Text String Extraction via Segmented Image Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in automating the collection of text-based information, such as expiry dates, from retail items due to inconsistent positioning, varying formats, and difficulties in distinguishing target text strings from non-target strings using optical character recognition.
Innovation Solution
A data capture device equipped with a camera and a processor that captures images, selects a search area using an initial classifier to detect associated strings, processes the area via a primary classifier to identify candidate target strings, and validates them based on criteria to ensure accurate extraction and display of the target text strings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If optical character recognition is used to extract text-based information from retail items, then automation of information collection is achieved, but accuracy deteriorates due to inconsistent positioning, varying formats, and difficulty in distinguishing target text strings from non-target strings
Solution Approach 1:
The patent segments the image processing into distinct stages: an initial classifier identifies candidate text strings and their locations, a search area is defined around each candidate, and then a primary classifier processes only that search area to determine if it's a target string. This segmentation allows the system to handle inconsistent positioning and varying formats by processing information in manageable portions rather than attempting to parse the entire image at once.
Solution Approach 2:
The patent applies preliminary action by using an initial classifier to first identify and locate candidate text strings before the main extraction process. This preliminary identification stage filters out non-target strings and establishes search areas in advance, allowing the primary classifier to focus only on relevant regions and improve overall extraction accuracy.
2Loss of information
If the entire image is processed to extract text strings, then completeness of information extraction is improved, but processing time and computational resources increase
Solution Approach 1:
The patent divides the image processing into segmented stages where the initial classifier first identifies candidate text strings and their locations, then defines search areas around each candidate. The primary classifier then processes only these localized search areas rather than the entire image, significantly reducing processing time while maintaining completeness through systematic coverage of all candidate regions.
Solution Approach 2:
The preliminary action of the initial classifier identifies and locations candidate text strings before the main processing stage. This advance identification allows the system to prepare search areas in advance and process only the necessary regions, avoiding unnecessary computation on the entire image while ensuring all potential target strings are captured.
Data Source
AI summary
A method of extracting a target text string includes: at a controller of a data capture device, obtaining an image of an item having the target text string thereon; at the controller, selecting a search area from the image; at the controller, processing the search area via a primary image classifier, to identify a candidate target string; at the controller, validating the candidate target string based on a validation criterion; and displaying the validated candidate target string via an output device of the data capture device.