figure 2 shows analysis of a fingerprint treated with
physical developer on counterfeit U.S. currency using
luminescence chemical imaging. An excitation filter
at 575nm was employed while tuning the imaging spectrometer
from 580 to 720nm at 5nm increments. The analysis involved
dividing the latent fingerprint chemical images by a
background image to ratio out the substrate emission.
This procedure was accomplished by selecting an average
of pixels in a region between the existing ridge patterns
defining this averaged spectrum as background and dividing
all pixels in the image by the background spectrum.
Subsequently, each pixel in the resulting chemical image
was subtracted by a global minimum value to reduce offset,
and then a vector normalization procedure was performed.
Vector normalization involves dividing each pixel spectrum
by the square root of the sum of the squares of all
the pixel spectra, which has the effect of bringing
intense image features on approximately the same scale
as weak image features. Principal component analysis
was then applied to the normalized data to produce fingerprint
images for the light background (Figure 2B) and the
dark background (Figure 2C). Figure 2D is produced by
averaging the principal component analysis extract images
in ChemAnalyze 6.0.
Figure
3. |
Figure
3A.
Optical image of a blue drug bag. |
|
|
|
Figure
3B.
Visible-absorption
chemical imaging of a ninhydrin-treated fingerprint
following multivariate statistical analysis. |
|
Figure
3C.
Digital
image of a ninhydrin-treated fingerprint following
conventional digital photography and image processing.
|
|
Images
are reprinted with permission of the Allegheny County
Coroners Office,
Forensic Laboratory Division, Pittsburgh, Pennsylvania. |
Figure 3 shows a latent fingerprint present on a drug
bag (Figure 3A) treated with ninhydrin. The fingerprint
was examined and chemically imaged from the 420 to 720nm
range in 10nm increments using white light excitation.
Background correction, offset correction, and normalization
procedures comparable to those applied in Figure 2 were
employed to produce Figure 3. Principal component analysis
was then applied to the normalized data for visualization
of the fingerprint (Figure 3B). The same ninhydrin-treated
fingerprint was photographed using a digital camera
and processed using More Hits (PC Professionals,
Inc., Lakewood, Washington) image-enhancement software
(Figure 3C).
Why is Chemical
Imaging Important?
The successful analysis of untreated
latent fingerprints on paper surfaces by chemical imaging
shows immense promise, given that ridge detail can be
detected on fresh fingerprints using chemical imaging
followed by ChemAnalyze software analysis. This
is an area for continued studies, because a nondestructive
optical method for detecting untreated prints would
be of significant benefit. Results presented here demonstrate
the enhanced sensitivity of chemical imaging for latent
fingerprint examination that exceeds the capabilities
of conventional imaging techniques.
Visualization of the fingerprint on the counterfeit
$10 bill demonstrates the ability of chemical imaging
to generate contrast in the presence of complex substrates
by using the optical properties of the substrate as
a digital signature of the background. Whereas the fingerprint
was nonluminescent, dividing the background spectrum
from each pixel in the chemical image and applying robust
chemometric routines could still develop fingerprint
contrast. As a result, the visualization of the fingerprint
was acquired in seconds, where previous efforts using
conventional means have been unsuccessful. Once chemical
imaging revealed the optimal detection strategy, examiners
were able to replicate this study using a VSC 2000
system (Foster and Freeman, Worcestershire, United Kingdom).
The ninhydrin-treated fingerprint was developed using
chemical imaging and also by conventional digital photography.
The chemical image result was significantly better as
demonstrated by enhanced ridge detail when compared
to the results produced by conventional methods. This
comparison exemplifies the value of chemical imaging
to real-world case samples and its value as an enhanced
detection strategy.
Conclusions
In recent years, new technology has been
sought for more rapid examination of forensic evidence,
as well as for field use. This includes the need for
improved fingerprint visualization methods. A new class
of technology based on chemical (i.e., spectroscopic)
imaging has demonstrated the ability to provide substantial
improvements in detection capability. Chemical imaging
techniques integrate digital imaging with a number of
proven optical inspection analytical methods, including
fluorescence imaging spectroscopy and visible/NIR reflectance
spectroscopy. Chemical imaging is consistently being
proven to be a valuable tool for forensic science, by
enabling examiners to visualize and identify evidence
with improved detection limits.
This overview has demonstrated the application of chemical
imaging for the detection of untreated and chemically
treated latent fingerprints. Current studies are ongoing
to further validate and optimize the technique for a
wider range of fingerprint detection methods and for
latent fingerprints on a wider range of substrates.
Acknowledgments
The
authors would like to thank Christie Wallace and Claude
Roux from the University of Technology, Sydney, Australia,
and Chris Lennard from the Australian Federal Police,
Canberra, Australia, for their valuable contributions
in implementing this technology. We would also like
to thank Dr. Antonio A. Cantu from the U.S. Secret Service,
Washington, DC, for his insights into the applications
of this technology. Lastly, we would like to thank Wayne
Reutzel and Mike Fedor from the Allegheny County Crime
Laboratory, Pittsburgh, Pennsylvania, for their support
and assistance in introducing this technology to practical
casework methodologies.
References
Andrew, J. J. and Hancewicz, T. M. Rapid
analysis of Raman image data using two-way multivariate
curve resolution, Applied Spectroscopy (1998)
52:797-807.
Exline, D. L., Wallace, C., Roux, C., Lennard, C., Nelson,
M. P., and Treado, P. J. Forensic applications of chemical
imaging: Latent fingerprint detection using visible
absorption and luminescence, Journal of Forensic
Sciences (in press).
Morris, H. R., Hoyt, C. C., Miller, P., and Treado,
P. Liquid crystal tunable filter Raman chemical imaging,
Applied Spectroscopy (1996) 50:805-811.
Morris, H. R., Hoyt, C. C., and Treado, P. Imaging spectrometers
for fluorescence microscopy: Acousto-optic and liquid
crystal tunable filters, Applied Spectroscopy
(1994) 48:857-866.
Wallace, C. Applications of chemical imaging to forensic
science. Bachelor of Science (Honors) thesis, University
of Technology, Sydney, Australia, 2001.
Wold, S., Esbensen, K., and Geladi, P. Principal component
analysis, Chemometrics and Intelligent Laboratory
Systems (1987) 2:37-52.
|