Driver Drowsiness Image Extraction Using Face Direction Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems erroneously determine that a driver's eyes are closed based on the direction of their face when detecting dozing, leading to incorrect extraction of dozing scenes.
Innovation Solution
An information processing device and system that acquires and analyzes images to extract captured images where the eyes are closed and the face is facing forward, utilizing vehicle speed thresholds, face direction recognition, and gaze point validation to suppress erroneous determinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the degree of eye opening is used to extract dozing scenes, then the extraction process is simple, but erroneous determination occurs when the face direction is not forward
Solution Approach 1:
The extraction process is segmented into multiple independent judgment stages: first judging face direction, then judging eye opening state, and finally determining dozing scenes. This segmentation allows each judgment to be performed under appropriate conditions, improving reliability while maintaining reasonable complexity.
Solution Approach 2:
Face direction judgment is performed as a preliminary action before eye opening state judgment. By pre-establishing whether the face is in the forward direction, the system avoids erroneous dozing detection when the driver is looking in other directions, thereby improving detection accuracy.
2Device complexity
If only eye opening degree is considered, then the detection method is simple, but false positives occur during parking or rest periods
Solution Approach 1:
The detection criteria are made dynamic by incorporating vehicle speed as a variable condition. When vehicle speed is above the threshold, eye closing is considered a dozing state; when speed is below the threshold, the same eye closing state is not classified as dozing. This dynamic adjustment reduces false positives during parking while maintaining detection sensitivity during driving.
3Reliability
If multiple judgment criteria (face direction, eye opening, vehicle speed) are applied, then detection accuracy improves, but processing complexity increases
Solution Approach 1:
Multiple judgment criteria are segmented into distinct processing modules: face direction judgment module, eye opening state judgment module, and vehicle speed judgment module. Each module independently evaluates its specific criterion and provides output to the final determination, making the complex processing manageable and maintainable.
Solution Approach 2:
Face direction and vehicle speed are judged as preliminary conditions before final dozing determination. These preliminary judgments filter out cases that should not be considered dozing scenes, reducing the burden on subsequent processing stages and improving overall efficiency.
Data Source
AI summary
An acquisition unit acquires vehicle information and a moving image frame as image information captured by an imaging unit from a database. A recognition unit recognizes opening and closing of an eye of a driver, a direction of a face, and a gaze point. An extraction unit extracts, from among the captured images acquired by the acquisition unit, a captured image in which the eye is closed and the face is facing forward in a predetermined range.