Multi-Camera Person Retrieval With Biometric Reliability Prioritization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing person retrieval systems based on biometric information often fail to complete processing within desired time frames due to insufficient processing capacity or frequent requests, leading to inefficiencies.

Innovation Solution

A person retrieval apparatus that extracts biometric information, determines its quality, sets a reliability level based on quality assessment, and performs retrieval processing in descending order of reliability, optimizing processing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If person retrieval processing is performed in the order of request generation, then processing simplicity is maintained, but processing time exceeds desired time frame

Engineering Contradiction:
Improveprocessing speedVSAvoidprocessing time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary quality assessment and reliability level assignment to biometric information before actual retrieval processing. By pre-evaluating the quality of extracted biometric data and assigning reliability levels in advance, the system enables prioritized processing that reduces overall processing time while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system changes the processing parameter from time-order sequence to reliability-level sequence. Instead of processing requests in the order they are generated, the system reorders processing based on the reliability levels assigned to biometric information, allowing critical high-reliability cases to be processed first and thus reducing processing time for important tasks.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If biometric information extraction is performed for all images, then retrieval accuracy is improved, but processing capacity becomes insufficient

Engineering Contradiction:
Improveretrieval accuracyVSAvoidprocessing capacity
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system applies local quality assessment to evaluate the quality of biometric information extracted from each image individually. Based on this localized quality evaluation, the system assigns reliability levels that reflect the specific quality characteristics of each biometric data set, enabling selective processing that maintains accuracy for high-quality data while reducing processing load for low-quality data.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system performs partial processing by selectively processing biometric information based on reliability levels. Instead of uniformly processing all extracted biometric data, the system focuses processing resources on high-reliability biometric information, achieving sufficient accuracy for critical cases while reducing overall processing load to match available capacity.

Inventive Principle:
Principle #16Partial or excessive action

3Ease of operation

If frequent person retrieval requests are handled, then system responsiveness is improved, but processing time for each request increases

Engineering Contradiction:
Improvesystem responsivenessVSAvoidprocessing time per request
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The system performs preliminary quality assessment and reliability assignment when biometric information is first extracted, before retrieval requests are processed. This pre-processing establishes a priority framework that enables rapid, informed decision-making when frequent retrieval requests arrive, allowing the system to respond quickly by processing high-reliability requests first without increasing overall processing time.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2128819B1Person retrieval device
Publication Date: 2025.08.06 KK TOSHIBA
  • EP2128819B1 patent drawingFigure 1
  • EP2128819B1 patent drawingFigure 2
  • EP2128819B1 patent drawingFigure 3

AI summary

In a person retrieval apparatus (1), a plurality of extraction processing sections (113a) each extract personal biometric information from images taken by a plurality of cameras (111). A quality determination section (113b) determines a quality of each piece of biometric information extracted by the extraction processing sections (113a). A reliability level setting section (124) sets a reliability level to each piece of biometric information on the basis of the quality determined by the quality determination section (113b). The biometric information extracted by the extraction processing sections (113a) and the reliability level set by the reliability level setting section (124) are stored in a memory (126a). In this state, in the person retrieval apparatus (1), the face retrieval section (126) performs person retrieval processing on each piece of biometric information stored in the memory (126a) in descending order of the reliability level corresponding to each piece of biometric information.