Vehicle Object Recognition with Gaze-Based Relevance Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle safety systems struggle to accurately assess the relevance of detected objects in the vehicle's environment, particularly in relation to driving tasks and navigation, due to insufficient contextual information.
Innovation Solution
A method that utilizes gaze direction data from vehicle occupants to classify initially irrelevant objects as potentially relevant based on prolonged gaze duration and tracking, followed by confirming relevance through trajectory analysis and storing relevant trajectories in a digital map for future vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If objects are classified as irrelevant based on initial detection, then processing efficiency is improved, but safety-relevant objects may be missed
Solution Approach 1:
The system performs preliminary classification of objects as relevant or irrelevant based on initial detection. This preliminary action filters out obviously irrelevant objects early, allowing the system to focus computational resources on potentially safety-critical objects while maintaining high processing efficiency.
Solution Approach 2:
The system implements feedback mechanisms where classification results are continuously refined. Objects initially classified as irrelevant can be re-evaluated when new information becomes available (e.g., gaze data, trajectory changes), ensuring that safety-relevant objects are not permanently discarded but rather temporarily filtered with the possibility of reclassification.
2Loss of information
If gaze recording is used to identify relevant objects, then contextual information is improved, but system complexity increases
Solution Approach 1:
The patent introduces gaze recording devices as intermediary components that capture occupant attention patterns. These devices serve as mediators between the driver's cognitive state and the object recognition system, providing indirect but valuable information about which objects warrant closer inspection without requiring direct neural or cognitive interfacing.
Solution Approach 2:
The system replaces complex mechanical or manual monitoring of driver attention with optical sensing technologies. Gaze tracking using cameras and image processing algorithms substitutes for physical observation or manual driver reports, reducing mechanical complexity while capturing rich contextual information about driver focus and interest.
3Adaptability or versatility
If trajectories are stored in digital map, then navigation capability is improved, but data storage requirements increase
Solution Approach 1:
The digital map stores trajectory information with local quality - not all locations require the same level of detail. Safety-relevant trajectories (those associated with frequent hazards or important navigation points) are stored with high precision, while less critical areas use coarser representations. This selective detail approach optimizes navigation capability while controlling data storage requirements.
Solution Approach 2:
The system performs preliminary filtering and aggregation of trajectory data before storage. Rather than storing every raw trajectory point, the system pre-processes data to identify and retain only the most significant trajectory patterns and safety-relevant paths, reducing storage requirements while maintaining navigation effectiveness.
Data Source
AI summary
A method involves classifying a respective object, which has been recognized by an object recognition device of the vehicle and initially classified as not relevant to the vehicle, as potentially relevant if it is recognized by a gaze recording device in the vehicle that a gaze of at least one vehicle occupant of the vehicle lingers on the respective object for longer than a predetermined minimum duration. Using a recorded gaze direction of the at least one vehicle occupant, it is determined that a respective moving object classified as potentially relevant is tracked by the at least one vehicle occupant for at least a predetermined duration. A trajectory of the respective recorded and moving object classified as potentially relevant to the vehicle is identified as being relevant to the safety of the vehicle after the predetermined duration has elapsed if it is necessary for the vehicle to react to the respective object.