Biometric Image Sorting via Face Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
During large-scale sports events, participants face difficulties in identifying and accessing photographs due to the large number of images taken, as numbering on athletes' vests can be obscured or distorted, making automatic sorting algorithms ineffective.
Innovation Solution
An image processor device with a computer processor unit and memory that detects faces, extracts biometric templates, and stores image references, allowing for comparison and association based on similarity scores and context information, enabling the reconstitution of identifiers even if partially obscured.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic sorting algorithms based on participant numbers are used, then image sorting efficiency is improved, but the reliability of number detection deteriorates due to masking, distortion, and interference from other numbers
Solution Approach 1:
The patent introduces biometric templates as an intermediary mechanism to bridge the gap between image data and participant identification. Instead of directly relying on obscured numbers, the system extracts biometric features from faces in images, creates templates, and uses these templates as a reliable intermediary for matching and sorting images to participants, thereby maintaining sorting efficiency while overcoming number detection reliability issues
Solution Approach 2:
The patent transforms the identification parameter from visual number recognition to biometric feature matching. By changing the fundamental parameter used for identification from text-based (numbers on vests) to biological-based (facial features), the system achieves both high sorting efficiency and reliable identification even when numbers are partially visible or distorted
2Measurement precision
If participants manually scan through all photographs to identify themselves, then identification accuracy is improved, but the time required for image retrieval deteriorates significantly with large numbers of images
Solution Approach 1:
The patent performs preliminary actions by automatically extracting biometric templates from images and pre-processing the image database during or immediately after the event. This preliminary processing creates ready-to-use biometric references that enable rapid matching and retrieval later, eliminating the need for participants to manually scan through thousands of images while maintaining accurate identification
Solution Approach 2:
The patent replaces the mechanical manual scanning process with an automated biometric recognition system. Instead of participants physically or visually searching through image stacks, the system uses computer-based biometric template matching to automatically identify and retrieve images, dramatically reducing retrieval time while preserving identification accuracy
3Measurement precision
If the complete identifier is required for image association, then sorting accuracy is improved, but the adaptability of the system deteriorates when identifiers are partially obscured or not visible
Solution Approach 1:
The patent applies partial action by using only the visible portion of identifiers or biometric features that are present in the image, rather than requiring complete identifiers. The biometric template extraction process works with partial facial visibility and can associate images even when complete identification data is not available, maintaining both sorting accuracy and adaptability to partial information
Data Source
AI summary
An image processor device includes a computer processor unit (CPU), at least one memory connected to the CPU, and device for transferring images to the CPU. The memory contains an image-processing program for processing images showing at least one person. The program performs the following operations: detecting at least a face in each image and extracting therefrom a biometric template of the face; for each image, storing in a database an image reference, the biometric template, and if possible context information for the image; comparing the biometric templates corresponding to different image references with one another and associating together the image references for which the comparison has a similarity score greater than a predetermined threshold; and searching for context information corresponding to at least one of the references of the associated images, and if there is corresponding context information, establishing a link between the associated images.