Forensic
ELSEVIER
Science 82 (1996)
lbrensic Science Intern;ati
International 7-20
What shoe was that? The use of computerised image database to assist in identification Wayne Ashley Crime
Scene Section,
Received
Victoria
12 October
Forensic
Science Centre, Australia
1995; revised
19 January
Forensic
Drive,
1996; accepted
Macleod
1 February
3085, Victoria,
19%
Abstract The use of a computerised (digital image-based) reference system for the capture, storage and retrieval of shoe soles and uppers has the ability to give quick and reliable information to the investigator in relation to the brand and model of shoe responsible for latent impressions located at crime scenes. The success of the system described is due to the diversity and simplicity of the classification codes, coupled with the ability to search part or all of a shoe sole area. This gives the user greater discriminating power and, with the use of icons and ‘click on’ features, makes the system user friendly when classifying and searching for shoes, thus reducing the possibility of interpretation error. Classificationsystem; Shoes; Shoe sole impressions;Crime scenes;Database; Interpretation
Keywords:
1. Introduction
As identified by Hamm [l], ‘shoe impressions or more correctly track impressions have been around for a very long time, well before the identification of fingerprints’. Shoe impressions and fingerprints are but a few of the types of evidence a criminal may leave behind at a crime scene. Such impressions or other evidence have great potential to the investigator. Not only can impressions lead to the identification of the make and style of a particular shoe, they can also provide approximate sizing of that shoe and the direction of gait, and other actions such as dragging or scuffing. The quick ,and timely identification of the type of shoe impression located at a 0379-0738/96/$15.00 @ 1996 Elsevier PZI SO379-0738(96)01962-7
Science
Ireland
Ltd.
All
rights
reserved
8
W. Ashley
I Forensic
Science
International
82 (1996)
7-20
crime scene may assist the police investigator in identifying possible suspects or the linking of possible similar offences. The provision of this valuable evidence to the investigator in the shortest possible time must not only be cost effective but provide sufficient information to warrant further enquiries. Shoe impression or reference files are not new: they are maintained worldwide by such bodies as the FBI, the Metropolitan Police, the National Bureau of Investigation in Finland and the Bundeskriminalamt in Germany. Reference files form an important part of the overall contribution forensic science has to play in the pursuit of justice. Without such assistance, time and resources are expended in compiling information about a single incident or enquiry that at its conclusion may be discarded. With increasing computer power and image recording systems, such reference files can now become more useful in their application. A potential problem is when systems become unwieldy due to their size or the lack of a clear and proper classification system. At a recent FBI International Symposium on Footwear and Tyretrack Evidence [2] held at the Quantico Training Academy in June/July 1994, several different image based computer systems were reported and demonstrated. The current shoe reference system - SHOE@operating in Australia was also demonstrated by the author.
1.1. History
A shoe reference file has operated within the Crime Scene Section of the Victoria Forensic Science Centre (VFSC) since 1982. The Shoe Sole Pattern File (SSPF), as it was known [3], was born out of the need to identify partial shoe impressions at scenes of crime, so that the information obtained could assist in the investigation process and possibly form a nexus between the offender and the scene, and to other scenes. A revised system was introduced in 1984 when the old system of photographs and codes was no longer able to cope with the increase in the diversity of shoe types or varieties and the complexity of their sole patterns [4] at that time. To overcome this, and to facilitate easy searching, a revised system focused on a comprehensive classification system of primary and secondary codes. The primary codes represented the major groups of bars, studs, waves, etc., while the secondary codes classified design features of brand name, motif, numerals and more. The benefit of the system was that the classification codes were computer based and searching was carried out by the computer. The system continued to grow with the classification codes coping with the increase in numbers but the problem remaining was searching the large number of photographs depicting the known shoe sole and upper. As mentioned above, the SSPF system records known shoes and their sole patterns rather than shoes collected from offenders charged with offences or impressions located at scenes of crime. The latter appears to be the main thrust of a number of other shoe reference systems currently being used around the world [2]. This is not to say that the linking of evidence is not carried out, as a secondary
W. Ashley
I Forensic
Science
International
function of the new system allows cross linking patterns recorded at crime scenes.
