Face Collator Partial Region Exclusion for Occlusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face image collation systems face challenges in displaying search results with high reliability due to environmental factors and partial occlusions, which affect the accuracy of feature quantity comparisons.
Innovation Solution
A collator and method that perform partial collation by excluding specific areas from both registered and search face images, calculating feature quantities for these areas, and displaying results based on similarity calculations, thereby excluding abnormal values and enhancing reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If environmental parameters and lighting conditions vary during face image capture, then the system must process images under different conditions, but this reduces the reliability of collation results
Solution Approach 1:
The patent divides the face image into multiple regions (eye region, nose region, mouth region, etc.) and performs collation separately for each region. This segmentation allows the system to exclude regions affected by occlusions or abnormal lighting conditions, thereby maintaining high reliability while adapting to varying environmental conditions. The collation unit calculates similarity for each region independently and combines these results to produce the final collation outcome.
2Loss of information
If the entire face image is used for collation, then more information is available for comparison, but occlusions or abnormalities in certain areas reduce the accuracy of the result
Solution Approach 1:
The patent extracts and excludes specific regions from the face image that are affected by occlusions or abnormalities. The collation unit identifies regions with abnormal values (such as occluded areas or regions with poor lighting) and excludes these from the collation process. This extraction of problematic regions ensures that the collation is performed only on reliable, unoccluded portions of the face image, thereby maintaining high measurement precision while still utilizing available information from other regions.
3Productivity
If feature quantities from all regions are compared, then comprehensive comparison is achieved, but abnormal values in certain regions degrade the overall collation accuracy
Solution Approach 1:
The patent applies different quality standards to different regions of the face image. Each region is evaluated individually to determine whether its feature quantities are abnormal. Regions with normal feature quantities are included in the collation with full weight, while regions with abnormal values are excluded or given reduced weight. This local quality assessment ensures that the comprehensive comparison is performed only on reliable regions, maintaining both productivity and reliability in the collation process.
Data Source
AI summary
A collator includes at least one processor and a storage unit storing a plurality of registered face images, the processor performs a partial collation for collating a feature quantity of a first target area excluding partial areas in each of the plurality of registered face images with a feature quantity of a second target area excluding a partial area in a search face image to be searched, and displays a result of the partial collation.