Face Component Model for Occlusion Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face recognition systems face challenges in accurately detecting and recognizing faces in unconstrained environments due to factors like pose, illumination, and partial occlusion, which affect their performance and efficiency.
Innovation Solution
The implementation of a face component model that detects facial parts instead of the entire face, combined with a fixed-size face descriptor and a trained convolution deep neural network for feature extraction and classification, improves accuracy and detection rates under partial occlusion and varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a face recognition system uses traditional face detection methods, then it works well on frontal facial photos, but it fails to handle faces in unconstrained environments with pose variations, illumination changes, and partial occlusion
Solution Approach 1:
The patent segments the face detection task into detecting individual face parts (eyes, nose, mouth) rather than detecting the entire face at once. This is achieved through a part-based model that independently locates facial components and then integrates them to identify the complete face, enabling robust detection under pose variations and occlusions where traditional whole-face detection fails
2Measurement precision
If a face recognition system processes all detected face images, then it achieves high recognition accuracy, but the computation complexity increases with the number of images
Solution Approach 1:
The patent extracts and utilizes geometric information from detected face parts (positions, orientations, shapes) to create a compact face representation. This extracted geometric model serves as a discriminative feature that enables accurate face recognition while maintaining constant computation complexity regardless of the number of enrolled faces, as the comparison is based on geometric parameters rather than full image processing
3Productivity
If a face detection system uses a simple detection algorithm, then it maintains low computation complexity, but it cannot accurately detect faces under partial occlusion and varying conditions
Solution Approach 1:
The system employs a hierarchical detection approach where simple detectors first locate individual face parts, then a higher-level model integrates these parts to identify complete faces. This segmentation allows the use of computationally efficient part detectors while achieving high detection rates through the collaborative integration of multiple part detections, particularly effective under occlusion where the whole face may be partially hidden
Solution Approach 2:
The patent detects and utilizes more face parts than strictly necessary for basic face detection. By detecting multiple facial components (eyes, nose, mouth, eyebrows) and their geometric relationships, the system creates a more robust face representation that improves detection accuracy under challenging conditions while the computational cost remains manageable due to the modular nature of part detection
Data Source
AI summary
In an approach to face recognition in an image, one or more computer processors receive an image that includes at least one face and one or more face parts. The one or more computer processors detect the one or more face parts in the image with a face component model. The one or more computer processors cluster the detected one or more face parts with one or more stored images. The one or more computer processors extract, from the clustered images, one or more face descriptors. The one or more computer processors determine a recognition score of the at least one face, based, at least in part, on the extracted one or more face descriptors.