Automated Latent Fingerprint Recognition (1704.01925v1)

Abstract: Latent fingerprints are one of the most important and widely used evidence in law enforcement and forensic agencies worldwide. Yet, NIST evaluations show that the performance of state-of-the-art latent recognition systems is far from satisfactory. An automated latent fingerprint recognition system with high accuracy is essential to compare latents found at crime scenes to a large collection of reference prints to generate a candidate list of possible mates. In this paper, we propose an automated latent fingerprint recognition algorithm that utilizes Convolutional Neural Networks (ConvNets) for ridge flow estimation and minutiae descriptor extraction, and extract complementary templates (two minutiae templates and one texture template) to represent the latent. The comparison scores between the latent and a reference print based on the three templates are fused to retrieve a short candidate list from the reference database. Experimental results show that the rank-1 identification accuracies (query latent is matched with its true mate in the reference database) are 64.7% for the NIST SD27 and 75.3% for the WVU latent databases, against a reference database of 100K rolled prints. These results are the best among published papers on latent recognition and competitive with the performance (66.7% and 70.8% rank-1 accuracies on NIST SD27 and WVU DB, respectively) of a leading COTS latent Automated Fingerprint Identification System (AFIS). By score-level (rank-level) fusion of our system with the commercial off-the-shelf (COTS) latent AFIS, the overall rank-1 identification performance can be improved from 64.7% and 75.3% to 73.3% (74.4%) and 76.6% (78.4%) on NIST SD27 and WVU latent databases, respectively.

• The paper presents a comprehensive automated latent fingerprint recognition system leveraging ConvNets and multiple templates to improve accuracy.
• The system achieved rank-1 identification accuracies of 64.7% on NIST SD27 and 75.3% on WVU databases, surpassing prior methods and matching a commercial system.
• This research demonstrates the effectiveness of integrating deep learning, descriptor matching, graph-based algorithms, and template fusion for enhancing latent fingerprint recognition in forensic science.

Automated Latent Fingerprint Recognition

The paper "Automated Latent Fingerprint Recognition" by Kai Cao and Anil K. Jain presents a comprehensive automated system for latent fingerprint recognition, leveraging Convolutional Neural Networks (ConvNets) to address the persistent challenge of poor recognition rates in comparison to reference fingerprints. Latent fingerprints, crucial in forensic science, often suffer from low quality due to factors like smudging and incomplete imprints, which complicates their automated recognition against large fingerprint databases.

This paper outlines an algorithm for latent fingerprint recognition that improves the representation of such prints using three complementary templates: two minutiae-based templates and one texture-based template. These templates incorporate advanced methodology, such as ConvNets for ridge flow estimation and minutiae descriptor extraction, and dictionary-based methods for ridge enhancement and spacing estimation.

Key Findings and Results

The experimental evaluation of the proposed recognition system showed significant improvement over existing methods. When tested against a reference database of 100,000 rolled prints, the system achieved rank-1 identification accuracies of 64.7% for the NIST SD27 database and 75.3% for the WVU latent database. These results surpassed preceding studies and matched or exceeded the performance of a leading commercial off-the-shelf (COTS) latent Automated Fingerprint Identification System (AFIS), which reported rank-1 accuracies of 66.7% for NIST SD27 and 70.8% for WVU DB. Moreover, by implementing score-level and rank-level fusion with the COTS system, the overall rank-1 identification accuracies improved to 73.3% (74.4%) for NIST SD27 and 76.6% (78.4%) for WVU DB.

Technical Contributions

The research makes several technical contributions:

• Utilization of ConvNets: This application of deep learning advances, notable in other fields like image and face recognition, shows promise in effectively handling noisy and fragmented latent fingerprint data.
• Descriptor-based Matching: Multi-scale and multi-location window-based learning for minutiae descriptors presents a sophisticated approach for enhancing the discriminatory power of fingerprint features.
• Graph-based Matching Algorithms: Leveraging second-order and third-order graph matching algorithms for minutiae correspondence proved effective in reducing false matches and optimizing true mate identification.
• Template Fusion: The fusion of minutiae and texture templates captures diverse layers of information, further enhancing recognition accuracy and overcoming limitations of individual templates.

Implications and Future Work

The implications of this paper are profound for forensic science and automated fingerprint systems. The enhanced accuracy and reliability can significantly reduce the manual effort required by forensic examiners, thus increasing throughput and reducing bias. From a theoretical perspective, the use of ConvNets opens up opportunities for further refinement and adaptation of neural network models to fingerprint recognition tasks.

Future research can explore improvements in ConvNet architectures, integration of additional features, such as ridge count and singular points, and strategies for large-scale database matching efficiency. Additionally, constructing a larger and diverse database of latent fingerprints can aid in refining machine learning models and descriptors.

Overall, this paper contributes substantial advancements in the automated recognition of latent fingerprints, highlighting methodologies that can be further developed and integrated into operational forensic systems.