Report Project Fingerprint Authentication Course Biometric Authentication Systems.pdf

School of Information and Communication Technology

Project: Fingerprint Authentication

Course: Biometric Authentication SystemsGroup 4:

Nguyen Quang Hung (20214961)Nguyen Tuan Long (20214963)Do Quang Minh (20210579)

Instructors:Dr Tran Nguyen NgocDr Ngo Thanh Trung

1.1 Problem statement 3

1.2 Summary of the fields 3

1.3 Motivation of our study 3

2 Method details 42.1 Basic theory of fingerprint recognition 4

2.2.2 Oriented Field Estimation 8

2.2.3 Ridge Frequency Estimation 9

2.2.4 Gabor Filter 10

2.3 Recognition Algorithm 10

2.3.1 Introduction to ORB 10

2.3.2 Adaptability to Fingerprint Recognition 12

2.3.3 Fast and Efficient Matching 12

3 Experiment 133.1 Data 13

Group contribution:

• ORB algortihm, Reports: Nguyen Tuan Long• Data Preprocessing, Slides: Do Quang Minh• Presentation: Nguyen Quang Hung

1.1Problem statement

Fingerprint authentication encounters challenges in achieving high accuracy, particularly inscenarios with low-quality prints and variations in skin conditions, necessitating improvedmatching algorithms Striking a balance between false acceptance and false rejection ratesis critical for the reliability and effectiveness of the authentication system Robustness isessential to handle variations in fingerprints due to factors like aging, injuries, and differentsensor types, highlighting the need for adaptable solutions Security concerns arise from thesusceptibility to spoofing attacks, emphasizing the importance of implementing advanced anti-spoofing techniques to enhance system security To ensure widespread acceptance, addressingthese challenges requires advancements in technology, including faster matching speeds andstreamlined user enrollment processes.

1.2Summary of the fields

Fingerprint authentication encounters multifaceted challenges spanning accuracy, false tance and rejection rates, robustness, security, template protection, speed, user enrollmentprocesses, and interoperability Achieving high accuracy is imperative, particularly in dealingwith low-quality prints and variations in skin conditions, necessitating the development ofadvanced matching algorithms Balancing false acceptance and rejection rates is crucial forthe system’s reliability, ensuring accurate identification while minimizing the risk of unautho-rized access The technology must exhibit robustness to variations in fingerprints caused byfactors such as aging, injuries, and diverse sensor types, emphasizing the need for adaptableand resilient solutions Security vulnerabilities arise from the susceptibility to spoofing at-tacks, demanding the implementation of sophisticated anti-spoofing techniques to fortify thesystem against fraudulent attempts Secure template protection mechanisms are essential topreserve the privacy and integrity of stored fingerprint data, preventing unauthorized accessand breaches Enhancing matching speed without compromising accuracy is vital for deliver-ing a seamless and efficient user experience Streamlining user enrollment processes is equallycritical for widespread adoption, ensuring user-friendly and secure registration of fingerprints.The challenges of interoperability and standardization further underscore the need for com-mon industry standards to facilitate seamless integration across diverse fingerprint recognitionsystems and applications.

accep-1.3Motivation of our study

The primary objective of our project is to build and implement a fingerprint recognition tem This system will apply a machine learning methodology: Gabor Filter for data prepro-cessing and Oriented FAST and Rotated BRIEF (ORB) algorithm for recognition method.Themain goal of this project is to develop a system that will be able to recognize whether 2 fin-gerprints come from the same person or not For this purpose, the images are first collectedfrom a public data set Then digital imaging techniques are applied to the same images inorder to improve their quality Once the image is reprocessed, the so-called image is searchedcritical points that are later compared according to their Hamming distance.

sys-Page 3

2Method details

2.1Basic theory of fingerprint recognition

2.1.1 Ridge Patterns

Figure 1: Three basic fingerprint ridge patterns

• Distinctive Patterns: Fingerprint ridge patterns, comprising loops, arches, and whorls, areunique to each individual and serve as a fundamental basis for fingerprint recognition.• Individual Variation: The specific arrangement and combination of ridge patterns exhibit

considerable individual variation, contributing to the statistical rarity and uniqueness of eachperson’s fingerprints.

