Object Position Estimation via Fusion of Visual and Relationship Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing techniques fail to accurately identify the position of objects in images, especially when objects are partially hidden, due to insufficient training data and high labor costs in preparing object images and training information.

Innovation Solution

A position estimation device and method that fuse position information, visual information, and relationship information of a subject object with a target object to estimate the target object's position using an object position estimator, with a parameter update unit optimizing the estimation by reducing the distance between estimated and correct position information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional object detection methods are used, then training data preparation is required, but labor costs and time consumption increase significantly

Engineering Contradiction:

Improveobject position identification accuracyVSAvoidtime for preparing training data

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-extracting visual features from images and pre-computing relationships between objects before actual detection. The visual feature extractor learns to represent objects, and the relationship extractor pre-establishes spatial and semantic relationships, so that during actual object detection, only new information needs to be processed rather than preparing all training data from scratch.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces intermediary components including a visual feature extractor that mediates between raw images and object detection, and a relationship extractor that mediates between object pairs. These intermediaries process and transform information in a standardized way, reducing the need for manual training data preparation while maintaining detection accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If objects are partially hidden in images, then detection becomes more difficult, but conventional methods still fail to identify position

Engineering Contradiction:

Improvedetection reliabilityVSAvoiddifficulty of detecting hidden objects

Core Design Contradiction:

ReliabilityVSDifficulty of detecting and measuring

Solution Approach 1:

The system employs feedback mechanisms where the relationship between detected subject objects and target objects is continuously refined. The relationship extractor analyzes spatial and semantic relationships between object pairs, and this information feeds back to improve the detection of hidden objects by considering contextual relationships rather than relying solely on direct visual detection of the hidden object itself.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent transitions from two-dimensional image analysis to three-dimensional spatial relationship analysis by extracting and utilizing relationship information between objects. This adds a dimensional layer of understanding that helps identify hidden objects through their relationships with visible objects, effectively solving the detection problem for partially hidden objects.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Measurement precision

If extensive training data is collected, then model accuracy improves, but preparation labor and cost increase

Engineering Contradiction:

Improveposition estimation accuracyVSAvoidease of data preparation

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The system performs self-service by automatically extracting visual features and relationships from images without requiring manual annotation or preparation of training data. The visual feature extractor and relationship extractor automatically process image data, generating the necessary training information themselves, which eliminates the need for human labor in data preparation while maintaining model accuracy.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent changes the parameters of data processing by transforming raw image data into extracted visual features and relationship information. This parameter transformation allows the system to work with automatically generated data representations rather than requiring manually prepared training data, significantly reducing preparation labor while maintaining the accuracy needed for position estimation.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20240371148A1Location estimation device, location estimation learning device, location estimation method, location estimation learning method, location estimation program, and location estimation learning program

Publication Date: 2024.11.07 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

It is possible to identify a position of a target object that is difficult to recognize.A position estimation device includes: an information fusion unit that generates fusion information in which position information of a subject object that is an object corresponding to a subject, visual information of the subject object, and relationship information indicating a relationship with a target object paired with the subject object are fused; and an object position estimation unit that estimates a position of the target object by using an object position estimator learned in advance on the basis of the fusion information.