Graphical Object Recognition Using Geometric Element Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for image recognition on bit-mapped images, such as document images, are inefficient due to the use of full-sized images and models, which reduces productivity and reliability, and lacks noise immunity.
Innovation Solution
The method involves describing recognized objects as sets of standard elements with simple geometrical forms, allowing for reliable identification and recognition by comparing the object image with a classifier containing descriptions of these elements and their spatially parametrical correlations, enabling the composition of more complex elements from simpler ones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If full-sized images and models are used for comparison, then recognition reliability is maintained, but productivity is reduced
Solution Approach 1:
The patent divides the image and model into smaller elementary geometric elements (points, lines, curves) instead of using full-sized images. Each object is represented as a set of these basic elements with specific spatial and metric characteristics, allowing for faster comparison while maintaining recognition accuracy through the integrity principle that preserves geometric relationships.
2Productivity
If simple geometric structures are used for object representation, then productivity increases, but noise immunity is reduced
Solution Approach 1:
The patent enhances simple geometric elements by assigning them specific metric characteristics (length, angle, curvature radius) and spatial relationships. This parameter enrichment allows the system to distinguish between simple geometric patterns and noisy variations, maintaining noise immunity while preserving the computational efficiency of using basic geometric forms.
3Ease of operation
If elementary geometric elements are used as standard elements, then ease of identification is improved, but object complexity is reduced
Solution Approach 1:
The patent creates composite object descriptions by combining multiple elementary geometric elements with specific spatial and metric characteristics. Each object is represented as a composite structure of standard elements (points, lines, curves) arranged in specific configurations, allowing complex objects to be described through combinations of simple, easily identifiable building blocks.
Data Source
AI summary
The present invention discloses a method to increase reliability, correctness of objects recognition processing by performing a recognized object description as a set of special standard elements along with the spatial and parametrical correlation thereof. The said standard elements are preliminarily assigned graphic structures of elementary form and of easy identification and recognition. They may be provided with spatial and/or parametric details and thus may describe any object on the image including characters of text.