Brand Logo Detection via Image Text Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting illegal use of brand logos are ineffective when the logos are transformed or used with background images, leading to reduced detection performance.
Innovation Solution
A brand logo illegal use detection method that involves analyzing the logo to extract images and text, generating transformed forms of the logo, creating a machine learning dataset, and using a logo identification model to detect illegal use in suspect images, while also performing text analysis to determine the illegal use of logo text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional logo detection methods are used, then detection simplicity is maintained, but detection accuracy deteriorates when logos are transformed or used with background images
Solution Approach 1:
The system performs preliminary actions by generating multiple transformed versions of the logo (rotated, scaled, filtered, with different backgrounds) before the actual detection process. This creates a comprehensive machine learning dataset that prepares the detection model to recognize logos in various transformed states, thereby improving detection accuracy without requiring complex real-time analysis during actual detection
Solution Approach 2:
The detection system employs dynamic approaches by using machine learning models that can adapt to various logo transformations. The system dynamically adjusts detection parameters and uses multiple detection strategies (image-based and text-based) depending on the input characteristics, allowing accurate detection across diverse scenarios without requiring a separate detector for each transformation type
2Adaptability or versatility
If logo transformations and background images are used to avoid detection, then logo usage versatility is improved, but detection reliability deteriorates
Solution Approach 1:
The system preemptively generates training data covering all possible logo transformations (rotations, scales, filters, background combinations) that users might employ for versatile logo usage. By incorporating these transformations into the training dataset before model deployment, the system ensures that versatile logo usage does not compromise detection reliability
Solution Approach 2:
The detection system uses feedback mechanisms where detection results are continuously evaluated and used to refine the machine learning model. The system incorporates both image-based detection feedback and text-based verification feedback to continuously improve reliability, ensuring that even transformed logos with backgrounds are reliably detected
3Measurement precision
If only image-based detection is used, then detection process simplicity is maintained, but detection completeness deteriorates when logos are text-transformed
Solution Approach 1:
The detection system segments the logo detection task into two independent but complementary components: image-based detection for graphical logo elements and text-based detection for text-transformed logos. Each segment uses specialized detection methods optimized for its type, and the results are combined to achieve comprehensive detection completeness without requiring a single overly complex detection process
Data Source
AI summary
The present disclosure relates to a method, device, and computer program capable of detecting whether a brand logo is illegally being used by an unauthorized person using an image and text of the brand logo. The present disclosure provides a detection technique capable of accurately detecting illegal use of a brand logo even when variously transformed forms and background images are used to avoid legal use detection. According to embodiments of the present disclosure, since the logo identification model can be generated through the machine learning dataset based on the logo transformation images and the image illegal use information can be generated using the logo identification model, it is possible to more accurately detect the illegal use of the logo that is used in variously transformed forms. In addition, since the illegal use information on the logo text can be considered in addition to the logo image, it is possible to further increase the accuracy of the illegal use detection.