Abstract:Currently in the UK, if a person is arrested or charged with a recordable offence, they can have prints of their footwear taken whilst in custody. The tread pattern recorded in these prints can be searched for using the National Footwear Database to find out if the same footwear pattern has been recovered at previous crime scenes, generating forensic intelligence. TreadMatch is a digitised system for collecting footwear prints seized from detainees in custody for this purpose. Whilst its use for generating intelligence is accepted, validation experiments have not been conducted to understand its level of performance in assisting in forensic comparison purposes for identification, because in the absence of an incorporated scale, it is not known how well TreadMatch reproduces the pattern size of a tread, threatening the validity of the comparison. If it can be determined that the measurements of TreadMatch scans are consistent with the more commonly used aluminium powder test marks, this could save Police time and resources if the digital scans could be used for preliminary assessment prior to the footwear being physically submitted for evidential comparison. Therefore, this study set out to compare three different types of TreadMatch scans ('dynamic', 'zoomed' and 'rolled') for thirty different pieces of footwear, with test marks of the same footwear using the traditional method (fingerprint powder). Length and width measurements were obtained from each tread pattern using GNU Image Manipulation Program software. The resulting data were analysed to assess for agreement between TreadMatch scans and test marks using 95% Intraclass Correlation Coefficients (ICC) and 95% Bland-Altman plots of Limits of Agreement (LOA). Additionally, an intra-sample study using fifteen repeated measurements of the same piece of footwear for different TreadMatch scanning methods was carried out to support the larger validity study. 95% ICC3,1 resulted in coefficients ranging from 0.99 to 1.00 across all measurements. 95% LOA displayed close agreement. There was less agreement and more variation displayed between the test marks and the TreadMatch rolled scans for both length and width measurements. This variation for hand-rolled prints must be taken into consideration and a standard approach developed. The study suggests TreadMatch can be used for preliminary assessments in assisting forensic comparisons, particularly for dynamic and zoomed prints.

Spatial modeling of randomly acquired characteristics on outsoles with application to forensic shoeprint analysis

A Bayesian Hierarchical Model for Evaluating Forensic Footwear Evidence

Classification of footwear outsole patterns using Fourier transform and local interest points

Computerized Matching of Shoeprints Based on Sole Pattern

Estimate of the random match frequency of acquired characteristics in footwear: Part I - Impressions in blood

Technical note: The Next Step - a semi-automatic coding and comparison system for forensic footwear impressions

Improving and Evaluating Machine Learning Methods for Forensic Shoeprint Matching

Creating a Forensic Database of Shoeprints from Online Shoe Tread Photos

Sexing Caucasian 2D footprints using convolutional neural networks

CriSp: Leveraging Tread Depth Maps for Enhanced Crime-Scene Shoeprint Matching

Learning Wear Patterns on Footwear Outsoles Using Convolutional Neural Networks

3D Shape Measurement of Shoeprint Impression with Structured Illumination and Fringe Pattern Analysis

Score‐based likelihood ratios for barefootprint evidence using deep learning features

A preliminary study on the stability of bare footprint linear measurements in four motion states

The application of TreadMatch scans to aid the process of footwear mark comparison

Similarity Quantification of Bare Footprint Based on Linear Measurement and Shape Context Contour Combined Score-based Likelihood Ratio Evaluation

Forensic shoe-print identification: a brief survey

Classification of 3D shoe prints using the PointNet architecture: proof of concept investigation of binary classification of nike and adidas outsoles

Matching 3D Foot to Shoe Last Based on Axial Deformation Algorithm

Comparison Of Rarefication Techniques For Foot Simulation Using Subject Specific Three-Dimensional Anthropometry Data

The Use of Scale-Invariance Feature Transform Approach to Recognize and Retrieve Incomplete Shoeprints.