3D Ray Cloud Visual Localization for Privacy-Preserving Pose Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing visual localization technologies using three-dimensional point clouds face privacy concerns due to the potential leakage of personal information through reverse reconstruction, and existing methods for using three-dimensional line clouds suffer from degraded performance in camera pose estimation.

Innovation Solution

A method is proposed to generate three-dimensional ray clouds by connecting anchor points with three-dimensional points, clustering these lines, and estimating camera pose based on sampled ray cloud clusters, while setting anchor points to protect privacy and maintain estimation performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If three-dimensional point cloud spatial maps are used for visual localization, then location identification accuracy is improved, but privacy security deteriorates due to reverse reconstruction attacks

Engineering Contradiction:

Improvelocation identification accuracyVSAvoidprivacy leakage risk

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent extracts only the necessary geometric features (three-dimensional lines) from the complete point cloud data, removing unnecessary information that could enable reverse reconstruction while preserving localization functionality

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces three-dimensional line clouds as an intermediary representation between the original point cloud and the localization algorithm, acting as a privacy-preserving mediator that maintains utility while reducing information exposure

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If three-dimensional line cloud maps are used instead of point clouds, then privacy security is improved, but calculation speed deteriorates

Engineering Contradiction:

Improveprivacy leakage riskVSAvoidcalculation speed

Core Design Contradiction:

Object-affected harmful factorsVSProductivity

Solution Approach 1:

The patent segments the continuous line cloud data into discrete, manageable features that can be processed more efficiently by localization algorithms, improving calculation speed while maintaining privacy protection

Inventive Principle:

Principle #1Segmentation

3Object-affected harmful factors

If random anchor points are used in ray cloud generation, then privacy protection is maintained, but pose estimation performance deteriorates

Engineering Contradiction:

Improvereverse reconstruction accuracyVSAvoidpose estimation accuracy

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent applies different quality requirements to different parts of the system: using non-random anchor points only where needed for pose estimation while maintaining overall privacy protection through the ray cloud structure

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent performs preliminary selection of anchor points from the ray clouds before pose estimation, pre-identifying useful anchor points that balance privacy protection with estimation accuracy

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250329045A1Visual localization method using 3D ray clouds and apparatus for executing the same

Publication Date: 2025.10.23 INDUSTRY UNIVERSITY COOPERATION FOUNDATION HANYANG UNIVERSITY

AI summary

According to one embodiment of the present disclosure, A visual localization method comprises generating at least two or more anchor points from three-dimensional point clouds; generating three-dimensional ray clouds by connecting three-dimensional points included in the three-dimensional point clouds with one of the generated anchor points; extracting feature points of an input image; and clustering a plurality of lines included in the three-dimensional ray clouds based on the at least two or more anchor points, sampling two ray cloud clusters out of the clustered ray cloud clusters, and estimating a pose of a camera that captured the input image based on the sampled ray cloud clusters and the feature points.