Identity Document Glyph Authentication Using Discriminator Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The verification of identity documents is complicated due to variations in models, types, and authenticity issues, especially when close to the document's manufacturing date, requiring human and computer analysis with extensive adaptation, and existing methods like OCR focus on character recognition rather than individual glyph treatment.
Innovation Solution
A deep learning method using a machine learning algorithm with a discriminator neural network to classify and authenticate identity documents by processing individual glyphs from predefined zones, employing unsupervised learning and neural networks like convolutional and recurrent networks, with data augmentation and blurring to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine reading and character recognition techniques are used to extract data from identity documents, then data extraction speed is improved, but reliability deteriorates due to susceptibility to document defects and image quality variations
Solution Approach 1:
The patent introduces an intermediary verification mechanism where a second set of reading heads reads the same data field from a different physical location on the document. This intermediary approach allows cross-validation of data extracted by the first reading heads, thereby improving reliability without sacrificing the automated extraction speed. The intermediary verification step acts as a mediator that reconciles the speed-reliability contradiction.
Solution Approach 2:
The system performs preliminary actions by pre-positioning multiple reading heads at different locations on the document before the actual data extraction process. This preliminary arrangement enables parallel reading operations and facilitates quick verification without adding significant processing time, thus maintaining productivity while enhancing reliability through pre-planned redundancy.
2Reliability
If multiple reading heads are used to verify data from different locations, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by designing reading heads that can function in multiple roles: primary data extraction and verification. The same type of reading head component is used for both initial data capture and subsequent verification, reducing the need for specialized components and simplifying the overall device architecture while maintaining reliability improvements.
Solution Approach 2:
The system changes parameters by utilizing different physical locations on the document rather than adding complex verification hardware. By varying the spatial parameter (reading from different locations) rather than increasing hardware complexity, the system achieves reliability improvement through a simpler approach that maintains device straightforwardness.
3Reliability
If data is read from multiple locations on the document, then verification capability is improved, but processing time increases
Solution Approach 1:
The patent ensures continuity of useful action by organizing the reading process so that multiple reading heads operate simultaneously and continuously during the document verification process. Rather than sequential reading that would increase processing time, the system maintains continuous parallel operation, eliminating idle time and ensuring that verification occurs without significant time penalty.
Solution Approach 2:
The system applies partial action by selectively verifying only critical data fields rather than performing exhaustive verification on all document data. This partial verification approach maintains reliability for essential information while minimizing the time investment required, avoiding excessive processing time while still achieving verification goals.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to an image processing method for an identity document that comprises a data page, the method comprising steps of: - acquiring a digital image of the data page of the identity document. It is essentially characterized in that it furthermore comprises steps of: - assigning a class or a superclass to the candidate identity document by automatically classifying said digital image using an automatic learning algorithm implemented beforehand on a set of reference images during a learning phase, - processing the digital image so as to obtain a set of at least one intermediate image the weighting of which is less than or equal to the weighting of said digital image, - discriminating said intermediate image using a discriminator neural network, and - transmitting an output signal at the output of the discriminator neural network, the value of which is representative of the probability of the candidate identity document being an authentic document or a counterfeit.