Document Boundary Detection Using Visual Similarity Between Pages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual insertion of separator pages between documents in batches is time-consuming and expensive, and existing automated systems rely on character recognition or probabilistic methods that are inefficient for determining document boundaries.
Innovation Solution
A system and method using a neural network to autonomously separate documents by determining document boundaries based on visual similarity between pages, trained on pairs of consecutive pages to identify probabilities of belonging to the same or different documents, with indications of document boundaries provided if the probability exceeds a certain threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual insertion of separator pages is used to separate documents in batches, then document separation accuracy is improved, but processing time and cost increase significantly
Solution Approach 1:
The patent replaces the mechanical manual insertion process with an automated computer-based system that uses machine learning algorithms to detect document boundaries and automatically insert separator pages, eliminating the need for manual labor while maintaining high accuracy
Solution Approach 2:
The system enables documents to separate themselves through automated boundary detection and separator insertion without requiring human intervention, allowing the system to serve itself by processing batches autonomously
2Extent of automation
If character recognition methods are used to determine document boundaries, then automation is improved, but efficiency and accuracy deteriorate due to complexity
Solution Approach 1:
The patent extracts and focuses only on the essential visual features needed for boundary detection, removing unnecessary character recognition steps and complex processing, thereby improving efficiency while maintaining automation
Solution Approach 2:
The system segments the document analysis task into distinct stages: visual feature extraction, boundary probability calculation using trained models, and separator insertion decisions, allowing each stage to be optimized independently for efficiency
3Extent of automation
If probabilistic methods are used to determine document boundaries, then automation is improved, but measurement precision deteriorates
Solution Approach 1:
The system performs preliminary training of machine learning models on labeled document data before actual boundary detection, establishing accurate probability thresholds in advance that improve precision during automated operation
Solution Approach 2:
The system uses feedback from trained models that learn from labeled examples to continuously improve boundary detection accuracy, with the trained models providing refined probability assessments that enhance measurement precision
Data Source
AI summary
A system for separating documents in a batch of unseparated documents. In one example, the system comprises a scanner, a display, and an electronic processor. In another example, the system comprises an electronic source, a display, and an electronic processor. The electronic processor is configured to receive, as input, a batch of unseparated documents and apply, image processing to each page in the batch. The electronic processor is also configured to determine, for each pair of consecutive pages in the batch of documents, a probability that pages of the pair of consecutive pages belong to different documents using a predictive model. The electronic processor is further configured to generate a batch of separated documents by providing an indication of a document boundary if the probability generated by the predictive model is above a predetermined threshold.