Hierarchical Vehicle Event Detection for Real-Time Driving Risk
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle video systems are limited in real-time analysis and risk prediction, failing to accurately detect abnormal driving events for timely warnings or preventive measures.
Innovation Solution
A system utilizing a memory-linkable hierarchical structure for event detection, involving a first event processing step on a terminal and a second event processing step on a cloud server, with deep learning models to generate and extend event information hierarchically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vehicle videos are used for recording driving situations, then the basic recording function is achieved, but real-time analysis and risk prediction capabilities are limited
Solution Approach 1:
The patent segments event detection into two hierarchical levels: first event detection models process individual video clips to generate first event information, while second event detection models aggregate this information across multiple clips to generate second event information. This segmentation enables real-time analysis at the clip level while maintaining comprehensive risk prediction capabilities across extended time periods.
Solution Approach 2:
The patent introduces a temporal dimension by processing video data across multiple time clips. First event detection operates on individual clips, while second event detection operates across sequences of clips, effectively adding a time-based dimension to the analysis. This enables the system to distinguish between isolated incidents and recurring patterns, improving risk prediction reliability.
2Measurement precision
If deep learning models are applied to detect events in real-time, then event detection accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent divides the computational workload into two stages: first event detection models process individual video clips independently and quickly, while second event detection models aggregate results from multiple clips. This segmentation allows the system to maintain high accuracy through comprehensive analysis while reducing the time penalty by processing clips in manageable units rather than analyzing all data simultaneously.
Solution Approach 2:
The system performs preliminary event detection on individual video clips before aggregating results for comprehensive analysis. First event detection models generate initial event information that serves as input for second event detection models, enabling the system to prepare and pre-process data in stages, thereby reducing overall processing time while maintaining detection accuracy.
3Adaptability or versatility
If multiple event detection models are used for hierarchical analysis, then event detection comprehensiveness is improved, but system complexity increases
Solution Approach 1:
The patent segments the detection system into distinct first and second event detection models with clear functional boundaries. First models handle individual clip analysis while second models handle temporal aggregation, creating a modular architecture that improves comprehensiveness while managing complexity through functional separation.
Solution Approach 2:
The patent merges first event information from multiple video clips into second event information through the second event detection model. This merging process consolidates results from multiple detection instances, improving comprehensiveness by capturing patterns across time while presenting a unified output that manages system complexity.
4Measurement precision
If video data is processed and stored in cloud servers for model learning, then model accuracy is improved, but data transmission and storage requirements increase
Solution Approach 1:
The patent extracts only the essential first event information from video clips for transmission to cloud servers, rather than transmitting raw video data. This extraction process selectively transfers only the processed event detection results, significantly reducing data transmission volume while preserving the information needed for model learning and accuracy improvement.
Data Source
AI summary
Disclosed are an event detection method and device based on a memory-linkable hierarchical structure. The method includes a first event processing step including a step of collecting a real-time driving video from a running vehicle, on a terminal, and a step of applying the real-time driving video to a first event detection model to generate first event information about driving of the vehicle; and a second event processing step including a step of collecting the first event information, on the terminal, and a step of applying the first event information to a second event detection model to generate second event information about driving of the vehicle