3D Facial Model Rotation for Touchless Checkout Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional facial recognition systems face challenges in accurately identifying individuals due to reliance on single frontal images, which can be of poor quality due to lighting conditions or partial framing, and lack guidance for users to provide suitable images.
Innovation Solution
The system provides real-time feedback to ensure high-quality image capture and constructs a three-dimensional model of a user's face from multiple angles, rotating it to a frontal view for comparison with stored data, using multiple frames of video to increase recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single frontal image is used for facial recognition, then the system is simple and fast, but the recognition accuracy deteriorates due to poor image quality from lighting conditions or partial framing
Solution Approach 1:
The patent divides the facial recognition process into multiple independent stages: capturing multiple video frames, detecting faces in each frame, extracting features separately, and then aggregating results. This segmentation allows the system to process multiple images without overwhelming complexity, maintaining modularity while improving accuracy through diverse input sources
Solution Approach 2:
The patent transitions from two-dimensional static image comparison to three-dimensional spatiotemporal analysis by incorporating multiple video frames over time. This adds the temporal dimension, allowing the system to capture facial features under varying conditions (lighting, angle, expression) and aggregate information across time, thereby improving recognition accuracy without linearly increasing complexity
2Measurement precision
If multiple frames of video are used for facial recognition, then recognition accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary face detection and frame quality assessment before full facial recognition processing. By pre-screening frames to identify those with adequate quality and proper framing, the system avoids wasting computational resources on poor-quality frames, thus reducing overall processing time while maintaining high accuracy through selective use of multiple frames
Solution Approach 2:
The patent implements early termination logic where, if a sufficient number of high-quality frames are captured and processed successfully, the system can skip processing additional frames and proceed to result aggregation. This rushing through approach reduces unnecessary processing time while ensuring adequate sampling for accurate recognition
3Measurement precision
If real-time feedback is provided to users for image quality, then image quality improves, but system complexity and processing overhead increase
Solution Approach 1:
The patent implements a feedback loop where the system analyzes captured frames in real-time, assesses their quality based on predefined criteria (lighting, framing, clarity), and provides immediate guidance to users to improve their pose or lighting conditions. This feedback mechanism ensures high-quality input data is captured, significantly improving subsequent recognition accuracy while the modular architecture keeps the added complexity manageable
Solution Approach 2:
The system performs self-assessment of frame quality using automated algorithms that evaluate lighting conditions, face detectability, and framing adequacy. This self-service quality control reduces the need for complex manual intervention or additional hardware, as the system autonomously identifies and communicates quality issues to users, improving image quality without proportionally increasing system complexity
Data Source
AI summary
Facial recognition method comprises generating an average three-dimensional model from a series of two-dimensional, consecutive, captured video frames and rotating that three-dimensional model to a frontal view. The three-dimensional model may be reduced to a two-dimensional model in order to create a mean reference map. Next, the method includes accessing a facial database to compare the mean reference map to a series of previously-stored reference maps, one for each individual in the database, each of the previously-stored reference maps including a set of previously-stored reference points. Finally, an individual whose previously-stored reference map most closely matches the mean reference map is identified as a match.