Skeleton Key Point Extraction for Non-Contact Identity Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing identity authentication methods require user participation and strict environmental conditions, such as close-range image acquisition and specific angles, making them inefficient and dependent on high-definition imaging equipment.
Innovation Solution
A method using a convolutional neural network to extract skeleton key points from human body images, convert them into feature data, and process this data with a physique feature model for non-contact, non-perceptual identity authentication, allowing for recognition at intermediate or long distances with reduced user involvement and environmental dependency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If close-range image acquisition is used for face authentication, then authentication accuracy is improved, but user participation and environmental requirements increase
Solution Approach 1:
The patent extracts the essential authentication features from the full human body image by identifying and processing only the skeleton key points. This extraction approach allows authentication to be performed on simplified structural data rather than requiring high-definition close-range images, thereby reducing user participation requirements while maintaining authentication accuracy.
Solution Approach 2:
The patent transitions from two-dimensional image-based authentication to three-dimensional spatial coordinate-based authentication using skeleton key points. By converting image coordinates to three-dimensional spatial coordinates and using distance ratios for verification, the system achieves accurate authentication without requiring close-range imaging, thus resolving the contradiction between accuracy and ease of operation.
2Measurement precision
If high-definition image acquisition is used, then authentication precision is improved, but device complexity and environmental requirements increase
Solution Approach 1:
The patent extracts skeleton key points from standard-definition human body images and converts them into authentication features. This extraction method eliminates the need for high-definition imaging equipment while maintaining authentication precision, as the skeleton key points provide sufficient structural information for accurate verification.
Solution Approach 2:
The patent creates a simplified computational model (skeleton key point structure) that copies only the essential authentication information from the full human body image. This abstracted representation enables accurate authentication using standard-definition images, thereby reducing device complexity and environmental requirements.
3Reliability
If strict authentication conditions are imposed, then reliability is improved, but productivity and user convenience decrease
Solution Approach 1:
The patent enables automatic authentication by having the system independently identify skeleton key points, calculate three-dimensional spatial coordinates, and verify distance ratios without requiring user cooperation for positioning or posture adjustment. This self-service approach maintains reliable verification while significantly improving authentication efficiency and user convenience.
Solution Approach 2:
The patent performs partial authentication by verifying only the distance ratios between key skeleton points rather than analyzing the entire human body image. This partial action approach reduces processing complexity and time while maintaining sufficient reliability for accurate authentication.
Data Source
AI summary
A method, a system and a terminal for identity authentication, and a computer readable storage medium are provided. The method includes: acquiring a human body image of a person to be authenticated, and determining from the human body image a plurality of skeleton key points of the person to be authenticated; converting the skeleton key points into feature data, and combining the feature data to form physique feature information characterizing the person to be authenticated; processing the physique feature information using a physique feature model by inputting the physique feature information into the physique feature model, to obtain a processing result; and recognizing the identity of the person to be authenticated based on the processing result of the physique feature model.