Supervised OCR Training for Custom Forms Using Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) systems struggle to accurately determine which text corresponds to which fields in forms, requiring specific location definitions and uniform form designs for reliable results.
Innovation Solution
The system uses computer-vision image-segmentation and feature-detection techniques to partition forms into image segments, extract features, and classify them using a machine-learning model to identify field types and labels, allowing for the generation of key-value pairs without the need for OCR.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If specific location definitions and uniform form designs are used, then OCR reliability is improved, but form adaptability deteriorates
Solution Approach 1:
The form image is divided into multiple image segments using computer-vision image-segmentation techniques. Each segment represents a potential field or text region, allowing the system to process different parts of the form independently and accurately identify text locations without requiring uniform form designs
Solution Approach 2:
The patent replaces traditional OCR-based text identification with a machine-learning model that uses computer-vision feature-detection techniques. Instead of relying on OCR to recognize and match text, the system uses image segmentation and feature extraction to directly identify and classify form fields, eliminating the need for specific location definitions and uniform form designs
2Productivity
If traditional OCR is used to convert forms to text, then text extraction is achieved, but field association accuracy deteriorates
Solution Approach 1:
The patent introduces an intermediary machine-learning model between image segmentation and field identification. This model takes features from segmented images and classifies them to associate text with corresponding fields, serving as a mediator that connects visual information with semantic meaning without relying on traditional OCR text recognition
Solution Approach 2:
The system extracts only the essential features needed for field identification from form images using computer-vision feature-detection techniques, rather than converting entire forms to text through OCR. This selective extraction of relevant features improves both processing efficiency and field association accuracy by focusing only on discriminatory visual characteristics
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosed technology is generally directed to optical character recognition for forms. In one example of the technology, optical character recognition is performed on a plurality of forms. The forms of the plurality of forms include at least one type of form. Anchors are determined for the forms, including corresponding anchors for each type of form of the plurality of forms. Feature rules are determined, including corresponding feature rules for each type of form of the plurality of forms. Features and labels are determined for each form of the plurality of forms. A training model is generated based on a ground truth that includes a plurality of key-value pairs corresponding to the plurality of forms, and further based on the determined features and labels for the plurality of forms.