Iterative Identity Document Classification on Mobile Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile devices face challenges in processing images of identity documents due to limited processing power and image resolution, as well as issues with aspect ratio and dimension consistency, nonlinear sensor arrays, and variable illumination conditions, which hinder efficient image capture and processing for business workflows.
Innovation Solution
A method involving iterative classification of identity documents using feature vector data, where the classification process determines the document's class and subclass, enabling the integration of image capture and processing with business workflows on mobile devices, independent of universal standards like MICR characters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional scanner-based image processing algorithms are used on mobile devices, then image processing capability is improved, but processing time and computational cost become prohibitively high
Solution Approach 1:
The patent segments the image processing task into distinct phases: initial image capture and basic preprocessing on the mobile device, followed by selective transmission of processed data to a server for more complex analysis. This segmentation allows the mobile device to handle time-critical operations locally while offloading computationally intensive tasks to the server, thereby reducing overall processing time without sacrificing processing capability.
Solution Approach 2:
The patent performs preliminary actions by pre-processing images on the mobile device before transmission, including basic enhancements and extractions of key features. This preliminary processing reduces the complexity of subsequent server-side processing and enables faster initial responses, effectively reducing the perceived processing time for users while maintaining high-quality analysis.
2Ease of operation
If mobile device cameras are used to capture document images, then portability and ease of use are improved, but image quality and dimension consistency deteriorate
Solution Approach 1:
The patent implements feedback mechanisms where the system analyzes captured images for quality metrics such as focus, lighting, and alignment. Based on this analysis, the system provides real-time feedback to the user through the mobile interface, guiding them to adjust their capture technique. This feedback loop enables users to improve their capture skills over time, thereby enhancing image quality consistency while maintaining the portability advantages of mobile devices.
Solution Approach 2:
The patent dynamically adjusts processing parameters based on the specific characteristics of each captured image. Rather than applying fixed processing algorithms, the system analyzes image quality metrics and adapts processing intensity, filtering parameters, and enhancement techniques accordingly. This parameter adaptation allows the system to maintain high image quality consistency across varied capture conditions while preserving the ease of use provided by mobile devices.
3Measurement precision
If iterative classification with feature vector comparison is implemented, then classification accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the classification process into iterative stages, where each stage focuses on specific features or document types. Rather than performing complete feature extraction and comparison for all possible document types simultaneously, the system divides classification into hierarchical levels, processing only relevant features at each stage. This segmentation reduces computational complexity while maintaining high classification accuracy through progressive refinement.
Solution Approach 2:
The patent applies partial classification actions by performing basic classification on the mobile device for common document types, and only transmitting images requiring more complex classification to the server. This partial action approach handles the majority of cases efficiently on-device, reducing overall computational complexity while maintaining high accuracy for edge cases that require full iterative classification with feature vector comparison.
Data Source
AI summary
A method involves: receiving an image comprising an ID; iteratively classifying the ID; and driving at least a portion of a workflow based at least in part on the classifying; wherein at least some of the classification iterations are based at least in part on comparing feature vector data, wherein a first classification iteration comprises determining the ID belongs to a particular class, and wherein each classification iteration subsequent to the first classification iteration comprises determining whether the ID belongs to a subclass falling within the particular class to which the ID was determined to belong in a prior classification iteration. Related systems and computer program products are also disclosed.