Driver Eye Tracking for Attention Detection During Handover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems that monitor driver states using in-vehicle cameras are inadequate for detecting reduced attention levels in drivers transitioning from automated to manual driving, as they are not designed to handle free-style head and eye movements, leading to inaccurate determination of manual driving ability and increased risk of accidents.
Innovation Solution
An image processing device and method that captures wide-range images of drivers at a low frame rate, selectively reads high-speed images of the eyes within a narrow capture frame to track saccades and microsaccades, allowing for accurate determination of the driver's consciousness level and recognition of the surrounding environment without mechanical attachment or wearable devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing systems use fixed-attitude camera monitoring, then the system structure is simple, but the system cannot accurately detect driver attention levels when drivers have free-style head and eye movements during automated-to-manual driving transitions
Solution Approach 1:
The patent divides the image processing into multiple stages: first capturing wide-range images at low frame rates to monitor overall driver position, then selectively capturing high-speed images of the eyes only when needed. This segmentation allows the system to achieve high measurement precision for eye movements without requiring continuous high-speed processing of the entire scene, thus managing device complexity.
Solution Approach 2:
The patent implements dynamic adjustment of capture frame rates based on driver state. The system switches between low frame rates during stable periods and high frame rates when detecting potential attention issues or during automated-to-manual transitions. This dynamic approach enables accurate detection of rapid eye movements while avoiding unnecessary continuous high-speed processing that would increase system complexity.
2Measurement precision
If the system captures images at high frame rates continuously, then the measurement precision of eye movements is improved, but the energy consumption and data processing load increase significantly
Solution Approach 1:
The patent employs periodic high-speed image capture only during critical periods such as automated-to-manual driving transitions or when attention deterioration is detected. During normal operation, the system uses low frame rate monitoring. This periodic high-speed sampling maintains measurement precision for eye movements while dramatically reducing overall energy consumption compared to continuous high-frame-rate capture.
Solution Approach 2:
The system performs preliminary detection using low frame rate wide-range imaging to identify when high-precision eye tracking is needed. By detecting driver position and basic eye region movements in advance, the system can trigger high-speed capture only when necessary, preventing unnecessary energy consumption while maintaining the ability to accurately measure eye movements when critical events occur.
3Adaptability or versatility
If the system uses wide-range camera coverage, then the adaptability to driver movements is improved, but the precision of eye movement detection decreases
Solution Approach 1:
The patent segments the monitoring task into two parts: wide-range low-resolution monitoring for adaptability, and narrow high-resolution eye tracking for precision. The system first captures wide-range images to adapt to driver position changes, then selectively captures high-speed narrow images of the eyes. This segmentation allows the system to maintain both wide adaptability and precise eye movement detection without requiring a single camera to do both simultaneously at full capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are an image processing device and an image processing method for acquiring a consciousness level at which a driver recognizes a surrounding environment on the basis of an image of the driver picked up by an in-vehicle camera. An image of an eyelips is picked up by an in-vehicle camera 100 at low resolution and eye positions grasped in the image analysis are corrected for each frame while being subjected to local high-speed processing in parallel, whereby the tracking of the saccade among eyeball movement is realized. That is, active local high-speed eye tracking is performed only on a capture frame 102 in an image picked up by the in-vehicle camera 100, whereby it becomes possible to monitor the saccade of the eye movement without causing an attachment load.