Database Reference
In-Depth Information
method for improving the accuracy of the correlator, which is based on the selection
of effective correlation areas (ECAs), preventing bad areas from being used for
correlation purposes. Bad areas are areas of the ballistic image that do not contain
useful, reproduced toolmarks. Generally, these areas are consciously ignored by a
firearm examiner during his visual checking. But including data located in bad areas
in the correlation results in a lower correlation score, and can seriously compromise
the performance of an automated system.
Given that the cross-correlation function is important for the ballistic image
comparison as demonstrated by Chu et al. [ 167 ], Vorburger et al. [ 168 , 169 ] and
Weller et al. [ 170 ], in the current work, a method is presented for improving the
discriminating power of the correlator using automatic image registration and the
characterization of the ECA. The proposal for the adoption of automatic registration
takes into account the translation of images in all three axes, as well as a scaling
change for the proper alignment of the compared images. Unlike the works of
Vorburger et al. [ 168 , 169 ], in which the image correlation is obtained right after
registration, the current work extracts the ECA before obtaining the similarity score.
This increases the score value of the potential match, as well as the discriminating
capability of the system, increasing the accuracy of firearm identification.
6.1.1
Firearm Identification Procedure
Traditionally, investigators identify the ballistic specimen from a crime scene with
a reference specimen, by mapping the marks with visual images from two cameras.
Images of the cartridge case of the firearm are captured using a personal computer
and frame-grabber in conjunction with 40
microscopes fitted with two color
charge-coupled device (CCD) cameras (Fig. 6.1 ). A ring light source is adopted,
which can provide uniform lighting conditions. In order to match both specimens,
a number of features within the identifying marks are chosen for their apparent
uniqueness. A decision is made as to whether the same firearm made the marks
under examination at the crime scene and in reference specimens. The critical step
here is the selection of the mark or set of marks for examination and comparison.
Figure 6.2 shows some examples of matching marks obtained by the system.
In comparison, Fig. 6.3 shows the automatic system for matching cartridge base
case images. Images of the reference specimens are stored in the database, and each
of them is compared to the image of the ballistic specimen from the crime scene.
The identification process consists of the following three steps: L2-norm energy
normalization, image registration, and ECAs-based image matching.
The quality of the captured image of a cartridge specimen can be affected by
noise and lighting conditions. Poor lighting conditions and wrong setting of the lens
aperture during image acquisition can result in a low-contrast image. Thus contrast-
enhancement transformation is used when the image is obtained by the system.
×
Search WWH ::




Custom Search