Driver Looking-Away Detection Using Face-Missing Image Ratios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing looking away detection systems fail to determine if a driver is looking away during driving when a face is not visible in the captured image, leading to incomplete detection.
Innovation Solution
A looking away determination device that assesses the proportion of images without a detected face based on driving state information, determining a driver is looking away when this proportion exceeds a predetermined value, and adjusts facial and line-of-sight direction condition ranges according to driving state changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If looking away detection is performed only when a face appears in the captured image, then detection accuracy is improved when faces are visible, but detection capability is lost when faces do not appear
Solution Approach 1:
The system dynamically switches between two detection modes: face-based line-of-sight detection when faces are visible, and driving state-based detection when faces are not visible. This dynamic adaptation allows the system to maintain detection capability across varying conditions while preserving accuracy in each specific scenario
Solution Approach 2:
The determination device integrates multiple detection functions into a single system: it can perform both face-based line-of-sight detection and driving state-based detection. This multi-functionality ensures that the system can operate effectively regardless of whether faces are visible in the captured images
2Device complexity
If a fixed threshold is used for determining looking away, then the determination rule is simple, but it cannot adapt to different driving states and speeds
Solution Approach 1:
The system changes the determination parameters (predetermined values for time proportions) based on driving state information such as vehicle speed and acceleration. When the vehicle is stationary or moving slowly, more lenient thresholds are applied, while stricter thresholds are used during normal driving, allowing adaptation to different driving conditions without excessive complexity
3Reliability
If looking away determination is performed continuously regardless of driving state, then detection coverage is maximized, but false alerts increase due to normal head movements during turning or lane changes
Solution Approach 1:
The system uses driving state information (vehicle speed, acceleration, turning status) as feedback to dynamically adjust the looking away determination criteria. This feedback mechanism allows the system to distinguish between intentional head movements for driving maneuvers and actual looking away behavior, reducing false alerts while maintaining comprehensive detection coverage
Data Source
AI summary
A looking away determination device includes: a determination unit that determines that a driver is in a looking away state when a proportion of an image in which a face of the driver is not detected with respect to a plurality of images obtained by imaging the driver is equal to or greater than a first predetermined value, the first predetermined value being determined based on driving state information that indicates a state during driving by the driver.