Automated Fingerprint Identification System (AFIS)
AFIS is a biometric identification system that uses digital imaging technology to obtain, store and process fingerprint data. AFIS has the ability to classify, match and store fingerprints and palm prints from criminal records and applicants.
What are latent fingerprints?
The development of AFIS revolutionized the field of forensic fingerprint examination. It helped not only in accurate analysis as well as effective time analysis.
The very first AFIS was created in 1974 by the Federal Bureau of Investigation (FBI). Previously, this system contained only important fingerprint details or features because it was expensive to store the full fingerprint image.
Processing 1,00,000 impressions took 30 minutes and now it’s even faster just in the blink of an eye. The FBI uses the terms AFIS and IAFIS (Integrated Automated Fingerprint Identification System) interchangeably due to the longevity of the FBI’s IAFIS.
Now, identification systems support palm, iris and face prints as well as fingerprints, known as Automated Biometric Identification System (ABIS). The first AFIS Inovatrics of this type was deployed in 2009.
Fingerprint identification is based on the details or location/direction of ridge terminations or bifurcations along the ridge path. The information elucidated from the friction edges includes the flow of the edges, the presence or absence of features along the individual trajectories of the friction edges and their sequences, and the intricate details of a single edge. This information is called 3-level detail.
An AFIS is designed to interpret the stream of global peaks to assign a fingerprint classification and then extract fine detail – a subset of the total amount of information available but enough information to search effectively an extensive repository of fingerprints.
AFIS is a computerized system, so it has two parts: hardware and software.
The hardware includes the sensors that are used to collect the digital images of the fingerprint. The types of sensors used in AFIS are:
- Optical sensors– These sensors are the most commonly used because they only capture the optical image of fingerprints.
- Ultrasonic sensors– These sensors use high frequency ultrasonic waves or optical devices that use prisms to detect the change in light reflection associated with fingerprints.
- Capacitive sensors– The sensors determine each pixel value based on the measured capacitance which indicates that the capacitance of the valley is less than that of the friction ridges.
- Thermal scanners– When a finger is swiped across the scanner surface, the temperature difference over time is measured to create a digital image of the fingerprint.
Software employed in AFIS is responsible for fingerprint matching. The matching techniques used are of two types – detail-based matching and pattern-based matching.
Minutia-based matching relies on the 3-level detail of friction ridges, while pattern-based matching helps compare two prints and find duplicates. The comparison of fingerprints is done by matching the patterns present in the software. Templates are the mathematical representations of stored fingerprint images.
Digital fingerprint images are processed by computer algorithms that have been developed over the past decades. These computer algorithms improved the operational productivity of law enforcement and reduced the number of fingerprint analysts needed.
AFIS search algorithm
Algorithms involved in passing an AFIS exam include:
1. Acquisition of digital fingerprints
Fingerprint images are captured using the sensors mentioned above. The parameters that characterize a digital fingerprint image are the resolution area, the number of pixels, the geometric precision, the contrast and the geometric distortion.
Sensors often capture a series of images instead of a single fingerprint image. Depending on the application for which the scanner was designed, it can run one or more algorithms using a resource-limited on-board microprocessor (memory and processing power) or using an attached computer.
The different algorithms used are- Automatic Fingerprint Capture Algorithm, Image Data Compression Algorithm, Vitality Detection Algorithm, Fingerprint Matching Algorithm, Image Processing Algorithms and Cryptographic algorithms and protocol(s).
2. Image enhancement
The fingerprint images obtained may exhibit different types of noise introduced during the acquisition process. These noises can come from dust or dirt particles on fingers, poor image quality, incomplete prints, etc.
Image enhancement algorithms can locate those noisy areas that give the optimal match on a large collection of fingerprint images. Moreover, these noises are useful in the detection and individualization of features in later stages.
3. Feature extraction
An automatic feature extraction algorithm is used to mimic details. The characteristic minutiae considered are ridge terminations and bifurcations. Other features are not included because they are difficult to extract.
The step involves a binarization algorithm that converts the enhanced grayscale fingerprint image into binary shaped black pixels (ridges) and white pixels (valleys). Then a thinning algorithm is used to convert the binary images to a single pixel width around the center ridgeline.
Finally, a minutiae detection algorithm is used for the thinned image which locates the x and y coordinates as well as the orientation of the minutia points.
4. Corresponding to
An automatic matching algorithm is used for this process which works on the results of the feature extraction algorithm and finds the similarities or differences between the fingerprint sets. This algorithm can perform a comparison at the rate of 10,000 per second and results can be sorted based on similarities.
Automatic fingerprint matching algorithms can give imperfect results due to the difficult problem of large interclass variations (variability of different fingerprints of the same finger) present in fingerprints such as: displacement, rotation, partial overlap , nonlinear distortion due to finger pressure. elastic three-dimensional finger on a rigid two-dimensional imaging surface, pressure, skin conditions, noise introduced by the imaging environment, and errors introduced by automatic feature extraction algorithms.
A robust fingerprint matching algorithm must be able to handle all these intraclass variations in different fingerprints of the same finger.
5. Indexing and recovery
The above mentioned processes are very time consuming, that’s why indexing and retrieval algorithms are introduced to speed up the search. Indexing is performed using automatic fingerprint classification algorithms. A recovery strategy is also required based on applications such as accuracy and efficiency, matching, human review, etc. As soon as a match is found in the database, the search is automatically stopped.
The Indian version of the Automated Fingerprint Identification System (AFIS) is called FACTS, co-developed by NCRB and CMC Ltd., India. The system uses image processing and pattern recognition techniques to capture, encode, store and match fingerprints, including random fingerprint matching. This system is used by the Central Fingerprint Bureau of India.