Autofocus System with Periodic Face Recognition for Tracking Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional autofocus systems with face recognition and tracking functions face challenges in preventing erroneous tracking of unintended targets and efficiently changing the tracking target, requiring significant time and labor for operators to correct mistaken tracking and switch between faces.

Innovation Solution

An autofocus system that integrates face recognition processing with object tracking, allowing continuous detection of the specific person's face using face recognition processing even during tracking, and enabling quick and easy target changes through a designation operation on a screen, such as a tap operation, to prevent incorrect tracking and facilitate rapid target switching.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If face recognition processing is executed repeatedly during object tracking to prevent erroneous tracking, then tracking accuracy is improved, but processing time increases and tracking speed decreases

Engineering Contradiction:
Improvetracking accuracyVSAvoidtracking speed
Core Design Contradiction:
ReliabilityVSSpeed

Solution Approach 1:

The system executes face recognition processing at predetermined time intervals during object tracking, rather than continuously or only once. This periodic execution maintains tracking accuracy by regularly verifying the tracking target while controlling processing time and maintaining tracking speed through interval-based operations.

Inventive Principle:
Principle #19Periodic action

2Speed

If face recognition processing is executed once at the start and then object tracking is used, then processing time is reduced and tracking speed is improved, but tracking accuracy deteriorates due to erroneous tracking

Engineering Contradiction:
Improvetracking speedVSAvoidtracking accuracy
Core Design Contradiction:
SpeedVSReliability

Solution Approach 1:

The system uses face recognition processing as a feedback mechanism during object tracking to verify whether the tracked object is still the correct face. By periodically checking the tracking target against registered face data, the system can detect and correct erroneous tracking while maintaining overall tracking speed through efficient interval-based operation.

Inventive Principle:
Principle #23Feedback

3Reliability

If operators manually correct mistaken tracking or change tracking targets, then tracking accuracy is maintained, but operational time and labor increase significantly

Engineering Contradiction:
Improvetracking accuracyVSAvoidoperational time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system automatically detects and corrects erroneous tracking through periodic face recognition processing without requiring operator intervention. The automated verification and correction mechanism eliminates manual operations for correcting tracking errors or changing tracking targets, significantly reducing operational time and labor while maintaining tracking accuracy.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2339826B1Autofocus system
Publication Date: 2012.09.05 FUJIFILM CORP
  • EP2339826B1 patent drawingFigure 1
  • EP2339826B1 patent drawingFigure 2
  • EP2339826B1 patent drawingFigure 3

AI summary

An autofocus system which can prevent erroneous tracking of an unintended tracking target, is provided. In a recognition and tracking mode which uses an AF frame to automatically track a face of a specific person for which certification data is previously stored, a face of a specific person designated by an operator is detected by face authentication processing from a captured image obtained from a television camera . Thereafter, the face is detected from the captured image by a pattern matching processing in which an image of the face is used as a reference pattern for object tracking processing, and the AF frame is moved to the detected position. The face recognition processing is executed each time when the object tracking processing is executed a predetermined times, thereby preventing erroneous tracking by the object tracking processing.