• Forensic Significance: Fingerprint ridge patterns play a crucial role in forensic tions, where the analysis of latent prints and the identification of minutiae points within thesepatterns aid in linking individuals to crime scenes.

investiga-2.1.2 Minutiae Points

Figure 2: Minutiae based extraction in fingerprint

• Definition and Types: Minutiae points are specific features in fingerprints, including ridgeendings and bifurcations, representing unique locations where ridges end, split, or converge.• Individual Variation: The arrangement and distribution of minutiae points are highly

individualized, contributing to the uniqueness of each person’s fingerprint and forming thebasis for reliable biometric identification.

• Fingerprint Template Creation: Minutiae points are crucial for creating a condenseddigital fingerprint template used in matching algorithms These templates capture essentialfeatures for accurate identification.

• Forensic and Security Applications: Minutiae points play a pivotal role in forensic tigations, aiding in the comparison of latent prints from crime scenes with known fingerprints.They are also considered in anti-spoofing measures to enhance security in fingerprint recog-nition systems.

inves-Page 5

2.1.3 Fingerprint Imaging

Figure 3: Fingerprint imaging procedure

• Capture Methods: Fingerprint imaging involves various capture methods, such as optical,capacitive, and ultrasonic sensors These methods detect and record the unique patterns ofridges and valleys on the fingertip.

• Valuable Data: During imaging, both the raised ridges and indented valleys on the skin’ssurface are captured This comprehensive data is essential for creating accurate and detailedrepresentations of the fingerprint.

• Enhancement Techniques: Preprocessing techniques, including image enhancement, areapplied to improve the clarity and quality of captured fingerprint images These enhancementscontribute to more precise feature extraction and analysis.

• Applications: Fingerprint imaging is used in various applications, including access controlsystems, mobile devices, forensic investigations, and identity verification The captured imagesare processed to create digital templates used for matching and identification.

2.2Data Preprocessing

2.2.1 Normalization

Figure 4: Fingerprint imaging procedure

The enhancement in fingerprint image remains as a vital step to recognize or verify the identityof person The noise is influenced during the acquisition of fingerprint image The poor qualityimages are captured and leads to inaccurate levels of discrepancy in values of gray level beside theridges and furrows because of non-uniformity of ink and contact of finger on scanner This poorquality images are affect to the minutiae extraction algorithm which may extract incorrect minutiaeand affect to the fingerprint matching during post-processing Normalization is the preprocessingstep for increase the quality of images by removing the noise and alters the range of pixel intensityvalues The mean and variance are used in process to reduce variants in gray-level values alongridges and valleys.

Page 7

2.2.2 Oriented Field Estimation

Figure 5: Fingerprint - Discrete orientation field - Orientation field estimated by mean squaremethod

Fingerprint orientation field estimation is a critical preprocessing step in fingerprint recognition,aiming to determine the local directionality of ridges Using methods such as Gabor filtering, itanalyzes small, overlapping regions of the fingerprint image to provide a localized and adaptiveestimation of ridge orientation The result is an orientation map that visually represents the localridge orientations across the entire fingerprint This orientation field is essential for subsequentprocesses, including ridge frequency analysis, minutiae extraction, and overall enhancement of ridgestructures The accuracy of orientation field estimation directly influences the precision of minutiaeextraction, a key factor in fingerprint recognition systems Overall, fingerprint orientation fieldestimation enhances the visibility of ridge patterns and contributes significantly to the accuracyand reliability of fingerprint recognition.

2.2.3 Ridge Frequency Estimation

Figure 6: Ridge Frequency

Ridge frequency estimation is a crucial step in fingerprint image processing, focused on determiningthe frequency of ridge patterns within a fingerprint Utilizing techniques like Fourier analysis, itinvolves analyzing the variations in the ridge spacings across different regions of the fingerprintimage The resulting ridge frequency map provides a visual representation of the local frequencies,aiding in subsequent analysis and enhancement Accurate ridge frequency estimation is vital fortasks such as fingerprint normalization, orientation field correction, and overall improvement of fin-gerprint recognition system performance It enhances the precision of feature extraction processes,contributing to the reliability and effectiveness of fingerprint identification.

