Fingerprint Image Correction Model With Evaluation-Guided Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fingerprint images registered in databases often lack accurate corrections, leading to unreliable matching, as they are not corrected by forensic examiners, necessitating a method to automatically correct and detect uncorrected images with high accuracy.
Innovation Solution
A model generation apparatus and method that selects training data based on evaluation values to train and update a model for correcting fingerprint images, using machine learning to improve accuracy and detect uncorrected images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If forensic examiners manually correct skeleton lines in fingerprint images, then correction accuracy is high, but productivity is low due to the large number of images requiring correction
Solution Approach 1:
The patent replaces the manual mechanical correction process performed by forensic examiners with an automated image processing system that uses machine learning models. The system automatically detects skeleton lines in fingerprint images and corrects defects without human intervention, thereby maintaining high accuracy while dramatically increasing productivity by processing large volumes of images simultaneously.
Solution Approach 2:
The patent implements a self-correcting system where the image processing apparatus autonomously identifies and corrects skeleton line defects in fingerprint images without requiring manual intervention. The machine learning model continuously learns from training data to improve its self-correction capability, enabling the system to serve itself in maintaining and improving its correction accuracy over time.
2Productivity
If uncorrected fingerprint images are registered in the database, then registration speed is high, but reliability of matching decreases
Solution Approach 1:
The patent applies preliminary correction action by automatically processing and correcting fingerprint images before they are registered in the database. The system performs skeleton line detection and correction as a preliminary step, ensuring that only properly corrected images are registered. This maintains high registration speed while guaranteeing matching reliability through pre-registration quality assurance.
Solution Approach 2:
The patent implements a feedback mechanism where the image processing system continuously evaluates the quality of registered fingerprint images and automatically reprocesses those that do not meet correction standards. The system uses evaluation criteria to provide feedback on image quality and iteratively improves corrections, ensuring high reliability of registered images without significantly impacting registration speed.
3Extent of automation
If a machine learning model is trained to correct fingerprint images, then automation level increases, but training data selection and model updates increase system complexity
Solution Approach 1:
The patent segments the machine learning system into distinct functional modules: a training data selection unit that automatically identifies suitable training images, a model training unit that generates correction models, and a model update unit that refines existing models. This modular segmentation manages system complexity by allowing each module to operate independently with well-defined interfaces, while maintaining high automation through coordinated operation of these segmented components.
Data Source
AI summary
In a model generation apparatus, a training data selection means selects sets of training data from a plurality of fingerprint images. A learning means trains a model that corrects a fingerprint image, by using the sets of training data. An evaluation value calculation means calculates evaluation values of results acquired by inputting the plurality of fingerprint images into the trained model. A model update means updates a model to be trained, based on the evaluation values. Next, the training data selection means determines sets of training data to be selected from the plurality of fingerprint images based on the evaluation values.