Face Recognition Neural Network Rotation Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object identification systems using artificial neural networks face challenges in accurately recognizing rotated faces, as they require frontal or specific angle views, leading to reduced reliability in automated machine recognition.
Innovation Solution
The object identification unit divides the image into triangles defined by characteristic points, ascertains the rotational angle based on symmetry properties, applies a correction factor to the triangle areas, and scales them to a rotation-invariant dimension, enabling the artificial neural network to compensate for rotation and improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the artificial neural network uses standard image input without rotation compensation, then the system is simpler to implement, but the identification reliability decreases when faces are rotated
Solution Approach 1:
The face image is divided into multiple triangular regions with vertices at characteristic points (eyes, nose, mouth, etc.). Each triangle's area is calculated and used as a feature descriptor. This segmentation transforms the rotation problem into a set of local geometric measurements that can be individually corrected.
Solution Approach 2:
The patent changes the parameter representation from raw pixel values to triangle area ratios. By using ratios of triangle areas (which are invariant to uniform scaling) and applying correction factors based on detected rotation angles, the system compensates for rotation while maintaining simplicity in the neural network input.
2Adaptability or versatility
If the system requires frontal views for identification, then the processing is simpler, but the adaptability to different viewing angles decreases
Solution Approach 1:
The system performs preliminary detection of characteristic points and calculation of triangle areas before the main identification process. Rotation angles are estimated from the triangular configuration in advance, and correction factors are pre-computed. This preliminary processing prepares the data in a rotation-invariant form, enabling the neural network to handle various viewing angles without requiring complex real-time adjustments.
3Measurement precision
If more characteristic points and triangles are used for rotation compensation, then the rotational angle detection becomes more accurate, but the computational complexity increases
Solution Approach 1:
The patent uses a selective set of characteristic points (typically 5-10 key points such as eyes, nose tip, mouth corners, and chin) rather than attempting to analyze the entire face or use excessive points. This partial approach provides sufficient information for accurate rotation detection while keeping computational complexity manageable. The triangle area ratios from these key points give robust rotation estimates without over-processing.
Data Source
AI summary
An object identification unit contains an artificial neural network and is designed to identify human faces. For this purpose, a face is divided into a number of triangles. The relative component of the area of each triangle in the total of the areas of all triangles is ascertained to ascertain a rotational angle of the face. The relative component of the area of each triangle in the total of the area of all triangles is then scaled to a rotation-invariant dimension of the face. The scaled area of the triangles is supplied to the artificial neural network in order to identify a person.