Rotating DToF Depth Capture for Precise Payment Face Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face recognition in face recognition payment systems is not precise enough when performed directly on cropped images, necessitating improved precision and reduced power consumption.
Innovation Solution
A method involving a Dtof sensor on a payment terminal that rotates to capture lattice depth images from different angles, followed by fusion processing to generate a dense lattice depth image, enhancing face recognition accuracy and reducing power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If face recognition is performed directly on the cropped to-be-recognized image, then the processing speed is fast, but the recognition precision is not sufficient
Solution Approach 1:
The patent applies preliminary action by performing face detection and image cropping before recognition. The system detects the face area in the original image, crops the to-be-recognized image from this detected area, and then performs recognition on the cropped image. This preliminary processing prepares the data in an optimal state for subsequent recognition operations.
Solution Approach 2:
The patent replaces traditional 2D image-based recognition with a 3D depth map-based recognition system. By using a depth sensor to capture three-dimensional facial information and constructing a depth map, the system substitutes mechanical/optical 2D imaging with 3D spatial measurement, significantly improving recognition precision while maintaining operational efficiency.
2Measurement precision
If multiple depth images are captured from different angles and fused, then the recognition accuracy is improved, but the power consumption increases
Solution Approach 1:
The patent applies partial action by capturing depth images from multiple angles (excessive action) but only processing and fusing the most relevant portions for recognition. The system captures more depth images than a single view would provide, then selectively processes these partial views to construct a comprehensive 3D depth map, achieving high accuracy without processing all possible data equally.
Solution Approach 2:
The patent transitions from 2D image processing to 3D depth map construction by capturing depth information from multiple angles. This dimensional change allows the system to build a three-dimensional representation of the face, improving recognition accuracy while the selective processing of angular data helps manage power consumption.
3Measurement precision
If a single depth image is used for recognition, then the power consumption is low, but the recognition precision is insufficient
Solution Approach 1:
The patent merges multiple depth images captured from different angles into a single comprehensive 3D depth map. By combining the depth information from multiple views, the system creates a unified three-dimensional representation that contains richer facial features, thereby improving recognition precision while consolidating multiple data sources into one integrated structure.
Solution Approach 2:
The patent uses another dimension (3D depth information) to compensate for the quantity constraint. Instead of relying on multiple 2D images, the system captures depth data from multiple angles and fuses them into a 3D depth map, effectively using spatial dimensionality to enhance recognition precision without proportionally increasing the quantity of processed images.
Data Source
AI summary
A face image processing method includes: obtaining a plurality of lattice depth images acquired by performing a depth image acquisition on a target face from different acquisition angles; performing a fusion processing on the plurality of lattice depth images to obtain a dense lattice depth image; and performing a face recognition processing on the dense lattice depth image to obtain a face recognition result.