Automated Tattoo Recognition: A Machine Learning Approach Integrating Facial Biometrics

Abstract

Tattoo identification has become an important forensic tool for crime investigations, suspect tracking, and tracing missing persons. Manual image retrieval and classification are at the heart of traditional approaches, which are tedious and susceptible to human error. In this work we propose an automated tattoo recognition system which uses machine learning approaches while combining with other biometrics, in this case faces, to improve the performance of the tattoo search. To do this, we use deep learning models to parse tattoos for their unique characteristics, assign them classes, and then compare the results to a central database. Also integrated are facial biometrics, which help to cross-reference individuals with facial features and provide an additional layer of verification for identifying individual users. To ensure robustness across skin tones and tattoo styles, the system is trained on a wide variety of tattooed skin and imaging conditions. Extensive experimental results show great advancements in recognition precision and retrieval speed over traditional approaches. This study demonstrates their potential to combine tattoo recognition and facial biometrics for law enforcement, personal or security applications.