Camera Character Recognition Using Location Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mobile phone camera-based character recognition systems face challenges in accurately recognizing store names due to varying fonts, backgrounds, lighting, and capturing angles, leading to incorrect recognition results and difficulty in searching for additional information.
Innovation Solution
An apparatus and method that utilize location information from a GPS to extract store names within a predetermined radius from a nationwide database and compare them with character information from a captured image, employing a capturer, pre-processor, recognizer, and post-processor to enhance text recognition accuracy by setting character combination ranges and prioritizing recognition results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional character recognition methods are used with dictionary functions, then common text can be recognized, but store names (proper nouns) cannot be recognized accurately
Solution Approach 1:
The system pre-extracts store names from a database within a predetermined radius around the user's current location before the recognition process. This preliminary preparation creates a customized local dictionary that enables accurate recognition of store names (proper nouns) that would otherwise be unrecognized by conventional dictionary functions.
Solution Approach 2:
Instead of using a universal dictionary for all text recognition, the system creates a location-specific dictionary containing only store names relevant to the user's current area. This local customization allows the system to accurately recognize proper nouns specific to the local environment while maintaining overall recognition accuracy.
2Adaptability or versatility
If character recognition is performed on all captured text, then comprehensive text can be identified, but recognition accuracy decreases due to various fonts, backgrounds, lighting, and angles
Solution Approach 1:
The system extracts only the text area from the captured image that contains potential store names, rather than attempting to recognize all text in the image. This extraction focuses the recognition process on relevant regions, improving accuracy by reducing the impact of varying fonts, backgrounds, lighting, and capturing angles.
Solution Approach 2:
The system performs preliminary text area extraction and segmentation before character recognition. By pre-processing the image to isolate text regions and segment characters, the system prepares optimized input for the recognizer, thereby improving recognition accuracy despite variations in capturing conditions.
3Adaptability or versatility
If a nationwide database is searched for text matching, then comprehensive store names can be found, but processing time increases significantly
Solution Approach 1:
The system filters the nationwide store name database by geographic location, extracting only store names within a predetermined radius of the user's current position. This local filtering dramatically reduces the number of database entries that need to be compared with recognized characters, thereby reducing processing time while maintaining comprehensive coverage of relevant store names.
Solution Approach 2:
The system pre-extracts store names from the nationwide database based on location information before the character recognition results are finalized. This preliminary extraction creates a reduced, location-specific candidate set that speeds up the matching process while ensuring all potentially relevant store names are considered.
4Device complexity
If character recognition is performed without location information, then the system is simpler, but recognition accuracy for store names decreases
Solution Approach 1:
The system introduces location information (GPS coordinates) as an intermediary element that bridges the captured image and the store name database. This intermediary enables the system to filter and select relevant store names based on geographic proximity, significantly improving store name recognition accuracy without adding substantial complexity to the overall system structure.
Data Source
AI summary
A method is provided for recognizing characters of an image captured using a camera in a mobile terminal. The camera captures an image of a signboard. A text area is extracted from the captured image. Characters are recognized from the extracted text area. At least one text is generated by combining the recognized characters. The at least one text is compared with at least one store name within a predetermined range. A comparison result is output.