Yaw Estimation via Truncated Ellipse Head Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing head pose estimation methods, particularly for yaw angle estimation, are sensitive to misalignment of face feature points, require high image resolution, are not tolerant to partial occlusions, and fail to accurately represent all human heads, making them unsuitable for critical applications like driver inattention detection.
Innovation Solution
A method that models the human head as a truncated ellipse in vertical projection, with the center of rotation fixed at the neck and the face center mapped to the nose, allowing for accurate yaw angle computation using the position of the nose and face boundaries, which are identified on an elliptic arc subtending ±60°.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If model based methods are used for yaw estimation, then computation speed is improved and real-time applications become feasible, but accuracy deteriorates due to sensitivity to misalignment of face feature points
Solution Approach 1:
The patent transforms the head model from a rigid structure with fixed feature points to a flexible parametric model defined by ellipse parameters (center position, semi-major axis, semi-minor axis, rotation angle). This parameterization allows the model to adapt to different head positions and orientations without requiring precise alignment of specific feature points, thus maintaining computational speed while improving accuracy
Solution Approach 2:
The elliptical head model serves multiple functions simultaneously: it represents the overall head shape, defines face boundaries through the ellipse perimeter, locates the face center at the ellipse center, and determines orientation through the ellipse rotation angle. This multi-functionality eliminates the need for separate feature point detection and alignment steps, resolving the contradiction between speed and accuracy
2Adaptability or versatility
If existing model based methods are used, then the system can operate with varying image resolutions, but reliability deteriorates due to sensitivity to partial occlusions
Solution Approach 1:
The patent extracts only the essential geometric parameters of the head (ellipse center, axes lengths, rotation angle) from the image data, separating the head representation from detailed facial features. This extraction makes the model robust to occlusions because the overall elliptical shape can be detected even when parts of the face are obscured, while still providing reliable yaw estimation
Solution Approach 2:
By representing the head as an ellipse (a curved, continuous shape), the model naturally handles partial occlusions better than methods relying on discrete feature points. The elliptical boundary can be partially observed and still fitted to determine head orientation, improving reliability under occlusion while maintaining adaptability to different image resolutions
Data Source
AI summary
Systems and methods of automatic detection of a facial feature are disclosed. Moreover, methods and systems of yaw estimation of a human head based on a geometrical model are also disclosed.