Abstract
Toolmark evidence is traditionally examined using comparison microscopy, where two surfaces are visually compared using two-dimensional (2D) images. This process relies heavily on examiner experience and subjective judgement, which can limit reproducibility and the ability to produce a statistical measurement of similarity. To improve objectivity in toolmark identification, this research investigates the application of areal topography (2.5D) surface topography measurements combined with correlation-based statistical analysis.Tool surfaces (chisels) and toolmark impressions were digitised using optical surface metrology techniques, including focus variation microscopy (Alicona IFM G5) for acquisition of areal datasets and digital microscopy (Keyence VHX-7000) to acquire 2D images. The resulting areal datasets were analysed using correlation metrics, including the maximum areal cross-correlation function (ACCFmax), which quantifies the similarity between two surface topographies, and the scale difference parameter (Ds), which evaluates differences in surface scale features. The preprocessing method wavelet decomposition was investigated as an alternative to the ISO 25178-3 Gaussian filtering approach to improve correlation reliability.
The initial studies showed that tool surface topography contains sufficient unique detail to allow differentiation between tools using statistical correlation methods. The results indicate that wavelet decomposition improves the separation between known matching and known non matching tool surfaces when compared to traditional Gaussian filtering methods.
The influence of substrate material on toolmark formation was then investigated using materials commonly encountered in forensic investigations, including wood, UPVC, aluminium and steel. These materials were selected as they are commonly encountered in forensic casework, particularly in burglary and forced-entry scenarios (internal/ external entryways and safe deposit boxes. Toolmarks were created using five chisels at controlled angles of attack (15°, 30°, and 45°), and the impressions formed were analysed using the same surface metrology and correlation methods. Hardwood substrates showed improved dimensional stability when compared with softwood due to reduced variability in indentation behaviour. Experiments on UPVC demonstrated that impressions formed at a 30° angle of attack produced the most consistent and reproducible toolmarks. Environmental exposure experiments showed that dimensional changes in impressions occurred during the first two weeks following toolmark creation.
Further experiments examined the effect of environmental exposure and time-based degradation on toolmark stability. Aluminium substrates showed gradual changes in correlation values when exposed to uncontrolled environmental conditions, while steel substrates initially preserved fine surface detail but were more susceptible to degradation due to oxidation over time.
Finally, double-blind case studies were conducted in which multiple participants produced toolmarks on realistic substrates including varnished and unvarnished wood, UPVC window frames and steel surfaces. Toolmarks were analysed using digital microscopy and correlation based surface comparison to determine the most likely matching tool. The results demonstrate that the proposed correlation-based methods can differentiate between matching and non matching toolmarks across a range of substrates while maintaining robustness under realistic forensic conditions.
Overall, this research demonstrates that 3D surface metrology combined with correlation-based statistical analysis provides a more objective framework for toolmark identification than traditional visual comparison methods and has the potential to improve the reliability and reproducibility of forensic toolmark evidence.
| Date of Award | 19 Jun 2026 |
|---|---|
| Original language | English |
| Sponsors | Engineering and Physical Sciences Research Council |
| Supervisor | Katie Addinall (Main Supervisor) & Liam Blunt (Co-Supervisor) |