82 (19%)
7-20
9
with similar offences or shoe sole
1.2. Why classify shoes ?
The main reason for classifying shoes is to form an ordered database which can be quickly and systematically searched and the information given to the police investigator in a timely and reliable manner. The use of computerised databases with the ability to search and identify objects by classification or other codes was the obvious method by which to improve on a manual reference file system. The ability of computer technology to also capture and compress images has assisted in producing a fast and effective tool that can be utilised for a myriad of forensic applications. The classification codes were developed to cover the ever increasing patterns that are available in shoe soles, particularly sporting shoes. There are over one hundred codes with descriptions and icons available so as to minimise any interpretation error during the classification and searching phases (Refer to SHOEview and SHOEadmin). 1.3. Data input
To obtain the maximum information available about a particular style of shoe it was decided to attend the manufacturer/importer of the shoe so as to obtain, (i) an image of the sole and upper of the shoe and (ii) details as to the manufacturing process, numbers produced and released for sale, and associated data relevant to the manufacturer/importer. The process by which the known shoes are recorded is by high quality negative film. The shoe earmarked for retail sale is mounted onto a frame that allows provision for a scale extending the length of the shoe. The scale is to assist with sizing of suspect shoes against the known shoe in the SHOE system. The outsole and the upper of the shoe are then individually photographed. Each photograph shows a full-frame of the outsole and the upper with or without side flash depending on the style, color and contrast of the outsole and upper. Each photograph is coded to cross reference with the make, model and associative data of the manufacturer/importer. Photographic prints measuring 150 X 200 mm are produced of each outsole and upper. These prints are then captured by a digital video camera, compressed and stored onto the SHOE system. Previous attempts at single recording of the outsole and upper of a shoe using a stand with a 45” mirror back did not produce quality images suitable for printing or faxing via the SHOE system. The reason for using photography and digital video capture is the speed with which new images can be captured and then classified onto the SHOE system. The Crime Scene Section of the Victoria Forensic Science Centre is currently trialling the use of negative scanners and digital 35 mm cameras with a view of increasing image quality for future advances in pattern recognition. Currently the system has 90 manufacturers/importers on file. It has been
10
W. Ashley
I Forensic
Science
International
82 (19%)
7-20
established [5] that 95% of shoe impressions located at crime scenes in Victoria were either one or more types of the following three groups - sporting shoes (runners), casual shoes and work-related shoes. Therefore, the reference collection was restricted to these groups of shoes with the system currently totalling 4000 shoes.
2. SHOE In 1992, research funding was made available by the Australian National Institute of Forensic Science (NIFS) to develop a computerised (digital image based), storage, retrieval and searching system that could be utilised on a national basis. The intelligence (classification codes) had already been developed and utilised in the revised system [3]. NEC Information Systems and an associated software company (Vision Control International) assisted with the development of the system to a viable operating product that could also be commercially marketed. Subsequently, NEC and Vision Control were contracted to assist with the further development of the existing system into a digital image capture, storage, and retrieval system. The new SHOE@’ system runs under WindowsTM and was implemented in 1994. The hardware required to run the system is a 486/33 PC with 8+ Mb RAM, 300Mb or more HDD, VGA 1 Mb recommended display, minimum 32K colour and a 17’ monitor. The system runs under Windows 3.1. The Shoe Application Program based around the classification system was devised by Vision Control and comprises two modules - SHOEview and SHOEadmin. Images are stored of the upper and sole of the shoe and compressed to between 13 and 17K of memory with a compression ratio of 25:l. This allows the 4000 images to be stored in 150 Mb of memory. SHOEadmin is designed to run with a Truevision TARGA+ video capture card and with the VGA display card supporting HiColor 32K. The SHOE system information database is stored within a relational SQL database, the specifications of which are Client/Server SQL DB in a standalone or multiuser environment. It meets ANSI SQL specifications and is compatible with IBM SAA standards. It also supports full transaction processing with built in referential and entity security and automatically uses expanded and extended memory to improve performance. 2.1. Classijication