Page 9

2.2.4 Gabor Filter

Figure 7: Gabor filter responses and binarized images according to the degree of orientation error

Fingerprint Gabor filtering is a powerful image processing technique used for enhancing the visibilityof ridge patterns in fingerprint images Gabor filters, inspired by mathematical functions, areemployed to capture specific frequency and orientation components of ridge structures The filtersare convolved with the fingerprint image, emphasizing ridge features while suppressing noise andirrelevant details This process aids in creating a clearer representation of the fingerprint, essentialfor subsequent analysis and recognition tasks Gabor filtering is particularly effective in capturingfine details and textures in fingerprint images, contributing to accurate feature extraction andmatching algorithms Its adaptability to different frequencies and orientations makes it a valuabletool for improving the overall quality and discriminative power of fingerprint recognition systems.

2.3Recognition Algorithm

2.3.1 Introduction to ORB

• FAST Keypoint Detection: ORB begins by employing the FAST algorithm for the rapididentification of keypoints in an image FAST identifies points where pixel intensities differsignificantly from their neighbors, providing a set of keypoints for further analysis.• BRIEF Descriptor Generation: After keypoint detection, the BRIEF algorithm is utilized

to generate binary descriptors for the keypoints BRIEF creates binary sequences by samplingpairs of pixel intensities and assigning binary values based on their relative magnitudes.• Rotation Invariance: ORB introduces rotation invariance by computing the dominant ori-

entation for each keypoint The binary descriptors are then rotated based on this orientation,ensuring consistent matching even when the images are rotated.

• Efficient Hamming Distance Matching: ORB employs the Hamming distance for cient binary descriptor matching The Hamming distance measures the dissimilarity betweentwo binary strings by counting the differing bits This approach significantly enhances com-putational efficiency compared to traditional methods using Euclidean distance.

effi-• Scale Invariance: While ORB primarily focuses on rotation and scale invariance, it alsoincorporates scale invariance to some extent, making it suitable for applications where thescale of features might vary.

• Applications: ORB is widely used in computer vision tasks such as object recognition, imagestitching, and visual odometry Its computational efficiency makes it particularly valuable inreal-time applications, including robotics and augmented reality.

The ORB algorithm’s ability to efficiently detect and describe features, along with its speed androbustness, has contributed to its popularity in various fields requiring rapid and reliable imageprocessing Its balance between accuracy and computational efficiency makes it well-suited forreal-world applications in both industry and academia.

Figure 8: ORB algorithm feature

Page 11

2.3.2 Adaptability to Fingerprint Recognition

• Texture and Structure: Fingerprint recognition heavily relies on the unique patterns andstructures of ridges and valleys While ORB is effective in capturing distinctive features intextured images, the specific requirements of fingerprint images may necessitate algorithmsdesigned specifically for ridge-based patterns.

• Rotation and Scale Invariance: While ORB introduces rotation invariance by computingdominant orientations, fingerprint recognition often requires more advanced methods to handlethe complex, non-linear deformations that can occur in fingerprint images Achieving robustrotation and scale invariance in fingerprint recognition may require additional adaptations.• Binary Descriptors: ORB uses binary descriptors, making it computationally efficient.

However, the binary nature might not capture the continuous and nuanced variations inridge patterns observed in fingerprint images Fingerprint recognition systems often employminutiae points and other specialized features.

• Adaptations Needed: Adapting ORB for fingerprint recognition might involve additionalprocessing steps and feature extraction methods specific to fingerprint patterns This couldinclude algorithms tailored for ridge orientation field estimation, minutiae extraction, andhandling complex ridge structures.

• Specialized Fingerprint Algorithms: Fingerprint recognition systems often rely on cialized algorithms designed explicitly for the unique characteristics of fingerprint images.These algorithms may include techniques such as ridge frequency analysis, local ridge orien-tation estimation, and advanced minutiae matching.

spe-In summary, while ORB is a versatile and efficient algorithm for general-purpose computer visiontasks, fingerprint recognition demands specialized approaches Researchers and practitioners oftenprefer algorithms specifically designed for fingerprint analysis, taking into account the unique char-acteristics and challenges posed by fingerprint images These specialized algorithms provide betteraccuracy and reliability in the context of fingerprint recognition systems.

2.3.3 Fast and Efficient Matching

The ORB (Oriented FAST and Rotated BRIEF) algorithm is known for its fast and efficient ing capabilities, making it suitable for real-time computer vision applications Here are key aspectsthat contribute to the fast and efficient matching of ORB:

match-• Real-Time Applications: ORB’s computational efficiency makes it well-suited for real-timeapplications such as object recognition, tracking, and augmented reality Its ability to perform