Road Sign Classification via Knowledge Graph and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current road sign recognition systems face challenges in efficiently annotating a vast number of road signs from diverse countries, requiring comprehensive datasets and posing a steep learning curve for human annotators due to varying conventions and sign features, leading to time-consuming and often poor results.
Innovation Solution
A system utilizing a knowledge graph and machine learning models to classify road signs and visual attributes, reducing the search space by providing ranked sign prototypes and assisting annotators, with human-machine collaboration for accurate classification and tracking across frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If human annotators manually classify road signs from multiple countries, then comprehensive annotation coverage is achieved, but annotation time and learning curve increase significantly
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary between road sign images and annotators. The knowledge graph stores structured information about road signs from multiple countries, including their visual characteristics and conventions. When annotating, the system queries the knowledge graph to retrieve relevant candidate signs based on detected visual attributes, thereby mediating the annotation process and reducing the time annotators need to spend learning and recalling signs from memory
Solution Approach 2:
The system performs preliminary classification by detecting visual attributes of road signs (such as shape, color, and text) before presenting them to annotators. Based on these detected attributes, the system pre-queries the knowledge graph to retrieve a limited set of candidate signs that match the observed characteristics. This preliminary action filters out irrelevant signs, allowing annotators to focus only on plausible candidates and significantly reducing annotation time
2Adaptability or versatility
If the candidate pool includes all road sign classes, then comprehensive classification is possible, but search space and annotation complexity increase
Solution Approach 1:
The knowledge graph is organized with local quality, where different sections store information about road signs from different countries and conventions. When a road sign image is being annotated, the system queries only the relevant local section of the knowledge graph based on detected attributes (such as geographic location or visual characteristics), rather than searching through all possible road sign classes globally. This localized approach maintains comprehensive classification coverage while reducing the effective search space and system complexity
3Measurement precision
If annotators review all possible road sign conventions, then accurate classification is achieved, but the learning curve becomes too steep
Solution Approach 1:
The system implements feedback by using detected visual attributes to query the knowledge graph and retrieve candidate signs that match those attributes. Instead of requiring annotators to review all possible road sign conventions, the system provides feedback in the form of a filtered candidate list that is already aligned with the observed visual characteristics. This feedback mechanism guides annotators toward the correct classification without requiring them to have comprehensive knowledge of all road sign conventions
Data Source
AI summary
A method of road sign classification utilizing a knowledge graph, including detecting and selecting a representation of a sign across a plurality of frames, outputting a prompt initiating a request for a classification associated with the representation of the sign, classifying one or more images including the sign, querying the knowledge graph to obtain a plurality of road sign classes with at least one same attribute as the sign, and classifying the sign across the plurality of frames in response to a confidence level exceeding a threshold.