Logo Recognition via Hierarchical Visual Word Indexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer vision technologies face challenges in efficiently recognizing objects, such as logos, in images due to the vast amount of data processing required, especially when dealing with varying orientations, scales, and distortions, which affects the robustness of feature matching across different images.
Innovation Solution
A process involving the extraction of local features from reference images, their representation as edge and triangle indices, and a hierarchical search approach to identify logos in sample images, using techniques like SIFT and RANSAC, allows for robust recognition across different perspectives and sizes by creating a vocabulary of visual words and employing a cascaded index structure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional image recognition methods are used to process enormous image data at high rates, then recognition accuracy may be maintained, but processing speed and efficiency deteriorate
Solution Approach 1:
The patent segments the image recognition process into multiple hierarchical stages: (1) extracting local features and representing them as visual words, (2) building edge and triangle indices from these visual words, (3) performing coarse matching using edge indices, and (4) performing fine matching using triangle indices. This segmentation allows parallel processing of different image regions and features, significantly improving processing throughput while maintaining recognition accuracy.
Solution Approach 2:
The patent performs preliminary actions by pre-processing reference images to extract local features, generate visual words, and build edge and triangle indices before actual recognition occurs. These pre-computed indices are stored and can be quickly queried during recognition tasks, eliminating the need to process entire images in real-time and thus reducing processing time while maintaining high recognition rates.
2Adaptability or versatility
If feature matching is performed across images with varying orientations, scales, and distortions, then recognition robustness improves, but matching accuracy deteriorates due to data transformation complexities
Solution Approach 1:
The patent applies local quality by extracting local features (edges, triangles) from specific regions of images rather than processing entire images globally. Each local feature is represented as a visual word with properties that capture its orientation, scale, and shape characteristics. This localized approach allows the system to adapt to local variations in orientation and scale while maintaining precise matching through the structured representation of visual words and their relationships in edge and triangle indices.
Data Source
AI summary
Subject matter disclosed herein relates to electronic image object or logo recognition.