Logo Recognition Under Warp, Occlusion, Blur, and Lighting Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current logo recognition systems face challenges in accurately detecting logos under conditions of warp, occlusion, blur, and varying lighting, require significant effort to train for new logos, and have limited capacity due to feature detector and neural network classifier limitations.
Innovation Solution
The system employs saliency analysis, segmentation techniques, and character stroke analysis to segment likely logo regions, uses multi-scale similarity comparison, and combines feature extraction, neural network classification, and optical character recognition to enhance detection accuracy. It also utilizes synthetic image training and transfer learning to quickly add new logos, and generates robust feature signatures for matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional feature detectors and neural network classifiers are used for logo recognition, then the system can identify logos, but the accuracy is limited and the number of recognizable logos is constrained
Solution Approach 1:
The patent segments the logo recognition task into multiple specialized components: rigid object detection for structured logos, flexible object detection for deformable logos, text recognition for text-based logos, and natural image search for scene-based logos. Each segment uses optimized features and classifiers tailored to its specific requirements, thereby improving both accuracy and the diversity of recognizable logos without overwhelming a single system
Solution Approach 2:
The system creates a universal logo recognition framework that handles multiple types of logos (rigid, flexible, text-based, and scene-based) through a unified architecture that selects and combines appropriate detection methods based on the input characteristics, enabling the system to recognize a broader range of logos with high accuracy
2Adaptability or versatility
If the system is trained to recognize new logos, then recognition capability improves, but significant time and effort are required to collect trainable images
Solution Approach 1:
The patent pre-computes and stores multiple types of features (color histograms, texture descriptors, shape descriptors, text features) for logos during an offline preparation phase. When new logos need to be recognized, the system can quickly match these pre-computed features against the database without requiring extensive retraining, significantly reducing the time and effort needed to add new logo recognition capabilities
Solution Approach 2:
The system creates comprehensive feature representations (copies) of logos in multiple formats and under various transformations (scaling, rotation, color variations) during the preparation phase. These feature copies enable rapid matching and recognition of new logos without requiring the system to relearn from raw images, thereby reducing training time while maintaining recognition capability
3Reliability
If the system attempts to detect logos under various conditions (warp, occlusion, blur, lighting variations), then robustness improves, but detection accuracy decreases due to challenging conditions
Solution Approach 1:
The patent applies different detection strategies and feature types to different regions and conditions: uses color and texture features for regions with good lighting, employs shape and structural features for occluded regions, applies text recognition for clear text-based logos, and uses rigid/flexible object detection appropriate to the logo type. This localized adaptation maintains high accuracy even when parts of the logo are affected by challenging conditions
Solution Approach 2:
The system combines multiple types of features (color, texture, shape, structure, text) and multiple detection methods into a composite recognition approach. By integrating results from different feature types and detection algorithms, the system achieves robust detection under varying conditions while maintaining overall accuracy through the complementary strengths of each component
Data Source
AI summary
Accurately detection of logos in media content on media presentation devices is addressed. Logos and products are detected in media content produced in retail deployments using a camera. Logo recognition uses saliency analysis, segmentation techniques, and stroke analysis to segment likely logo regions. Logo recognition may suitably employ feature extraction, signature representation, and logo matching. These three approaches make use of neural network based classification and optical character recognition (OCR). One method for OCR recognizes individual characters then performs string matching. Another OCR method uses segment level character recognition with N-gram matching. Synthetic image generation for training of a neural net classifier and utilizing transfer learning features of neural networks are employed to support fast addition of new logos for recognition.