Virtual Fingerprint Image Generation for Overlapped Ridge Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fingerprint extraction technologies face challenges in accurately separating overlapped fingerprints, particularly when ridge directions are ambiguous, and may damage samples or have limited applicability based on fluorescence intensity or mass spectrometry.
Innovation Solution
A fingerprint extraction apparatus and method utilizing a machine learning module to generate virtual fingerprint images through primary and secondary image processing, allowing for the extraction of target fingerprints from real images without damaging the sample, regardless of fingerprint formation time or background patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mass spectrometry is used to separate overlapped fingerprints based on spectrums, then separation accuracy is improved, but the sample is damaged due to ionization process
Solution Approach 1:
The patent uses virtual fingerprint images as copies to train the machine learning model instead of requiring multiple physical samples. The virtual images are generated through image processing techniques that simulate various fingerprint overlapping scenarios, allowing the model to learn separation without damaging real fingerprint samples.
Solution Approach 2:
The patent replaces the physical/chemical ionization process of mass spectrometry with a computational machine learning approach. The deep learning model processes images directly to separate overlapped fingerprints, eliminating the need for sample ionization and thus preventing sample damage while maintaining separation capability.
2Measurement precision
If fluorescence intensity detection is used to separate overlapped fingerprints, then separation is achieved, but it is limited to fingerprints formed with time difference
Solution Approach 1:
The machine learning model is trained to handle multiple scenarios including fingerprints formed at different times, fingerprints with various ridge directions, and fingerprints overlapping with background patterns. This universal approach allows the same system to separate any type of overlapped fingerprints regardless of formation time or background conditions.
Solution Approach 2:
The patent changes the approach from detecting temporal parameters (fluorescence intensity over time) to analyzing spatial parameters (ridge directions and patterns) through image processing. This allows separation of fingerprints regardless of when they were formed, as the model focuses on the structural characteristics visible in the image rather than temporal formation differences.
3Measurement precision
If ridge direction-based separation is used for overlapped fingerprints, then separation is achieved, but accuracy decreases when ridge directions are ambiguous
Solution Approach 1:
The machine learning model uses feedback from training data that includes various ridge direction scenarios to improve its separation capability. During training, the model learns from examples where ridge directions are ambiguous by adjusting its parameters to correctly identify and separate fingerprints based on subtle pattern differences, thereby improving reliability in ambiguous situations.
Solution Approach 2:
The patent performs preliminary image processing and feature extraction to enhance ridge direction information before separation. The model pre-processes the overlapped fingerprint images to emphasize ridge patterns and directions, making ambiguous directions more distinguishable and improving subsequent separation accuracy.
4Device complexity
If traditional fingerprint extraction methods are used, then processing is simpler, but extraction speed and accuracy are reduced
Solution Approach 1:
The machine learning model performs periodic training with diverse virtual fingerprint data to continuously improve extraction speed and accuracy. The model is trained on various overlapping scenarios, ridge directions, and background patterns, enabling it to quickly and accurately separate fingerprints in real-world applications without requiring complex manual processing.
Data Source
AI summary
A fingerprint extraction apparatus includes a fingerprint generation module configured to generate first and second virtual fingerprint images, perform primary image processing on the first virtual fingerprint image and primary and secondary image processing on the second virtual fingerprint images, and generate virtual overlapped fingerprint images by combining the first and second virtual fingerprint images on which image processing is performed; a machine learning module configured to generate a learning model by performing machine learning using the virtual overlapped fingerprint images as input data; and a fingerprint extraction module configured to extract a fingerprint located vertically on a center of a real image by inputting the real image to a target fingerprint extraction learning model, and the primary image processing comprises image processing on a curve forming a fingerprint, and the secondary image processing comprises image processing on a location of the fingerprint in the image.