system
Apart from the one hundred classification codes (refer to Appendix l), the shoe sole is divided into four areas - Toe, Ball, Instep and Heel. This division has increased the speed and the ability to locate a known shoe. The searching works on an inclusion/exclusion principle and where there is some doubt as to the
‘@SHOE
is copyright
to Vision
Control
International
W. Ashley
I Forensic
Science
International
82 (19%)
7-20
11
position of a certain feature, the system can make adjustment through the use of icon driven features. The classification codes have been divided into major groups and sub features of those groups so as to assist with ease of use. The major groups are Bars, Circles, Design, Mesh, Pattern, Studs, Waves and Zig Zags. A separate search feature of ‘Brand Name’ has been incorporated so as to provide greater discriminating power or the ability to selectively search known styles within a brand. The same four distinct areas of Toe, Ball, Instep and Heel are used in the classification of a known shoe when a new shoe is entered onto the system (see SHOEadmin). As identified in Data input (Section 1.3), the upper of a known shoe is photographically recorded on film and the subsequent print, digitally video captured onto the SHOE system. Classification codes have also been developed for the upper of the shoe and will form part of future upgrades of the system. Such a classification could assist with identification of a possible shoe, particularly those identified in bank security film or descriptions by eyewitnesses to a crime. To minimise the risk of errors or incorrect information entering the system, most fields have dialog boxes with pull down selected features or drop lists so that data entry is regulated and standardised.
3. SHOEview
SHOEview is the module that searches and matches the shoe to the latent print located at the crime scene. Under the Search Menu, choose Shoe and the user is presented with a Search Shoe dialog (Fig. 1). By choosing the Group category, then selecting a Feature of that Group the user then clicks onto the desired shoe area button (Toe, Ball etc.) and that listing is then identified in the specific listing box. The listbox will identify that feature in ‘text’ and will categorise its search criteria with an icon display. @The Tick indicates the feature MUST be present in this area for the shoe to match. c8 The Question Mark is generally not used as a single entry for a given feature, but for multiple shoe areas for a given feature - that is you don’t know where on the shoe the feature exists, but you know it has this feature somewhere. For example you could place a Question Mark for Studs, Hexagons in the Ball and Instep area, and when searching matching shoes must have Studs, Hexagons in either the Ball or Instep area or both. Note if the Question Mark is used as a single entry then it behaves as a Tick. 8 The Not indicates the feature MUST NOT be present in this area for the shoe to match. The use of the mouse to ‘click on’ to a feature and place that feature onto a specific area of the sole not only increases searching speed, but is a very user frieudly method of performing a search. It is also possible to undertake an exclusive search, such as ‘Brand Name’ only, or utilise a number of other features.
W. Ashley
-
Figure
Dimples MotiF Numerals Separate Heel Stud forming heel Writing
1. Shoe Search
I Forensic
Science
International
82 (1996)
7-20
I
I
Screen.
Fig.
1. Shoe
search
screen.
If the user is unsure as to where a feature appears on a latent shoe print, the feature may be entered in the ‘ALL’ feature bar. By filling in the search criteria, and clicking on the Search button, the user is presented with the number of matches and asked to proceed or return to the search dialog. In electing to proceed, the Shoe Match dialog (see Fig. 2) displays shoes matching the selected search criteria. If there is more than one match to the search criteria (and there usually is), they can be viewed using the Next/Previous buttons. The Shoe Match dialog also displays details about the shoe, including manufacturer/importer, brand, style, process, etc and the image as either a sole, upper or both.
3.1. Wanted file
When a shoe from the database is matched to a latent impression from a scene, details of the particular offence are entered into the database in a ‘Wanted File’ under the dialog box ‘New Wanted’ (Fig. 3). The user can also search the ‘Wanted File’ to produce a sequential listing of all wanted shoes from crime scenes. The ‘Wanted File’ (Fig. 4) enables links to be established between multiple cases which involve the same style of shoe along with Case, Offence, Complainant and Informant details. A closer examination of the latent images from these cases may reveal corresponding wear patterns or identifying features, thus linking the offences.
