Occlusion Detection for Facial Recognition Using Heat Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition processes are hindered by occlusion in captured images, where parts of the user's face are obscured, reducing the effectiveness of the authentication process.
Innovation Solution
The generation of landmark and occlusion heat maps to assess occlusion in facial images, estimating landmark locations and determining occlusion scores, which are used to control device operations and improve image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If facial recognition processes are performed on captured images, then authentication capability is provided, but occlusion of facial features reduces recognition accuracy and reliability
Solution Approach 1:
The patent segments the facial recognition process into multiple independent components: landmark detection, occlusion detection, and recognition processing. By dividing the face into multiple landmarks and assessing occlusion for each landmark separately, the system can identify which specific facial features are occluded and adjust the recognition process accordingly, thereby maintaining reliability despite partial occlusion.
Solution Approach 2:
The patent introduces an intermediary occlusion detection mechanism that acts between image capture and facial recognition. This intermediary layer assesses the occlusion status of facial landmarks and provides guidance to the recognition process, allowing the system to adaptively handle occluded images by adjusting recognition thresholds or requiring alternative authentication methods.
2Productivity
If occluded images are processed in facial recognition, then processing speed is maintained, but accuracy and false acceptance rates deteriorate
Solution Approach 1:
The patent performs preliminary occlusion detection and landmark assessment before the main recognition process. By pre-evaluating the occlusion status of facial landmarks and determining which landmarks are reliable, the system can adjust recognition parameters in advance, maintaining processing speed while improving accuracy by focusing computational resources on unoccluded or less-occluded landmarks.
3Measurement precision
If strict occlusion thresholds are applied to discard occluded images, then recognition accuracy is improved, but enrollment and authentication productivity decreases
Solution Approach 1:
The patent applies local quality assessment by evaluating occlusion levels for individual landmarks rather than applying a uniform threshold to the entire face. This allows the system to selectively use reliable landmarks for recognition while discarding only the occluded portions, thereby maintaining high accuracy without unnecessarily rejecting images that have partial but sufficient unoccluded landmarks for reliable recognition.
Data Source
AI summary
Occlusion of facial features may be detected and assessed in an image captured by a camera on a device. Landmark heat maps may be used to estimate the location of landmarks such as the eyes, mouth, and nose of a user's face in the captured image. An occlusion heat map may also be generated for the captured image. The occlusion heat map may include values representing the amount of occlusion in regions of the face. The estimated locations of the eyes, mouth, and nose may be used in combination with the occlusion heat map to assess occlusion scores for the landmarks. The occlusion scores for the landmarks may be used control one or more operations of the device.