Primary Uses

Fingerprint matching is a technique that is frequently used by law enforcement agencies in criminal investigations to identify suspects and link them to crimes. When a crime is committed, the perpetrator may leave latent fingerprints on surfaces that they touched. These latent prints can be collected by law enforcement officials and compared to the fingerprints of known suspects or to the prints stored in a database. If there is a match, this can provide strong evidence linking the suspect to the crime.

It is also often utilized as part of the background check process for employment or security clearance. In these cases, the individual's fingerprints are checked against databases to verify their identity and to look for any criminal history.

Fingerprint matching is also used in the immigration process to verify the identity of individuals applying for visas or seeking to enter a country. In these situations, the fingerprints are used to confirm the individual's identity and to check for any prior immigration violations or criminal activity.

Fingerprint Matching Algorithms

Fingerprint matching algorithms are used in both fingerprint verification (confirming that a fingerprint belongs to a specific person) and fingerprint identification (finding which of a group of people a fingerprint belongs to). These algorithms can be used to search a database of many fingerprints in order to identify a specific person. In order to speed up the search process, fingerprint classification and indexing techniques are often used. These techniques group fingerprints based on their characteristics, making it easier to find a match in the database.

Matching fingerprint images can be difficult due to the following factors:

• Displacement: the finger being in different locations on the scanning device
• Rotation: the finger being at different angles relative to the scanning device
• Partial overlap: a smaller overlap between the template and input fingerprints, especially for small scanning devices
• Non-linear distortion: the mapping of the 3D shape of a finger onto a 2D surface causing distortions in scans
• Pressure and skin condition: factors such as pressure, dryness, and humidity affecting the ridge structure of the fingerprint
• Noise: introduced by the fingerprint scanning system

Automatic fingerprint matching algorithms have been developed to compare fingerprints and determine if they belong to the same person. However, these algorithms can struggle to match low-quality and partial fingerprints. In human-assisted systems, a quality-checking algorithm can be used to ensure that only high-quality fingerprints are used for comparison.

Human Fingerprint Examiners

Human fingerprint examiners use various factors to determine if two fingerprints are a match, including the overall pattern and type of the fingerprints, the presence of identical minute details, and the relationship between corresponding minute details. They also follow complex protocols and use a detailed flowchart to guide their process. The minimum number of corresponding minute details required for a match is typically 12, according to United States forensic guidelines.

Automatic Fingerprint Matching

A fingerprint matching algorithm is used to compare two fingerprints and determine if they belong to the same person. To do this, the algorithm first converts the fingerprints into a simplified form, called a "template," which represents the unique characteristics of the fingerprints. The algorithm then compares the templates and produces a score indicating how similar they are. If the score is above a certain threshold, the algorithm determines that the fingerprints match.

However, in unattended systems that are frequently used in commercial applications, human intervention is not possible. Poor-quality fingerprints are often responsible for a large percentage of false non-match errors made by these algorithms. While significant progress has been made in fingerprint recognition technology, there is still a need to develop more robust systems that can accurately compare low-quality fingerprints, especially for large-scale applications or when using small, low-quality sensors.

Automatic fingerprint matching techniques can be divided into three categories: correlation-based, minutiae-based, and non-minutiae feature-based. These approaches do not necessarily follow the same guidelines as manual fingerprint matching, and have been developed specifically for automation.

• Correlation-based: comparing the pixels of two fingerprint images by superimposing them and computing the correlation between corresponding pixels for different alignments (e.g., various displacements and rotations)
• Minutiae-based: the most popular and widely used technique, which involves extracting minutiae from the two fingerprints and storing them as sets of points in a two-dimensional plane. The alignment between the template and the input minutiae feature sets that result in the maximum number of minutiae pairings is found.
• Non-Minutiae feature-based: used when minutiae extraction is difficult in extremely low-quality fingerprint images. Features of the fingerprint ridge pattern, such as local orientation and frequency, ridge shape, and texture, are extracted and compared. Correlation-based techniques can be considered a subfamily of this approach, as the pixel intensity of an image can be considered a feature of the fingerprint.

It is difficult to determine the best algorithm for matching fingerprints because different applications have different performance requirements and it is hard to compare the performance of different algorithms on different benchmark datasets. Some applications may prioritize accuracy, while others may prioritize efficiency or scalability. Additionally, it is difficult to determine whether improvements in performance are due to the matching algorithm or other factors, such as changes in the feature extraction method. To objectively compare fingerprint matchers, it is necessary to use the same set of features, such as minutiae for minutiae-based matchers.