Palm ROI Expansion for Reliable Liveness Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional biometric liveness detection systems face challenges in distinguishing between real and fake palm images due to variations in imaging conditions and the use of fake biometrics, leading to unreliable outcomes and loss of information through padding with zero intensity pixels.
Innovation Solution
The system generates variations of palm images under different conditions and uses individual liveness detectors for each type of fake palm, combining decisions to determine liveness, and adjusts ROIs by expanding them in specific directions using image content rather than padding with zeros to maintain semantic information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If zero-padding is used to expand ROI to meet specified dimensions, then the ROI dimension requirement is satisfied, but information loss and distortion occur
Solution Approach 1:
The solution segments the image into multiple overlapping patches instead of using a single padded ROI. Each patch is processed independently and then aggregated, avoiding the need for zero-padding while maintaining dimensional requirements through the aggregation of multiple smaller regions.
Solution Approach 2:
The solution transitions from expanding a single ROI in spatial dimensions to expanding the detection coverage through multiple patches in the sequence dimension. By processing multiple patches and aggregating their results, the system achieves comprehensive coverage without introducing artificial padding.
2Device complexity
If a single generic fake palm detection model is trained, then the device complexity is reduced, but the ability to detect various types of fake biometrics deteriorates
Solution Approach 1:
The detection system is segmented into multiple specialized detectors, each trained to detect a specific type of fake biometric (photo-based, video-based, 3D mold-based, etc.). This segmentation allows each detector to specialize in particular fraud types while the ensemble maintains overall system manageability through modular architecture.
Solution Approach 2:
The system changes the parameter of detector specialization from a single generic model to multiple models with different expertise parameters. Each detector is configured with specific training parameters tailored to its target fake biometric type, improving overall detection reliability across diverse fraud scenarios.
3Adaptability or versatility
If palm images are captured under varying imaging conditions, then the adaptability of the system is improved, but the measurement precision of liveness detection deteriorates
Solution Approach 1:
The system performs preliminary actions by capturing multiple patches under varying conditions before the actual liveness detection. These preliminary captures include different lighting conditions, angles, and distances, which are then aggregated to compensate for the variations and improve the precision of the final detection result.
Solution Approach 2:
The system merges multiple image patches captured under different imaging conditions by aggregating their feature representations. This combining process integrates information from varied conditions, allowing the system to maintain high measurement precision despite the adaptability required to handle diverse capture scenarios.
Data Source
AI summary
The present teaching relates to detecting palm liveness. When an image is received with visual information claimed to represent a palm of a person, an initial region of interests (ROI) is identified from the image that corresponds to the palm and an initial dimension thereof is determined. When the initial dimension is smaller than a specified dimension, the initial ROI is extended in some respective directions to some expansion region with certain expansion dimension to generate an ROI using the visual information in the ROI from the image. A plurality of decisions are obtained with respect to the ROI, each of which is made individually on whether the ROI represents a specific type of fake palm. The decisions are then combined to derive a liveness detection decision on whether the palm captured in the image is live.