A Survey on Fingerprint Verification Algorithms Essay Example
A Survey on Fingerprint Verification Algorithms Essay Example

A Survey on Fingerprint Verification Algorithms Essay Example

Available Only on StudyHippo
  • Pages: 7 (1710 words)
  • Published: February 2, 2017
  • Type: Case Study
View Entire Sample
Text preview

Worldwide, fingerprints are recognized as the principal and most dependable biometric for uniquely identifying individuals in various applications like verification and law enforcement. Different strategies used for matching these prints encompass filter-based, minutiae or Galton feature's-based, correlation-based, and pattern matching techniques. The frequently employed method is rooted on minutiae that identify distinctive features of a fingerprint such as ridge endings, ridge bifurcation, islands, core, delta, crossover among others.

The specific positions and orientations of detected minutiae points are kept in a database to evaluate the techniques suggested by various studies. Fingerprints usually have approximately 20-60 such minutiae points. The effectiveness of the system greatly depends on the number of these points taken into account. The objective of the algorithms presented by the assessed research is to enhance computational speed. Each algo

...

rithm is designed for particular uses, possessing its own set of pros and cons.

Minutiae points, which mark the unique features of a scanned fingerprint image, serve as the basis for fingerprint matching. However, the reliability of minutiae recognition can vary depending on different factors such as the scanner used, the quality of the scanned image and the compatibility of the scanning devices. Factors like available resources (e.g., storage space and speed requirements) influence how minutiae characteristics, their positions or coordinates, orientations and types are utilized during fingerprint recognition and verification processes.

The precision can be enhanced by adding more specifics such as the count of ridges between two chosen minutiae points. During the verification process, not all minutiae points are taken into account due to intermediary processes like binarization and skeletonization which might cause incorrect points or overlook less

View entire sample
Join StudyHippo to see entire essay

noticeable minutiae points. Moreover, including every minutiae point could potentially reduce the system's effectiveness and accuracy. For a match to be deemed successful, a particular group of minutiae points (which varies depending on the application) from the input image must correspond with those in the reference image if they originate from the same person.

Most fingerprint matching methods employ the steps outlined above, but procedures such as binarization and skeletonization are not mandatory. This article delves into the topic of fingerprint minutiae recognition and validation systems, based on genetic algorithm. The system works by extracting minutiae points from fingerprints and comparing them to an already stored set of minutiae points. Multiple steps are involved in the extraction of minutiae. It starts with obtaining an image, followed by a block direction process which utilizes the Coetzee algorithm to get a directional image based on pixel intensity. Following this, the image is segmented and binarized. Segmentation involves categorizing and labeling identical pixels based on their gray scales.

With the use of suitable low pass filters, the grayscale image is subjected to a binarization process which transforms it into binary format. Following this, the image undergoes skeletonization - a method that ensures each object in the image has only a single pixel width. This can be achieved through either Zhong and Suen's algorithm or Marthon's algorithms. The primary features utilized in this procedure are ridge endings and bridge points; these features as per literary research findings, are identified as the most accurate minutiae attributes. The positions of these minutiae points are preserved and later used for fingerprint matching.

Identification of fingerprints can be performed either

through the Hough transformation or the Genetic algorithm. In the process involving Hough's algorithm, a specific scaling factor transforms the minutiae set of one image to coincide with that of another image. Alternatively, genetic algorithms assume a binary approach; 1 representing a correlation between two fingerprints and 0 indicating a lack of any relation. This is instrumental in encoding the population traits of fingerprints. The procedures of both the Hough transformation and the genetic algorithm are detailed step by step.

The article tackles the issue of inadequate storage capacity and reduced processing speed inherent in integrated circuit cards in the context of fingerprint matching algorithms. The typical preprocessing stages such as image filtering, skeletonization and binarization are bypassed, due to their significant computational expense. Instead, minutiae characteristics are directly obtained from grayscale fingerprint images following the method suggested by D Maio et al. Additionally, not all minutiae details like their types are kept. Verification of the fingerprints relies solely on the locations and respective orientation fields of minutiae points.

The initial phase of this algorithm involves computing the orientation field of the fingerprint template. It's essential for the orientation field to be accurately computed and saved as a feature. The input fingerprint image is segmented into suitably sized blocks. Following that, the x-gradient and y-gradient for each pixel in each block are computed before estimating the local orientation for each block as per the provided formula. The average and standard deviations of the orientation fields are also preserved. To expedite these computations, a sliding window technique is utilized. The subsequent stage outlined is the registration of feature patterns, which necessitates carrying out

the similarity transform on the two fingerprints derived from the same finger.

