matching algorithm of finger print in biometrics notes ppt
.Overview of fingerprint matching
Purpose: To compare a new fingerprint image (query print) against one or more stored templates to
verify or identify an individual.
General process:
1. Image acquisition: Capture a fingerprint using a sensor (e.g., optical, capacitive).
2. Image enhancement: Use filters to improve image quality by clarifying ridges and valleys.
3. Feature extraction: Isolate key, unique characteristics of the fingerprint pattern.
4. Matching: Compare the extracted features to a stored template to calculate a similarity
score.
5. Decision: Match or no-match is determined by comparing the similarity score to a
predefined threshold.
Minutiae-based matching
This is the most widely used and classical technique for fingerprint matching due to its high accuracy
with good-quality images.
Core concept
Focuses on the unique local ridge characteristics known as minutiae points, such as ridge
endings and bifurcations.
Each minutiae point is defined by its position (x, y), direction (
Matching process
1. Minutiae extraction: The fingerprint image is preprocessed (enhanced, thinned) to a 1-pixel-
wide ridge skeleton, from which minutiae points are extracted.
2. Alignment: The minutiae sets of the input and template fingerprints are aligned. This
involves applying geometric transformations (translation, rotation) to correct for placement
variations.
3. Local and global matching:
o Local matching: Analyzes the spatial relationships and direction differences between
nearby minutiae.
o Global matching: Matches the overall alignment of the minutiae sets to find the
maximum number of matched pairs.
4. Similarity scoring: A score is computed based on the number of matching minutiae pairs. A
higher number of matching points indicates a better match.
Advantages and disadvantages
Advantages: Efficient, widely adopted, and robust for high-quality images.
Disadvantages:
o Difficulty with low-quality fingerprints or partial prints, where minutiae extraction is
unreliable.
o The matching process is sensitive to non-linear distortions caused by skin elasticity.
Correlation-based matching
This approach overcomes some of the challenges of low-quality images by comparing the entire
fingerprint image pattern directly.
Core concept
Aligns two fingerprint images and computes the pixel-wise correlation to determine how
much they overlap.
The correlation value indicates the similarity between the two prints.
Matching process
1. Alignment: All possible alignments (translations and rotations) between the template and
query print are attempted to maximize the correlation score.
2. Correlation calculation: The correlation is calculated for each alignment. The highest score
corresponds to the best alignment.
3. Feature selection (optional): To reduce computational cost, the correlation can be applied
only to specific regions or feature sets.
Advantages and disadvantages
Advantages: Can handle poor-quality or partial prints more effectively than minutiae-based
methods, as it uses the richer grayscale information.
Disadvantages:
o Computationally intensive, as it requires testing all possible alignments.
o Less tolerant of significant non-linear distortions, as it treats the fingerprint as a rigid
object.
Feature-based (texture) matching
This method uses features like ridge flow and texture to create a unique fingerprint representation,
often combined with other techniques.
Core concept
Uses a bank of Gabor filters to capture the unique texture and orientation of fingerprint
ridges.
This creates a "FingerCode" which is a fixed-length feature vector.
Matching process
1. Image filtering: Gabor filters are applied to the image around a central reference point
(core).
2. Feature vector creation: The filtered results are compiled into a feature vector that
represents the fingerprint's unique pattern and texture.
3. Matching: The Euclidean distance between the two FingerCode vectors is used to determine
a matching score.
Advantages and disadvantages
Advantages: Computationally fast for matching, as it only requires a simple vector
comparison. It can also be more robust to partial prints and minor distortions.
Disadvantages: Reliant on accurate core point detection and may lose fine detail that
minutiae-based methods capture.
Hybrid and advanced techniques
Modern systems often combine multiple methods to improve accuracy and robustness.
Minutiae and texture fusion: Some algorithms combine minutiae information with texture-
based features to improve overall accuracy, leveraging the strengths of both methods.
Deep learning (CNN): Convolutional Neural Networks (CNNs) are increasingly used for
fingerprint matching. They can learn to extract features directly from the raw image data,
bypassing traditional preprocessing steps and offering high accuracy.
Delaunay triangulation: A topology-based method that constructs a unique geometric
pattern from minutiae points. This method can be more robust to non-linear distortion by
modeling the elastic deformation of the skin.