Face Rotation Angle Determination Using Symmetrical Feature Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current facial recognition technologies struggle to accurately determine the face rotation angle due to the complexity of texture feature analysis, often resulting in rough estimates and potential inaccuracies.
Innovation Solution
The method involves obtaining multiple pairs of symmetrical facial feature points and calculating line segment ratios to query a correspondence between these ratios and face rotation angles, providing a precise determination of the face rotation angle.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If texture feature analysis is used to determine face rotation angle, then the determination process can be performed, but the precision of face rotation angle determination is poor and the process is complex
Solution Approach 1:
The patent extracts specific geometric features (symmetrical facial feature points and line segments) from the complex texture feature analysis, isolating the essential geometric relationships needed for angle determination. This extraction simplifies the analysis by focusing only on key geometric elements rather than processing entire texture patterns.
Solution Approach 2:
The patent changes the parameter basis from texture features to geometric parameters (coordinates of symmetrical feature points and line segment lengths). By transforming the determination basis from complex texture analysis to simple geometric measurements, the precision is improved while the complexity is reduced.
2Reliability
If texture feature analysis is used to determine face rotation angle, then the determination process can be performed, but the determination process is complex and error-prone
Solution Approach 1:
The patent segments the face into symmetrical feature point pairs, analyzing each pair independently to determine line segment lengths. This segmentation approach breaks down the complex texture analysis into simple, independent geometric measurements, reducing errors and improving reliability.
Solution Approach 2:
The patent uses symmetrical feature points as copies of each other across the midline, leveraging the natural symmetry of facial features. By copying the measurement approach from one side of the face to the other, the method ensures consistency and reduces determination errors.
3Measurement precision
If conventional texture feature method is used, then face rotation angle determination can be performed, but only rough angle can be determined and specific angle cannot be determined
Solution Approach 1:
The patent pre-establishes the geometric relationships between symmetrical feature points and the midline, creating a ready-to-use geometric model. This preliminary setup allows for direct angle calculation without time-consuming texture analysis, enabling both high precision and fast determination.
Data Source
Figure 1
Figure 2A
Figure 2B~2C
AI summary
A method and device for determining the rotation angle of a human face, and a computer storage medium. The method comprises: obtaining first position information of preset multiple human face characteristic points in a to-be-determined human face image, the number of the multiple human face characteristic points being an odd number, and the multiple human face characteristic points comprising multiple pairs of symmetrical human face characteristic points and one first human face characteristic point, and the multiple human face characteristic points being not located on a same plane (101); obtaining first position information of a symmetrical middle point of each pair of human face characteristic points according to the first position information of the human face characteristic point comprised in each pair of human face characteristic points among the multiple pairs of human face characteristic points (102), and determining the rotation angle of the to-be-determined human face image according to the first position information of the symmetrical middle point of each pair of human face characteristic points and the first position information of the first human face characteristic point (103).