Postal OCR Collation Database for Address Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Postal OCR devices face inefficiencies in sorting postal items due to incomplete postal address databases, particularly lacking house number and door number registration, leading to increased reliance on video keying and repeated operator input for addresses that cannot be identified by OCR.
Innovation Solution
A postal optical character recognition device generates collation data from destination images, which includes graphical characteristics of the address, and registers this data with corresponding sort codes, allowing for the retrieval of previously determined sort codes for recurring destinations, reducing the need for repeated video keying.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the postal address database is updated to include all possible address formats, then the OCR recognition rate improves, but the database maintenance complexity and cost increase
Solution Approach 1:
The patent creates a collation database that stores images of destination fields alongside their sort codes. Instead of maintaining a comprehensive address database with all possible formats, the system captures and stores actual images from postal items, creating a copy-based reference system that naturally adapts to various address formats without requiring proactive database updates.
Solution Approach 2:
The system implements feedback by storing successful OCR recognition results and video keying results in the collation database. When the OCR fails to recognize an address, the system refers back to the collation database to find matching images and retrieve corresponding sort codes, creating a self-learning mechanism that improves recognition without manual database maintenance.
2Reliability
If video keying is performed for all unrecognized addresses, then the sorting accuracy improves, but the operator burden and processing time increase
Solution Approach 1:
The system performs preliminary action by pre-storing sort codes associated with destination field images in the collation database during video keying operations. This allows the system to quickly retrieve previously determined sort codes without requiring operators to re-key the same addresses, effectively caching results for future use.
Solution Approach 2:
The system creates a copy of the destination field image and stores it in the collation database with its corresponding sort code. This image copy serves as a reference that can be quickly matched against future postal items, eliminating the need for repeated video keying of identical or similar addresses.
3Measurement precision
If the collation database stores detailed address information, then the retrieval accuracy improves, but the storage requirements and data processing complexity increase
Solution Approach 1:
The patent extracts only the essential visual characteristics of destination fields needed for identification, storing images in a optimized format in the collation database. Rather than storing complete address databases with all possible variations, the system extracts and stores only the actual images that appear on postal items, reducing storage requirements while maintaining retrieval accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A keystroke saving device includes a collation data generating portion that acquires an image including information of a destination; reads in character strings from the image with a predetermined marker serving as a reference and cuts out a plurality of lines that those character strings show; and generates collation data that is identifier information corresponding to the destination address included in the image using characters at predetermined positions in each line and position information of each of the predetermined positions in each line; a collating portion that uses the collation data to look up address sorting information corresponding to the collation data, and in the case of having found it, reads out the address sorting information and outputs it; and a registering portion that, in the case of not finding the address sorting information, registers the collation data that the collation data generating portion has generated and address sorting information corresponding to the destination information included in the image in a corresponding manner.