Predicting suitability of finger marks using machine learning techniques and examiner annotations

Interpol review of fingermarks and other body impressions (2019 – 2022)

“…LQMetrics delivers a quality map and a series of quality scores including a probability of a rank-1 correspondence on AFIS. Swofford et al [ 65 ] proposed the DFIQI tool which measures the clarity of friction ridge features (locally), evaluates the quality of impressions (globally), and maps them to three distinct value scales (value, difficulty and complexity, as defined by Eldridge [ 66 , 67 ]). In an operational environment, the tool is intended to provide an empirical foundation to support experts' subjective judgments and to provide transparency to the overall quality of a given mark.…”

“…Four aspects of suitability need to be distinguished depending on the intended purpose: hence for scales named value , complexity , AFIS suitability and difficulty [ 70 ]. Taking advantage of machine learning techniques, the researchers have proposed a predicting model for value decisions based on automatically extracted quality or selectivity measures in conjunction with a limited set of user inputs [ 67 ]. The model achieved accuracy at a similar level to that of examiners asked to make the same suitability determinations across all four scales.…”

Toward better AFIS practice and process in the forensic fingerprint environment

“…Additionally, strategies aimed at reducing the variability involved in subjective human-based methods, such as consensus approaches, are also helpful [ 53 ]. Research is currently underway to explore automated approaches to reduce human elements, such as automated quality assessments and evaluative approaches [ 54 , 55 ] (D. A [ 6 , 56 ]. Promisingly, the discipline is beginning to explore automated lights-out latent processing in daily workflows [ 57 ].…”