Image Management System Using Multi-Feature Subject Indexing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video management systems struggle to effectively identify and extract images of specific subjects, such as building construction materials, from recorded videos, especially when the subjects have similar shapes or are not captured as part of the index information.

Innovation Solution

A system that uses a convolutional neural network for subject recognition, incorporating image capture position and time information to determine the position of subjects within images, and stores index information including target identifiers, positions, and capture times, allowing for efficient image retrieval based on search instructions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If facial image extraction is used for indexing video content, then images of specific persons can be easily retrieved, but images of subjects with similar shapes (such as building construction materials) cannot be effectively identified

Engineering Contradiction:

Improvesubject identification accuracyVSAvoidapplicability to different subject types

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system changes the indexing parameters from facial features to multiple attributes including shape features, color, texture, and position information. This allows the same indexing mechanism to effectively identify both human faces and building construction materials by adapting to different subject types through parameter adjustment

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The indexing system is designed to be universal by extracting multiple types of features (shape, color, texture, position) that can apply to various subject types. The system can handle both facial images and building construction materials using the same multi-feature extraction and matching mechanism

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Ease of operation

If comprehensive index information is stored for all possible subjects, then any subject can be retrieved, but the system complexity and storage requirements increase significantly

Engineering Contradiction:

Improveimage retrieval capabilityVSAvoidindexing system complexity

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The system extracts only the most relevant features (shape, color, texture, position) needed for identification rather than storing all possible attributes. This selective extraction reduces indexing complexity while maintaining effective retrieval capability for construction materials and other subjects

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The indexing process is segmented into distinct feature extraction steps (shape extraction, color extraction, texture extraction, position determination) that can be applied independently and组合ed. This modular approach reduces overall system complexity by breaking down the complex indexing task into manageable segments

Inventive Principle:

Principle #1Segmentation

3Productivity

If only facial images are used as index information, then person-specific search is efficient, but subjects that are not captured as facial images (such as building materials) cannot be indexed

Engineering Contradiction:

Improvesearch efficiencyVSAvoidexclusion of non-facial subjects

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The system changes the indexing parameters from facial-specific features to general object features including shape, color, texture, and position. This parameter transformation enables the system to index and search for building construction materials and other non-facial subjects while maintaining search efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

Instead of trying to make all subjects fit the facial recognition framework, the system inverts the approach by using general feature extraction that can accommodate facial images as one subset among many other subject types. This inversion allows building materials and other objects to be indexed using the same mechanism that indexes faces

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS11227007B2System, method, and computer-readable medium for managing image

Publication Date: 2022.01.18 OBAYASHI CORP

AI summary

A system for managing images obtains an image that is captured by an image capture device at an image capture position and an image capture time; and stores the image, the image capture position, and the image capture time. The system further determines a position of a subject included in the image based on the image capture position. The system further stores index information comprising a target identifier of the subject, the position of the subject, and the image capture time in association with the stored image. The system further determines index information corresponding to the search instruction when obtaining a search instruction of the image; and extracts the image corresponding to the determined index information.