Vehicle Trajectory Clustering for Knowledge Graph Ontology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intelligent traffic management systems face challenges in efficiently modeling and understanding high-level activities and root causes of events like traffic jams and accidents, due to differences in cultural and linguistic descriptions, which require collective intelligence and effective representation of contextual information.
Innovation Solution
A hybrid approach combining AI-mediated crowdsourcing, concept mining, and natural language processing to collect and summarize diverse scenarios, using a system that includes object-tracking machine-learning models and clustering algorithms to build and augment knowledge graphs, facilitating the integration of human input and machine perception for improved ontology construction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional traffic management systems use conventional data processing methods, then system simplicity is maintained, but the ability to model and understand high-level activities and root causes of events is insufficient
Solution Approach 1:
The system segments the complex task of understanding traffic events into multiple components: trajectory extraction from video data, clustering of trajectories into patterns, ontology-based event detection, and knowledge graph construction. Each component handles a specific aspect, making the overall system more manageable while achieving high-level activity modeling capability
Solution Approach 2:
The patent introduces intermediate representations including trajectory clusters and ontology structures that mediate between raw video data and high-level event understanding. These intermediaries bridge the gap between simple data processing and complex semantic understanding, enabling the system to model high-level activities without requiring the entire system to be uniformly complex
2Adaptability or versatility
If the system processes diverse cultural and linguistic descriptions of traffic events, then coverage of diverse scenarios is improved, but the difficulty of representing and integrating this information increases
Solution Approach 1:
The system employs a standardized ontology structure that provides homogeneous representation for diverse traffic events. By mapping different cultural and linguistic descriptions onto a common ontology framework with standardized classes, properties, and relationships, the system achieves uniform representation while accommodating scenario diversity
Solution Approach 2:
The knowledge graph integrates multiple types of information (trajectory data, event descriptions, contextual information) into a composite structure that combines heterogeneous data sources. This composite representation unifies diverse scenario descriptions while maintaining the ability to represent complex contextual relationships
3Measurement precision
If manual ontology construction is used, then accuracy of event modeling is improved, but the time and resources required increase significantly
Solution Approach 1:
The system performs preliminary automated processing of trajectory data and event descriptions to generate candidate ontology structures and relationships. This preliminary action prepares the data in a structured format that requires minimal manual refinement, significantly reducing the time and resources needed for final ontology construction while maintaining accuracy
Solution Approach 2:
The system incorporates feedback mechanisms where automated ontology generation results are evaluated and refined through iterative processes. Trajectory clustering outcomes and initial event detections feed back into ontology construction, allowing continuous improvement of modeling accuracy while reducing manual intervention requirements over time
Data Source
AI summary
Methods and systems of building a knowledge graph based on event-based ontology of a scene and vehicle trajectory in the scene. Image data corresponding to a plurality of scenes captured by one or more cameras is received. Event-based ontology data corresponding to events occurring in the plurality of scenes is received. Via an object-tracking machine-learning model, the system determines (i) a presence of a plurality of vehicles in the image data, and (ii) a plurality of vehicle trajectories, each vehicle trajectory associated with a respective one of the vehicles. Using a clustering model, the vehicle trajectories are clustered. A knowledge graph is augmented based on the clustered vehicle trajectories and the event-based ontology.