Published in Applied Sciences on March 30, 2021 / MDPI-Open Access Journals

The study focuses on the impact of small errors in latent fingerprint identification systems. Several experiments were conducted where ground-truth minutiae were removed from latent fingerprints. The effects on matching scores and rank-n identification were evaluated using two different matchers and the NIST SD27 dataset. The findings indicate that even the absence of a single minutia can greatly affect identification performance. Additionally, fingerprints that were initially ranked highly can be demoted to lower ranks when multiple minutiae are missing. Certain minutiae were found to be more significant than others in correctly identifying a latent fingerprint. Based on this discovery, a dataset was created to train machine learning models to predict the impact of each minutia on the matching score of a fingerprint identification system. The best-trained model can successfully determine whether a minutia will increase or decrease the matching score of a latent fingerprint.

Click on the link below to take you to the research.

View This Research