W. Ashley
I Forensic
Science
International
82 (1996)
Date Entered Shoe Key :2 Brand
0187
: REEBOK
Style
: BB 5000
Size
: ?
Manufacturer (NEXT Process /CEMENT
: 08/l
13
7-20
: HON
PTY LTD
:
1 of 645
Figure
2. Shoe Match
Screen.
Fig. 2. Shoe
match
screen.
~~
[06/04/94
Suburb
j
Comments
:
Complainant
: I------l
:
II
gum 3. Wanred
Screen
for
New Cases.
Fig. 3. Wanted
screen
for new cases.
1
W. Ashley
14
Old Number Style
Date Entered Case
Number
Crime Scene
Science
: 064001
Brand
I
I Forensic
Date :
: KANGAROOS
/I
: BJH 4420
I
:
11103/68
:
International
I
II~OWBB
7-20
1
Location street
1
3082/878
82 (1996)
Suburb Offsncsltl
: I] : MELBDURNE
1
:
No. :
LEAPRefNo.: Member
:
1 ‘9pgq
II igure 4.
Wanted Screen for Old Cases. Fig. 4. Wanted
3.2. Unidentified
screen
for old cases.
file
Should the search against the database not identify the brand and style of shoe responsible for the impression recovered from the crime scene, then the classification codes are recorded and stored under the ‘Unidentified File’ (Fig. 5) where the latent image from the crime scene is recorded using the SHOEadmin module (refer to Fig. 7). This facility allows shoe searching at a later date, when additional shoes may have been added to the system that might match the latent impression. The Unidentified records are displayed in chronological order with the most recent records displayed first. This dialog box (Fig. 5) also allows the user to search and match or cross reference existing unidentified shoes located at crime scenes. 3.3. Reporting
If a search has matched a crime scene latent impression to a known shoe in the database, then the user has the option to print the shoe sole or upper or both on a formal report in either black and white or colour. The user then has the ability to ‘direct fax’ the result to the investigator’s station, identifying the shoe brand, style, manufacturer details, offence details and similar offences, thus ensuring a quick and effective reporting service. 4. SHOEadmin SHOEadmin is the administration module for the SHOE system and is used to add new shoes to the database by video capturing the soles and uppers and then
W. Ashley
I Forensic
Science
1
International
Date
Entered
15
7-20
: [30/03rS4
Case Numb= Gime Scene
82 (19%)
:
3456
No. :
4567
I
LEAP ReF No. : ,
Member:
II
Informant
Strect Subulb Offence(t]
Complainant(t]
IWJA
: 20284
: 11123 SMITH
ST
I
1
: COLLtNGWOOD :
:
JONES
Fig. 5. Search
screen
for unidentified
impressions.
classifying these new shoes using the same principles as used for searching. It also has the ability to edit existing data on the system. Should a feature on a shoe cover more than one area of the shoe sole, then it may be entered into both areas, or similarly when a feature covers the entire sole, the user only has to ‘click on’ the ‘ALL’ field for the feature to appear in each divided area (Fig. 6). Manufacturer and shoe details are added and edited in this module. The system will not allow the user to save the entry until all three functions of, (i) classification, (ii) shoe and manufacturer details, and (iii) image capture of the sole and upper, are performed and completed. Images of unidentified latent prints are also added to the system in this module and linked to an existing Unidentified Wanted record (Fig. 7). The administration module (SHOEadmin) will also allow for offenders’ shoes, test impressions of same and crime scene impressions to be added to the system if required. 4.1. Experiences The new system has increased the speed with which a shoe can be entered onto the database and classified. The classification system is user friendly and, with the use of icons and descriptors, significantly reduces interpretation error. Searching
16
W. Ashley
I Forensic
Science
International
82 (1996)
7-20
Circlet.Cir&
II-
1 I-
-
With
2 Rings
Pattcm Studs
ChanneVtrough. Gmbination
Fine/broad
Circlef.Circl.5 CirckSingls DesignErand
Wilh 2 Ring* (rolidlhallo~] name
l
Figure 6.
