Image Processor Hierarchical Weak Classifier Target Registration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image processors face difficulties in detecting human and pet faces accurately, often requiring users to register precise domains, which can be operationally complex and lead to incomplete target detection due to unclear boundaries and incorrect domain assignments.

Innovation Solution

An image processor that includes a classifier with weak classifiers performing hierarchical detection processing, a search image generator to extract domains near user-specified positions, and a registration image generator to produce a registration image based on the largest processing domain hierarchy, allowing for accurate target detection even with slight user position deviations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If users manually specify precise domains for target detection, then detection accuracy improves, but operational complexity increases and users may incorrectly assign domains due to unclear target boundaries

Engineering Contradiction:
Improvedetection accuracyVSAvoidoperational complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs self-service by automatically specifying the search domain based on the user's simple position input, eliminating the need for users to manually define complex domain boundaries. The processor autonomously calculates the search domain range and performs detection within this automatically generated domain.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary action by pre-specifying the search domain before actual target detection occurs. The processor determines the appropriate search domain range based on the user's position input and detection requirements, preparing the detection framework in advance to guide subsequent detection operations.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If the search domain is expanded to include entire targets with unclear boundaries, then detection completeness improves, but processing time and computational load increase

Engineering Contradiction:
Improvedetection completenessVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system applies parameter changes by dynamically adjusting the search domain range based on detection requirements and target characteristics. The processor modifies the domain parameters to optimize the balance between including complete targets and maintaining efficient processing speed.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system uses partial action by specifying a search domain that is larger than the minimal target area but not excessively large. This partial expansion ensures complete target inclusion while avoiding unnecessary processing of irrelevant regions, achieving optimal detection completeness with reasonable computational cost.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If multiple weak classifiers are applied in sequence to improve detection accuracy, then detection precision improves, but processing complexity and time increase

Engineering Contradiction:
Improvedetection precisionVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system applies segmentation by dividing the detection process into multiple sequential weak classifier stages. Each weak classifier processes the data in sequence, with results passed to the next stage, breaking down the complex detection task into manageable segments that collectively achieve high precision.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system uses an intermediary approach by introducing a search domain specification mechanism that mediates between user input and classifier processing. This intermediary layer pre-processes and structures the input data, making subsequent classifier operations more efficient and reducing overall processing complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP2639744B1Image processor, image processing method, control program, and recording medium
Publication Date: 2019.04.10 OMRON CORP
  • EP2639744B1 patent drawingFigure 1
  • EP2639744B1 patent drawingFigure 2
  • EP2639744B1 patent drawingFigure 3

AI summary

A registration image including a desired target can easily be registered. A domain near an assigned position assigned by a user on a photographed image is extracted from the photographed image to generate a search image, a classifier performs the processing to the generated search image, and a processing domain having the largest number of hierarchies to which a weak classifier can perform the processing is extracted from the photographed image to generate the registration image.