Person Recognition Apparatus Using Time-Based Image Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing person recognition methods fail to accurately identify individuals over time due to significant changes in facial features, especially in children, as the method disclosed in JP2012-27868A updates identification elements infrequently, leading to inaccurate face recognition when the time gap between shooting dates is large.

Innovation Solution

A person recognition apparatus and method that extracts and classifies feature amounts from images based on shooting dates and times, determines similarity within groups, and integrates feature amounts between adjacent groups with a predetermined threshold, allowing for accurate recognition of individuals with time-dependent face changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If identification elements are updated infrequently to reduce processing complexity, then device complexity is reduced, but measurement precision of face recognition deteriorates over time

Engineering Contradiction:
Improveprocessing complexityVSAvoidface recognition accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent divides the image data into multiple time periods and creates separate identification elements for each period. Instead of maintaining a single identification element that becomes outdated over time, the system segments the recognition process into multiple time-based groups, each with its own accurate identification element. This resolves the contradiction by maintaining low processing complexity through structured segmentation while preserving high measurement precision within each time segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a dynamic identification element management system where identification elements are created and updated based on time-dependent changes. The system dynamically adjusts which identification element to use based on the shooting date of images, ensuring that the most current and accurate identification element is applied. This dynamic approach maintains measurement precision without requiring continuous full-system reprocessing.

Inventive Principle:
Principle #15Dynamics

2Productivity

If the time range for updating identification elements is extended to cover longer periods, then productivity is improved by reducing update frequency, but measurement precision deteriorates due to facial feature changes

Engineering Contradiction:
Improveupdate efficiencyVSAvoidperson recognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the time range into multiple periods, each with its own identification element. Instead of creating one identification element for a long time period (which would maintain productivity but reduce precision), the system creates multiple segmented identification elements for shorter time periods. This allows the system to maintain high productivity through efficient segmentation while ensuring high measurement precision within each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter of time range duration based on the specific time period being processed. By adjusting the time range parameter to match appropriate segmentation intervals, the system optimizes both productivity and measurement precision. The identification elements are created with time-range parameters that balance update efficiency with accuracy requirements for different time periods.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If identification elements are updated frequently to maintain accuracy, then measurement precision is improved, but device complexity increases due to more processing operations

Engineering Contradiction:
Improveface recognition accuracyVSAvoidprocessing operations
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent reduces device complexity by segmenting the processing into organized time-based groups rather than performing continuous or frequent full-system processing. Each time period is processed independently with its own identification element, which simplifies the overall processing architecture while maintaining accuracy within each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-creating identification elements for each time period before actual recognition tasks. This preliminary organization of identification elements by time period eliminates the need for complex real-time processing decisions, reducing device complexity while ensuring measurement precision is maintained through pre-prepared accurate reference data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS9443145B2Person recognition apparatus, person recognition method, and non-transitory computer readable recording medium
Publication Date: 2016.09.13 FUJIFILM CORP
  • US9443145B2 patent drawing
  • US9443145B2 patent drawing
  • US9443145B2 patent drawing

AI summary

There are provided a person recognition apparatus, method, and a non-transitory computer readable recording medium which can perform accurate person recognition according to a time dependent face change. A sorting section sorts a plurality of images by shooting date and time. A group division section divides the plurality of images into a plurality of groups according to a predetermined shooting date and time range. A face recognition section extracts feature amounts by face recognition for each group. An in-group person determination section determines a person having a similarity of a predetermined reference threshold value or higher as the same person and integrates the feature amounts relevant to the person for each group. An inter-group person recognition section recognizes persons having a similarity of a predetermined recognition threshold value or higher as the same person between two groups based on the feature amounts integrated in adjacent groups.