Claim Document Character Recognition via Region Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current OCR technologies for recognizing characters in insurance claim documents suffer from low recognition precision, requiring significant human and material resources for verification, leading to inefficiencies and document recording mistakes.

Innovation Solution

A method involving region segmentation based on the frame line layout of claim document images, followed by analysis using a CNN or LSTM model to differentiate between easily recognizable and difficult regions, employing appropriate recognition engines or models for each, thereby improving character recognition precision.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If traditional OCR technology is used for character recognition in claim documents, then automation is improved, but recognition precision deteriorates

Engineering Contradiction:
ImproveautomationVSAvoidrecognition precision
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The patent divides the claim document into multiple regions based on frame line layout (e.g., claim number region, claim content region, signature region). Different recognition strategies are applied to different regions: simple OCR for clear regions, enhanced preprocessing for difficult regions. This segmentation allows the system to maintain high automation while improving overall recognition precision by treating different document areas with appropriate methods.

Inventive Principle:
Principle #1Segmentation

2Productivity

If traditional OCR technology is used for character recognition, then processing speed is improved, but recognition precision deteriorates, leading to more verification needs

Engineering Contradiction:
Improveprocessing speedVSAvoidrecognition precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies different quality levels of processing to different document regions. For regions with clear, standard formatting (like claim numbers), simple fast OCR is used. For regions with handwritten text or complex layouts (like claim content), more sophisticated preprocessing and multiple recognition passes are applied. This local quality approach maintains high processing speed for simple regions while ensuring high precision for difficult regions, reducing overall verification needs.

Inventive Principle:
Principle #3Local quality

3Device complexity

If uniform recognition processing is applied to all document regions, then device complexity is reduced, but recognition precision deteriorates

Engineering Contradiction:
Improvedevice complexityVSAvoidrecognition precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent implements a dynamic recognition system that automatically adjusts processing strategies based on document characteristics. The system first performs region segmentation, then analyzes each region's characteristics (handwritten vs. printed, clarity, layout complexity) to dynamically select appropriate recognition methods. This dynamic approach achieves high precision without requiring manually configured complex systems for each document type, as the system adapts automatically.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3432197B1Method and device for identifying characters of claim settlement bill, server and storage medium
Publication Date: 2022.07.06 PING AN TECH (SHENZHEN) CO LTD
  • EP3432197B1 patent drawingFigure 1
  • EP3432197B1 patent drawingFigure 2
  • EP3432197B1 patent drawingFigure 3

AI summary

The disclosure discloses a method, device and server of recognizing characters of a claim document, and a storage medium. The method includes: after the server receives a claim document image to be subjected to character recognition, performing region segmentation according to the frame line layout of a claim document frame format, thus obtaining one or multiple segmented regions; and calling a predetermined analysis model to analyze all the obtained segmented regions , and respectively performing character recognition on all the analyzed segmented regions by using a predetermined recognition rule to recognize characters in all the segmented regions. In the disclosure, in consideration of the influence of the claim document frame format on the recognition precision, the region segmentation is firstly performed according to the frame line layout of the claim document frame format before the character recognition, and then the character recognition is performed on all the segmented regions, so that the influence and the interference of frame lines in the document on the character recognition during uniform recognition of the characters in the whole claim document image are avoided, and the recognition precision for the characters in the claim document can be effectively improved.