3D Facial Recognition Using Depth Maps and Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional facial recognition technologies that combine 3D and 2D data face high computational costs and large storage requirements, limiting their widespread adoption.
Innovation Solution
A three-dimensional facial recognition method and system that performs pose estimation, reconstructs facial depth images, and detects local grid scale-invariant feature descriptors to generate recognition results, reducing computational costs and storage needs while enhancing recognition accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If three-dimensional facial recognition technology combines 3D and 2D data to improve recognition accuracy, then recognition accuracy is improved, but computational costs and storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential 3D facial depth information from the complete 3D facial data, separating it from redundant 2D image data. By focusing on the depth map and key 3D facial landmarks rather than processing full 3D models and multiple 2D images, the system achieves accurate recognition with significantly reduced computational requirements and storage space.
Solution Approach 2:
The patent segments the facial recognition task into distinct modules: pose estimation using 3D reference models, depth image reconstruction from binocular vision, and feature extraction using scale-invariant descriptors on the reconstructed depth maps. This segmentation allows each component to process only necessary data, reducing overall computational burden while maintaining accuracy.
2Measurement precision
If three-dimensional facial recognition technology combines 3D and 2D data to improve recognition accuracy, then recognition accuracy is improved, but storage space requirements increase
Solution Approach 1:
The patent extracts only the essential 3D facial depth information from the complete 3D facial data, separating it from redundant 2D image data. By focusing on the depth map and key 3D facial landmarks rather than storing full 3D models and multiple 2D images, the system achieves accurate recognition with significantly reduced storage requirements.
Solution Approach 2:
The patent transforms facial data from high-volume 3D mesh models and multiple 2D images into a compact representation using depth maps and key point coordinates. This parameter transformation reduces the data volume while preserving the essential geometric information needed for accurate recognition.
3Use of energy by moving object
If traditional two-dimensional face recognition is used, then computational costs are low, but recognition accuracy fails in extreme application scenarios
Solution Approach 1:
The patent transitions from two-dimensional image-based recognition to three-dimensional depth-based recognition. By using binocular vision to reconstruct 3D depth maps and incorporating pose estimation in 3D space, the system gains the ability to handle extreme angles and occlusions that plague 2D methods, while managing computational costs through efficient 3D processing techniques.
Data Source
AI summary
The present disclosure provides a three-dimensional facial recognition method and system. The method includes: performing pose estimation on an input binocular vision image pair by using a three-dimensional facial reference model, to obtain a pose parameter and a virtual image pair of the three-dimensional facial reference model with respect to the binocular vision image pair; reconstructing a facial depth image of the binocular vision image pair by using the virtual image pair as prior information; detecting, according to the pose parameter, a local grid scale-invariant feature descriptor corresponding to an interest point in the facial depth image; and generating a recognition result of the binocular vision image pair according to the detected local grid scale-invariant feature descriptor and training data having attached category annotations. The present disclosure can reduce computational costs and required storage space.