Graph Neural Encoding of Fingerprint Minutiae for Image-Free Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fingerprint encoding methods require fingerprint images, lack interoperability between databases, and are inefficient for identifying a single fingerprint from a set of candidates, especially in databases lacking images or with varying image formats and quality.
Innovation Solution
A method that encodes a list of minutiae into a fixed-size vector using a trained graph neural network, projecting and aggregating minutiae coordinates through a series of steps to create an encoding vector, independent of image presence and format, enhancing database interoperability and computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fingerprint images are used for encoding and identification, then identification accuracy can be maintained, but database interoperability deteriorates due to varying image formats and quality requirements
Solution Approach 1:
The invention extracts only the essential minutiae coordinates from fingerprint images, separating the identification task from image format dependencies. By encoding only the (x, y, θ) coordinates of minutiae points rather than full images, the system achieves format independence while maintaining identification accuracy through the preserved geometric relationships of fingerprint characteristics.
Solution Approach 2:
The encoding method creates a universal representation that works across different database formats and quality levels. The minutiae-based encoding scheme can process and identify fingerprints from various sources (different scanners, resolutions, formats) using the same coordinate encoding approach, making the system universally applicable without requiring format-specific processing.
2Reliability
If fingerprint images are stored in databases, then complete information is preserved, but storage space and processing time increase significantly
Solution Approach 1:
The invention extracts only the critical minutiae coordinates from complete fingerprint images, storing merely the (x, y, θ) values rather than full image data. This extraction preserves the essential identification information while dramatically reducing storage requirements and accelerating processing speeds by working with compact coordinate sets instead of large image files.
Solution Approach 2:
The invention creates a simplified copy of the fingerprint data in the form of minutiae coordinates rather than storing the original image. This coordinate-based representation serves as a sufficient copy for identification purposes, eliminating the need to store and process the complete image while maintaining identification reliability.
3Ease of operation
If traditional encoding methods are used that require fingerprint images, then identification can be performed, but the system cannot process databases lacking images or with varying formats
Solution Approach 1:
The encoding method is designed to be universally applicable to any database containing minutiae coordinates, regardless of whether the original images are present or what format they were in. The system processes (x, y, θ) coordinates directly, making it compatible with diverse database types and enabling operation on databases that lack images entirely.
Solution Approach 2:
The invention uses minutiae coordinates as an intermediary representation that bridges the gap between different database formats and the identification process. Rather than requiring direct access to images in various formats, the system mediates through the standardized coordinate representation, enabling seamless processing of heterogeneous database sources.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A computer-implemented method for encoding a list of minutiae of a fingerprint, said method taking, as input data, the coordinates associated with each minutia of a list of minutiae of a fingerprint, and providing, as output data, a fixed-size encoding vector representative of the list of minutiae of said fingerprint, the method comprising the following steps: (a) concatenating the coordinates of each minutia of the list of minutiae in the form of a source matrix of dimension; (b) projecting the source matrix of dimension to a space of a dimension greater than the dimension of said source matrix using a previously trained projection model to form a projected matrix of dimension; (c) inferring a dimension inference matrix by applying, to the intermediate matrix, a previously trained graph neural network;(d) aggregating the values of the inference matrix into a fixed-size vector using a previously defined aggregation model, said fixed-size vector being the fixed-size encoding vector representative of the list of minutiae of the fingerprint.;