Face Orientation Detection Using Vertical Sum Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face orientation detection systems require large memory capacity and long processing times due to the need to store multiple orientations of image data and perform time-consuming correlation comparisons, and are inflexible when faced with varying conditions or partial face images.
Innovation Solution
The system detects face orientation by calculating sum values in the vertical direction of image data, using characteristic tables to reduce memory usage and processing time, and accurately determines orientation using eye and nose position detection methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple orientations of image data are stored in advance for comparison, then face orientation detection accuracy is improved, but memory capacity requirement increases significantly
Solution Approach 1:
The patent extracts only the essential characteristic data (sum values in vertical direction) from complete image data for storage in the characteristic table. By storing only these extracted features rather than full image data, the memory capacity is dramatically reduced while still enabling accurate face orientation detection through comparison of these characteristic values.
Solution Approach 2:
The patent segments the face detection process into distinct steps: first calculating sum values in the vertical direction to identify the face region, then detecting eye positions, and finally determining face orientation. This segmentation allows the system to process and store only the necessary characteristic data at each stage rather than requiring complete image data storage.
2Measurement precision
If complete image data is stored and compared for each orientation, then detection accuracy is maintained, but processing time increases
Solution Approach 1:
The patent performs preliminary calculation of sum values in the vertical direction for each row of pixels and stores these pre-calculated characteristic values in a characteristic table. When detecting face orientation, the system only needs to retrieve and compare these pre-computed sum values rather than processing complete image data, significantly reducing processing time while maintaining detection accuracy.
3Measurement precision
If four specific points (tear ducts and lips) are extracted for orientation detection, then detection precision is improved, but the system becomes inflexible when lips are not visible
Solution Approach 1:
The patent uses sum values in the vertical direction as a universal characteristic that can be calculated for any face image regardless of which specific facial features are visible. This approach allows the system to detect face orientation using any combination of visible features (eyes, nose, mouth, or partial features) rather than requiring specific landmarks like tear ducts and lips, thereby achieving both precision and flexibility.
4Measurement precision
If image data is stored in original resolution, then detection accuracy is maintained, but data processing amount increases
Solution Approach 1:
The patent extracts only the sum values in the vertical direction from complete image data for storage and comparison. By calculating and storing only these one-dimensional characteristic values rather than two-dimensional image data, the data processing amount is dramatically reduced while the face region detection accuracy is maintained through comparison of these essential characteristic patterns.
Data Source
AI summary
The CPU of a face orientation detection apparatus which acquires an image frame photographed by a camera through a cable detects the face region in the horizontal direction of the acquired image frame by executing computer programs stored on a hard disk. Moreover, the CPU detects the eye position in the vertical direction from the image frame, and detects the nose position in the horizontal direction based on the detected eye position. Furthermore, the CPU detects the orientation of the face included in the image frame, based on the detected nose position and face region.