Facial Recognition with 3D Mesh and Multi-Biometrics Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current facial recognition systems face challenges in accurately identifying faces under varying conditions such as different ages, lighting, poses, and obstructions, leading to high false acceptance and rejection rates, especially when dealing with large databases and diverse populations.
Innovation Solution
The implementation of a facial recognition system that uses an automatically adjustable camera rig, a machine learning model combining 2D and 3D analyses, and infrared sensors to assess body temperature, allowing for improved identification and tracking even with obstructed facial features, and employing multi-biometrics fusion for enhanced accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional 2D facial recognition methods are used, then the system is simpler and faster, but accuracy deteriorates under varying conditions such as different ages, lighting, poses, and obstructions
Solution Approach 1:
The patent transitions from traditional 2D facial recognition to 3D facial analysis by capturing multiple images from different angles using a camera rig. This dimensional change enables the system to understand facial geometry in three-dimensional space, improving accuracy under varying conditions such as poses, lighting, and obstructions while maintaining manageable system complexity through structured data processing
2Measurement precision
If multiple images from different angles are captured using a camera rig, then facial recognition accuracy improves under varying conditions, but the time required for image acquisition and processing increases
Solution Approach 1:
The system performs preliminary actions by capturing multiple images from different angles before the actual recognition task. These pre-acquired images are stored and processed in advance to create a comprehensive facial model, which then enables rapid and accurate recognition even when the subject is partially obscured or in non-standard poses during actual use
3Measurement precision
If 3D mesh analysis is applied to facial features, then identification accuracy improves for obstructed faces, but computational requirements and system complexity increase
Solution Approach 1:
The patent segments the facial recognition task into distinct components: capturing multiple images, generating a 3D mesh model, identifying key facial points, and performing recognition based on spatial relationships. This segmentation allows each component to be optimized independently, reducing overall computational complexity while maintaining high accuracy for obstructed faces through focused analysis of critical facial regions
4Reliability
If multi-biometrics fusion is implemented, then false acceptance and rejection rates are reduced, but system complexity and processing requirements increase
Solution Approach 1:
The patent merges multiple biometric data sources including 2D facial images, 3D mesh models, key point coordinates, and spatial relationship data into a unified recognition system. This combining approach leverages the complementary strengths of each biometric modality to reduce false acceptance and rejection rates while managing system complexity through integrated processing architecture
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This solution significantly reduces false acceptance and rejection rates, achieving a 97.9% true acceptance rate with a false acceptance rate of 0.001, and enables effective identification and tracking across multiple sensing locations, including in scenarios with off-angles and obstructions like masks, by utilizing a 3D mesh and multi-analysis fusion.
Implementation Method 1
a body temperature assessment that uses the infrared sensor to assess the body temperature of the person corresponded to the temporary identification at the target zones
Data Source
AI summary
A facial recognition system, comprising: an automatically adjustable camera rig comprising a plurality of movable cameras, wherein the plurality of movable cameras are moved by a camera control platform according to take enrollment images; a first input for receiving images from the automatically adjustable camera rig; a second input for receiving a plurality of images from an comparative input; a first computing memory for storing a machine learning model that includes a three dimensional and a two dimensional comparison between the received first input and the received second input, wherein the comparison uses key facial points to compute a distance between the first input and the second input; and a match output in a case of a distance within a predetermined threshold.