Vehicle ROI Detection Using Occupant Gaze for Sensor Load Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional advanced driver assistance systems face performance issues due to data overload from environment sensors, as they attempt to process information from multiple sensors equally, exceeding the processing power and restricting sensing capabilities.
Innovation Solution
A system that uses a machine-learned predictor, trained on data including the line of sight of a human driver, to identify a region of interest within environment sensor data, allowing the control unit to operate the vehicle based on this determination, thereby focusing processing resources on relevant areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the environment sensor processes all information captured by multiple sensors equally, then comprehensive sensing coverage is achieved, but data overload occurs exceeding processing capabilities
Solution Approach 1:
The patent segments the sensor data processing by dividing the field of view into multiple regions of interest (ROIs). Instead of processing all sensor data uniformly, the system identifies and prioritizes specific ROIs based on relevance criteria, thereby reducing the overall data volume requiring intensive processing while maintaining comprehensive sensing coverage through selective focus on critical areas.
Solution Approach 2:
The patent applies local quality by differentiating processing intensity across different spatial regions. High-priority ROIs receive enhanced processing resources and attention, while low-priority regions undergo reduced processing. This non-uniform allocation of processing power optimizes system performance by concentrating computational resources where they are most needed, resolving the contradiction between comprehensive sensing and processing capacity.
2Measurement precision
If the environment sensor captures high-resolution data from all directions, then sensing precision is improved, but data volume increases causing performance issues
Solution Approach 1:
The patent segments the high-resolution sensor data into multiple regions of interest, processing only the necessary portions at full resolution. By identifying ROIs that require detailed analysis and processing other regions at lower resolution or aggregating their data, the system maintains measurement precision for critical areas while significantly reducing the total data volume that contributes to performance issues.
Solution Approach 2:
The patent applies local quality by varying the resolution and processing depth across different spatial regions. High-priority ROIs maintain high measurement precision with full-resolution processing, while low-priority regions use reduced resolution processing. This spatially differentiated approach preserves sensing precision where needed while reducing overall data volume to prevent system overload.
Data Source
AI summary
A method for a vehicle includes determining a region of interest based on environment sensor data corresponding an environment of the vehicle and a machine-learned predictor configured to identify, within the environment sensor data, a region as the region of interest that at least statistically coincides with a line of sight of an occupant of the vehicle. The method also includes classifying a detected object within the determined region of interest using an object detection algorithm. The method also includes operating the vehicle in based on at least one of the detected object and the determined region of interest.