Character Recognition Device Using Degradation-Based Dictionary Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition methods face challenges in accurately and efficiently recognizing degraded character images, particularly in outdoor environments where image size and blur can vary, leading to increased processing time and suboptimal dictionary selection.
Innovation Solution
A character recognition device that stores multiple dictionary databases corresponding to different degradation levels, determines the usage order based on the calculated degradation degree of the input image, and sequentially uses dictionary databases to achieve accurate and fast recognition, ending the process when prescribed conditions are met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple dictionary databases are prepared for different degradation levels and all are used for recognition, then recognition precision is improved, but processing time increases
Solution Approach 1:
The patent pre-calculates and stores the correspondence relationship between degradation degrees and degradation levels before recognition. When recognition is performed, the system calculates the degradation degree of the input image, queries the pre-stored correspondence to determine the appropriate dictionary database, and performs recognition using only that specific dictionary. This preliminary preparation eliminates the need to test multiple dictionaries during recognition, significantly reducing processing time while maintaining high recognition precision.
2Productivity
If a single dictionary database is used for recognition, then processing time is reduced, but recognition precision decreases for degraded images
Solution Approach 1:
The patent creates multiple dictionary databases, each corresponding to a different degradation level. Instead of using a single dictionary for all images, the system adapts the dictionary selection to match the degradation characteristics of the input image. By changing the parameter of dictionary selection based on image degradation degree, the system maintains high processing speed while achieving accurate recognition even for degraded character images.
3Ease of manufacture
If the camera is fixedly arranged on a lamp-post for multi-lane road recognition, then device cost is reduced, but character image quality varies and degradation occurs
Solution Approach 1:
The patent applies different dictionary databases to different regions (degradation levels) based on the local characteristics of each character image. Instead of using a uniform recognition approach for all images, the system evaluates the degradation degree of each image and selects the appropriate dictionary from multiple options. This local adaptation allows the fixed camera system to maintain high recognition precision across varying image qualities from different lanes and distances.
Data Source
Figure 1(a)~1(b)
Figure 2
Figure 3
AI summary
A character recognition device comprising: a storage means storing a plurality of dictionary databases corresponding to each of a plurality of degradation levels, said databases being generated from images wherein an original character image has been degraded using a plurality of different degradation levels; an image input means that receives input of images; an extraction means that extracts character images from said images; a usage order determination means that determines an usage order for dictionary databases on the basis of degradation degrees for extracted character images; and a character recognition means that uses dictionary databases in accordance with said usage order, recognizes characters included in the extracted character image and, if recognition results for the character recognition fulfil prescribed conditions, outputs the recognition results without performing recognition using the dictionary database next in said usage order. As a result, accurate and fast recognition of degraded character images can be performed.