Receipt Identification Using Neural Network Region Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing receipts require manual recording of information, which is time-consuming and prone to errors, and lack efficient automated solutions for extracting relevant details such as time information, store names, and payment amounts.
Innovation Solution
A receipt identification method using neural network-based region and character identification models to automatically identify and extract time information, store names, and payment amounts from receipt images by identifying row regions and using preset keywords to determine the total amount region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual recording of receipt information is used, then操作简单 (operation simplicity) is maintained, but processing efficiency is low and error-prone
Solution Approach 1:
The patent replaces the manual mechanical recording process with an automated optical character recognition (OCR) system. The OCR device captures receipt images and automatically converts them into structured data, eliminating the need for manual typing and data entry while significantly improving processing efficiency and accuracy.
Solution Approach 2:
The system enables self-service by allowing the receipt information extraction to occur automatically without human intervention. The OCR technology processes receipt images independently, extracting relevant information such as merchant names, amounts, and dates automatically, thereby improving productivity without adding operational complexity.
2Productivity
If automated OCR recognition is implemented, then processing efficiency is improved, but recognition accuracy deteriorates due to varied receipt formats
Solution Approach 1:
The patent segments the receipt image processing into distinct regions (header, body, footer) and identifies key information fields separately. By dividing the receipt into structured zones and processing each region with targeted recognition rules, the system maintains high accuracy despite format variations while preserving automated processing efficiency.
Solution Approach 2:
The system dynamically adjusts recognition parameters based on detected receipt characteristics. Different recognition strategies are applied to different receipt types, and the system adapts parameters such as character encoding, font recognition thresholds, and field extraction rules to match the specific format being processed, thereby maintaining accuracy across diverse formats.
3Loss of information
If detailed information extraction is performed, then data completeness is improved, but processing time increases
Solution Approach 1:
The patent extracts only the essential information fields from receipts (merchant name, amount, date, and other key details) rather than processing the entire document. By selectively extracting only the most important data elements needed for accounting purposes, the system achieves high information completeness for critical fields while minimizing processing time.
Solution Approach 2:
The system performs partial processing by focusing on key information fields first, and only processes additional detailed information when necessary. This selective approach ensures that essential data is always captured completely while avoiding unnecessary processing of less critical information, thereby balancing completeness with processing efficiency.
Data Source
AI summary
The disclosure provides a receipt identification method, apparatus, electronic device and computer-readable storage medium. The method includes: obtaining (S101) an image of the receipt to be identified; identifying (S102) each row region of the receipt in the image by using a region identification model, wherein the row region is the region where each line of characters is located; identifying (S103) the character content in each row region by using the character identification model; determining (S104) time information, store name and payment amount on the receipt according to the character contents in each row region. The solution provided by the disclosure can automatically identify the relevant information on the receipt.