Logo Detection in Video Streams Using Gradient Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition technologies face challenges in accurately identifying logos in video streams due to their dynamic and transparent nature, which complicates the use of convolutional neural networks (CNNs) that require extensive training and struggle with overlaying logos in changing video frames.
Innovation Solution
A method involving identifying areas within images using optical character recognition (OCR) and analyzing image gradient vectors to extract logo features, followed by using a machine learning model trained on received image data to accurately identify logos, even in video streams with dynamic and transparent logos.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If convolutional neural networks (CNN) are used to identify logos in video streams, then logo identification capability is provided, but training time and computational resources increase significantly
Solution Approach 1:
The patent segments the logo identification process into distinct stages: extracting color features from video frames, analyzing temporal variations in color values, and identifying logos based on these features. This segmentation eliminates the need for extensive CNN training by breaking down the complex recognition task into manageable feature extraction and comparison steps.
Solution Approach 2:
The system performs preliminary actions by pre-defining color feature extraction methods and temporal analysis approaches before actual logo identification. By establishing these preprocessing frameworks in advance, the system avoids the time-consuming CNN training phase while maintaining identification capability.
2Reliability
If CNN is trained to identify logos overlaying various image frames, then identification accuracy improves, but the amount of training data required increases to hundreds or thousands of images
Solution Approach 1:
The patent extracts essential color features from video frames and temporal variations independently of specific logo appearances. By focusing on these fundamental characteristics rather than training on numerous logo instances, the system achieves identification accuracy without requiring hundreds or thousands of training images.
Solution Approach 2:
The system changes the approach from learning logo-specific parameters through extensive training to analyzing temporal variations in color values. This parameter shift allows the system to identify logos based on their dynamic color characteristics rather than static visual patterns, reducing training data requirements.
3Loss of information
If logos with transparent regions are analyzed in video streams, then complete logo information is captured, but identification difficulty increases due to changing pixel values across frames
Solution Approach 1:
The patent applies periodic action by analyzing temporal variations in color values across multiple frames. This approach converts the challenge of changing pixel values into a detectable pattern, where periodic or consistent color variations indicate the presence of logo regions with transparent areas.
Solution Approach 2:
The system maintains continuous analysis of color features across video frames rather than processing individual frames independently. This continuity allows the system to track logo regions through temporal variations, preserving complete logo information while simplifying identification through consistent feature monitoring.
Data Source
AI summary
A method for identifying a logo within at least one image includes identifying an area containing the logo within the at least one image, extracting logo features from the area by analyzing image gradient vectors associated with the at least one image, and using a machine learning model to identify the logo from the extracted logo features, wherein the machine learning model is trained to identify at least one target logo based on a received image data containing the logo features.