C/ass
Fig. 6. Classification/editing
screen.
parameters have been refined and database searching time is under five seconds. The capturing and classification of new data is very quick with a single entry being processed in under a minute. The discriminating power of the shoe application software has the ability to reduce a selection of 300 images from a single common feature to under 20 with the addition of one further feature. The use of Windows and the ability to restrict information to a limited number of screens or fields have assisted in producing a system that can be operated with little computing knowledge. This system has been designed to operate as a standalone unit or as a multiuser in network format. The Crime Scene Section of the VFSC has a fileserver with a SHOEview workstation and a SHOEadmin workstation, thus allowing each module to operate in isolation. SHOE can be networked to remote sites or operated on a notebook personal computer, allowing portability when attending crime scenes.
5. Discussion
The use of a computerised capture, storage and retrieval system is solely dependent on the quality of the information being entered and the ability of the user to be competently trained to understand and use the system. Research and
W. Ashley
I Forensic
Science
El
International
Lf%PRefNo.:
82 (1996)
7
Member: Informant
W.JA : 120284
Date : 3o/ow94 Location Str-ct:
I
123 SMITH
Suburb: Offence
7-20
ST
COLUNGWOOD
I
:
Cmpfainant[r] : IRAVE ROBBING
Fig. 7. Capture
screen
for unidentified
impressions.
practical experience at the VFSC over a number of years have ensured that such systems are user friendly, not complex in functionality, and produce results. The SHOE system is one that provides the information needed by crime scene investigators and one that can be updated to allow for changes in technology. Future developments will allow the matching of unidentified latent impressions against each other to see if one unidentified type is occurring in large numbers or at specific locations, automatic image matching (pattern recognition) as a first run through the database, and image enhancement processes for images and comparison viewing.
Acknowledgments
The author would like to thank the National Institute of Forensic Science for a Research Grant to support the development of the SHOEQystem, NEC Information Systems for equipment and technical support, Vision Control International for system development, equipment and technical support (SHOE is a trademark of Vision Control International. For further information - contact Vision Control International on E.Mail -
[email protected].), and Messrs Zwolak, Liddy, Lake and Ms. McKelvey for their assistance.
18
Appendix
Classification
W. Ashley
I Forensic
Science
International
8.2 (19%)
7-20
1
codes
BROAD BROKEN
ji..::::::::::::l
DIMIUS
BROAD UNBROKEN FINE COMBINATION FINE/BROAD CHANNELS
TROUGHS
DIAMOND
•l0
CIRCLE WITH
> 2 RINGS
CIRCLE WITH 2 RINGS A SINGLE CIRCLE
CREPE
BRAND
EDGE
NAME
W. Ashley
PROVISION BRANDNAME
[/
I Forensic
Science
International
WAVES
Igl
gggrAL
1+++1
CROSSES
w
%~!ENGTH
m]
EZENGTH ARCS c
+
1
c
/c
DIAMONDS
COMPLEX
HEXAGONS
m -
7-20
PLAIN
FOR
STUDS
j
82 (1996)
Lia
ZIGZAG
KT-26 I
lZGl
tt%Zi.~cTH AMP>% WAVELENGTH
ROUND
OVAL
ml COMPLEX ZIG-ZAG
SQUARWRECTANGULAR
PARTIAL
STARS
TRIANGLES
UNUSUAL BY
CAUSED
CHANNELS/TROUGHS
19
20
W. Ashley
I Forensic
Science
International
82 (1996)
7-20
References [l] E.D. Hamm, Track identification: an historical overview. J. Forensic Zdentif., 39 (1989) 333-338. [2] FBI International Symposium on Footwear and Tyretracks, Presentation of Shoe Reference Files, June 1994, Quantico, Virginia. [3] W.J. Ashley, Shoe Sole Pattern File, Presented at the 12th Meeting of the International Association of Forensic Sciences, Adelaide, 1990. [4] W.J. Bodziak, Footwear Impression Evidence, Elsevier, New York, 1990, pp 242-250, 331. [5] Victoria Forensic Science Centre, Crime Scene Section, Annual Statistical Reports 1992-1994.