ID Card Recognition via Neural Network Outline Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ID card recognition methods based on image processing are vulnerable to environmental factors like shading and background interference, leading to slow recognition speeds and inaccuracies, particularly in capturing ID card images.
Innovation Solution
A deep learning-based method using multiple neural network models for extracting the ID card outline, modifying the image into a reference form, and enhancing text information to improve recognition accuracy and speed, while also determining the validity of the ID card image by identifying recognition inhibitory factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If OCR technology is used for character recognition in ID cards, then character recognition can be automated, but recognition accuracy deteriorates under environmental factors like shading and background interference
Solution Approach 1:
The patent transforms the ID card image into a reference form by detecting corner positions and applying geometric transformation parameters. This changes the spatial parameters of the image to eliminate shading and background interference, thereby improving recognition accuracy while maintaining automation
Solution Approach 2:
The patent introduces an intermediary processing step between image capture and OCR recognition. The corner detection and image transformation process acts as a mediator that prepares the image by removing environmental distortions, enabling accurate automated recognition
2Device complexity
If existing image processing methods are used for ID card recognition, then the system structure remains simple, but recognition speed becomes slow
Solution Approach 1:
The patent divides the ID card image processing into distinct segments: corner detection, outline extraction, image transformation to reference form, and text recognition. This segmentation enables parallel processing and optimizes each stage, significantly improving recognition speed without requiring complex hardware
3Ease of manufacture
If traditional OCR methods are applied to ID card images, then implementation is straightforward, but recognition accuracy deteriorates due to shading by hand or shadow
Solution Approach 1:
The patent performs preliminary actions before OCR recognition by detecting corner positions, extracting outline regions, and transforming the image to a reference form. This preliminary processing removes shading and shadow effects caused by hand holding, ensuring accurate text recognition while keeping the overall system easy to implement
Data Source
AI summary
A method for recognizing an identification (ID) card of a user terminal using deep learning includes extracting an outline region of the ID card included in an input image using a first neural network model, modifying the ID card image of the image to a reference form using at least a partial value of the extracted outline region, and determining whether the modified ID card image is valid and recognizing text information in the valid ID card image. A recognition rate of the ID card may be increased by modifying an ID card image into a reference form using a trained neural network model.