A hierarchy-based tactic is implemented to align the two minutiae patterns' positions. The proceeding phase outlined is known as coarse registration. In this stage, the presumption is that the template fingerprint and the entered fingerprint share at least one similar minutiae point that can be precisely extracted. The set of transformation parameters are defined prior to conducting additional transformations. This stage accommodates a 360-degree variance in the base and the input fingerprint's orientation. Subsequently, the input fingerprint's orientation field is calculated, which differs from the approach used for the template image. For this calculation, only the pixels situated on the ridges are taken into account because external ridges are deemed irrelevant. The matching score gets defined. The rough estimation of the two fingerprints' alignment is determined upon completing this stage. The final step involves conducting fine registration to enhance the matching score even further.

The calculation of the matching score utilises a significant number of test cases. The tree comparison-based algorithm measures the ratio of relational distances. This study also includes a fingerprint comparison algorithm. The assurance is that it has the capability to compare two images captured from similar or various sensors. The primary image is the one first scanned and kept. The input images are those scanned later and contrasted against the initial image for identification purposes. This study's algorithm calculates the ratio of relational distances between minutiae points during the comparison process. It incorporates: identifying common minutiae points in the primary and subsequent image and subsequently undertaking a comparison between these points. In the initial phase of

the algorithm, intersecting the two sets of minutiae unveils common points. By calculating Euclidean distances, the nearest minutiae points to each point are identified.

The procedure entails computing the relevant 10 ratios and angles, which are subsequently saved in tuple format for each minutiae point. This step is implemented for both the reference and entered images. During the matching procedure, two tuples are deemed identical if they share more than 2 identical ratios and angles. The points that match are referenced as Candidate common points. Following this, the matching stage ensues. In this stage, the candidate points are categorized into confirmed candidate points and false points. Solely those points from the reference image that appear in the candidate list are acknowledged as confirmed points and a tree structure is created from the base upwards. Each point signifies a vertex and the lines connecting them become edges. These edges embody details such as quadrant, angle relative to the preceding edge, proportion of its length to the former edge, etc.

The method is likewise applied to the input picture. The analysis is carried out by sequentially erasing the borders, and accordingly, a matching rating is derived. The number of items in the Confirmed Common points list is represented as C(N), and N is the maximal count of points either in the base image or input picture, hence C(N) should be greater than or equal to N/2. If this condition holds correct, it implies that both images are identical. On the contrary, if it isn’t true, a negative rating will be reflected, signifying dissimilar images.

Despite the abundance of minutiae points in fingerprints, ridge endings

and branch points are the most recognizable. Some systems do not need information about these point types; knowing their locations can sometimes be enough. This lessens memory consumption, making it suitable for devices with restricted memory like PDAs or integrated circuit boards. Nonetheless, this method does not accommodate minor overlaps between base and input images. Compared to Marthon's technique, Zhong and Suen's algorithm is more efficient for thinning purposes despite its potential to create superfluous minutiae points. Hence, scanning the image without implementing any pre-processing methods is advised.

In evaluating the likeness of two minutiae maps, it is crucial that both maps are accurately oriented and aligned. However, a system that utilizes an algorithm which builds a minutiae tree based on the ratios of relational distances, doesn't require image orientation prior to comparison. Not using enhancement filter technologies, like Gabor filters, can potentially minimize the accuracy of the matching algorithm but on the other hand, it enhances the versatility of use and integration into various sensors, thereby improving the portability and simplicity of the algorithm.

The attributes applied in every matching algorithm differ based on the particular problem addressed in the research. Taking into account the number of ridges between the minutiae points alongside their locations and types could enhance accuracy. Moreover, the image quality significantly affects accuracy. For the development of applications that are interoperable with various sensors, an algorithm that isn't dependent on a single sensor is crucial. Going forward, there could be a requirement to use diverse devices in tandem.

It is essential for efficient minutiae maps storage in order to utilize fingerprint matching software in devices with inadequate

memory such as mobiles, integrated circuit chips, and PDAs. Moreover, innovative techniques like genetic programming can serve to enhance the precision of matching algorithms, ensuring efficiency, limited minutiae points storage, and a smaller overlap region. While iris matching is an option for biometric verification, fingerprint matching remains the most straightforward and cost-efficient method. Thus, refinement of current fingerprint verification techniques is vital and continues to pique considerable interest.

Get an explanation on any task
Get unstuck with the help of our AI assistant in seconds
New