Minutiae Graph Comparison for Interoperable Fingerprint Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fingerprint identification methods require fingerprint images, which may not be available in older databases, and are often not interoperable due to varying image formats and qualities, and methods using minutiae-encoded graphs struggle with identifying a single dactylogram among a set.
Innovation Solution
A method that projects minutiae coordinates into a higher-dimensional space using a trained projection model, applies graph neural networks for inference and encoding, and aggregates the results into a scalar score, enabling comparison of minutiae lists without requiring image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If fingerprint images are used for identification, then accurate comparison can be performed, but the method cannot be applied to older databases that only store minutiae coordinates
Solution Approach 1:
The patent transforms the identification approach by changing the input parameters from image-based features to minutiae coordinate-based features. The system accepts minutiae lists with coordinates and orientations as input, processes them through neural network models trained specifically on coordinate data, and produces identification results comparable to image-based methods. This parameter transformation enables compatibility with databases storing only minutiae coordinates while maintaining identification accuracy.
2Adaptability or versatility
If different image formats and qualities are used in databases, then more fingerprints can be stored, but interoperability between systems decreases
Solution Approach 1:
The patent extracts the essential identification features from fingerprint images during the enrollment phase, storing only the critical minutiae coordinates, orientations, and ridge flow directions in the database. This extraction approach eliminates the need to store complete images in various formats, enabling seamless interoperability between different systems while maintaining the ability to store and process a large quantity of fingerprints using a standardized, compact data structure.
3Productivity
If traditional screening algorithms are used to eliminate candidates, then processing speed improves, but precision in identifying the correct fingerprint decreases
Solution Approach 1:
The patent implements preliminary action by pre-training neural network models on comprehensive datasets during an offline phase. The models learn optimal feature representations and comparison metrics in advance. During online identification, these pre-trained models rapidly process minutiae lists and compute similarity scores without requiring iterative screening, thereby achieving both high processing speed and high precision through the preliminary learning phase.
4Measurement precision
If more computational resources are allocated for processing, then identification accuracy improves, but processing time increases
Solution Approach 1:
The patent replaces traditional mechanical comparison algorithms with neural network-based computational models. The neural networks, trained offline to learn optimal feature representations and decision boundaries, perform rapid similarity computations during online identification. This substitution leverages the parallel processing capabilities of neural networks and optimized tensor computations, achieving high identification accuracy with reduced processing time compared to traditional iterative comparison methods.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Method for comparing two dactylograms from the lists of their minutiae, said method takes, as input data, a first source matrix MS1 of the coordinates of each minutiae of a first list L1 of minutiae of a first dactylogram and a second source matrix MS2 of the coordinates of each minutiae of a second list L2 of minutiae of a second dactylogram, and provides, as output data, a correspondence score between the two lists L1, L2 of minutiae.