An algorithm to compare two-dimensional footwear outsole images using maximum cliques and speeded-up robust feature

Area of Science:

• Forensic Science
• Computer Vision
• Pattern Recognition

Background:

• Footwear impression comparison is crucial in forensic investigations.
• Existing methods face challenges with degraded or smudged crime scene prints.
• Class characteristics like pattern and size can make comparisons difficult.

Purpose of the Study:

• To develop and evaluate a novel algorithm for comparing shoe outsole impressions.
• To improve the accuracy and robustness of footwear examination in forensic science.
• To provide a reliable tool for matching crime scene prints to known exemplars.

Main Methods:

• Feature extraction using Speeded-Up Robust Features (SURF).
• Image alignment based on Maximum Clique (MC) algorithm.
• Feature combination and similarity scoring using Random Forest (RF) and a novel algorithm (MC-COMP).

Main Results:

• The proposed MC-COMP-SURF algorithm demonstrated superior classification precision compared to other methods.
• The algorithm effectively identifies unique features, outperforming alternatives on degraded and smudged impressions.
• Random Forest implemented on SURF features showed strong performance in initial comparisons.

Conclusions:

• The MC-COMP-SURF algorithm offers a significant advancement in footwear impression analysis.
• This method enhances the ability to accurately match shoe prints in challenging forensic scenarios.
• The algorithm is available as an R-package (shoeprintr) for practical application.