Altered Document Detection Using Dynamic Region Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to accurately detect unauthorized alterations in alterable documents, such as cheques, without relying on standardized templates or historical data, due to non-standardized document formats and potential unauthorized modifications.
Innovation Solution
A system and method that utilize image segmentation and object detection models to identify regions of interest in alterable documents, reconfigure these regions to include surrounding features, and use convolutional neural networks to generate prediction values on potential fraud, based on image data alone, without prior knowledge of document templates or historical data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If automated object detection is used to identify regions of interest without template matching, then the system can handle non-standardized document formats, but the precision of region boundary identification decreases
Solution Approach 1:
The system dynamically adjusts the region of interest boundaries by generating multiple candidate regions with different dimensions and positions. Instead of using fixed template-based regions, the object detection model generates variable regions that adapt to each document's unique layout, resolving the contradiction between handling non-standardized formats and maintaining precise boundary identification.
Solution Approach 2:
The system changes the parameters of region boundaries by generating multiple candidate regions with varying dimensions, positions, and aspect ratios. This allows the system to adapt to different document formats while maintaining accurate identification of alterable parameters through subsequent verification steps.
2Measurement precision
If the region of interest is tightly bounded to only include the alterable parameter, then the detection precision for that parameter improves, but background information that could provide fraud detection cues is excluded
Solution Approach 1:
The system segments the document into multiple regions: a primary region of interest containing the alterable parameter and surrounding context, and additional candidate regions that may contain relevant background information. This segmentation allows the system to focus detection efforts on the primary region while preserving access to background cues in candidate regions for comprehensive fraud analysis.
Solution Approach 2:
The system implements a nested region structure where multiple candidate regions are generated at different scales and positions. The primary region of interest is nested within a larger context region, which itself is nested within the full document. This nested structure allows simultaneous access to both precise parameter boundaries and broader background information for fraud detection.
3Reliability
If multiple candidate regions are generated and evaluated, then the fraud detection accuracy improves, but the computational processing time increases
Solution Approach 1:
The system generates multiple candidate regions but does not evaluate all of them with equal computational resources. Instead, it performs a primary evaluation on the most likely region and only conducts additional verification on candidate regions that show potential anomalies or have high relevance scores, reducing unnecessary computational overhead while maintaining high detection accuracy.
Solution Approach 2:
The system performs preliminary filtering and scoring of candidate regions before conducting full fraud detection analysis. By pre-evaluating regions based on basic features and confidence scores, the system identifies and prioritizes the most promising candidates for detailed analysis, thereby reducing overall processing time while maintaining high accuracy.
Data Source
AI summary
Systems and methods of electronic altered document detection. The system may conduct operations of a method to: retrieve image data representing an alterable document and determine a target region of interest representing a boundary of an alterable parameter associated with the alterable document. The system may conduct operations to generate a tuned region of interest by calibrating the target region of interest based on an object detection model. The tuned region of interest may include a re-dimensioned boundary of the alterable parameter of interest. The object detection model may be prior-trained based on non-standardized alterable documents. The system may conduct operations to generate, based on the tuned region of interest, a prediction value representing whether the alterable document was subject to unauthorized alteration and transmit a signal representing the prediction value for identifying alterable documents for downstream document deconstruction operations.