The development of an automatic recognition system for earmark and earprint comparisons

Area of Science:

• Forensic Science
• Biometrics
• Pattern Recognition

Background:

• The reliability of earmarks for personal identification is debated due to a lack of structured data and research on feature selectivity.
• Existing methods often require manual feature extraction, limiting efficiency and objectivity.

Purpose of the Study:

• To develop and evaluate an automated system for comparing earprints and earmarks without manual intervention.
• To provide a robust data basis for assessing the value of earmarks in forensic identification.

Main Methods:

• A novel system for automatic earprint and earmark comparison was developed, creating donor models from multiple reference prints.
• Image alignment and a proximity score based on normalized 2D correlation coefficient are used for comparison.
• The system derives a likelihood ratio to assess the probability of a match from a known source.

Main Results:

• The system achieved an equal error rate (EER) of 2.3% for mark-to-print comparisons using a dataset of 1229 donors.
• Approximately 88% of marks were found within the top 3 positions of a hitlist.
• Print-to-print comparisons yielded a lower EER of 0.5%.

Conclusions:

• The automated system demonstrates significant potential for enhancing the efficiency and reliability of earprint-based personal identification.
• The system's performance, validated on both research and real-case data, supports its utility in forensic applications.
• Further research and validation are warranted to fully establish the system's role in forensic identification practices.