Visual Localization Using 2D Line Features in Sparse Skyline Scenes

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Satellite map-based visual localization methods suffer from low localization success rate and accuracy, particularly in scenes with insufficient or sparse skylines.

Innovation Solution

A visual localization method that utilizes two-dimensional line feature information, such as boundaries between buildings and non-buildings, combined with location and magnetometer angle deflection data to determine the localization pose, employing semantic segmentation and iterative optimization techniques to enhance accuracy and robustness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If satellite map-based visual localization is used, then large scene localization is enabled, but localization success rate and accuracy deteriorate

Engineering Contradiction:

Improvelocalization coverage areaVSAvoidlocalization accuracy

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent segments the localization process into multiple stages: first using satellite maps for large-area coverage, then using two-dimensional line features (building boundaries, road boundaries) for precise localization. This multi-stage segmentation allows the system to achieve both large coverage area and high localization accuracy by applying different methods at different stages.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transitions from three-dimensional satellite map data to two-dimensional line feature extraction from images. By projecting 3D building models to 2D boundaries and matching them with 2D line features in captured images, the system achieves more precise localization while maintaining large scene coverage capability.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Area of stationary object

If traditional satellite map-based localization is used, then large scene localization is achieved, but localization robustness deteriorates in scenes with sparse skylines

Engineering Contradiction:

Improvelocalization coverage areaVSAvoidlocalization robustness

Core Design Contradiction:

Area of stationary objectVSReliability

Solution Approach 1:

The patent introduces two-dimensional line features (building boundaries, road boundaries, sidewalk boundaries) as intermediary elements between the satellite map and the localization process. These line features serve as reliable matching markers even in scenes with sparse skylines, thereby improving localization robustness while maintaining large scene coverage.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the localization parameters from relying on skyline features to utilizing two-dimensional line features such as building boundaries and road boundaries. This parameter change makes the localization system more robust in environments where traditional skyline-based methods fail, while still covering large areas through satellite map integration.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12591996B2Visual localization method and apparatus

Publication Date: 2026.03.31 HUAWEI TECH CO LTD

AI summary

A visual localization method includes obtaining an image captured by a terminal device; obtaining two-dimensional line feature information of the image that includes at least one of information about a boundary between a building and a non-building or information about a boundary between a non-building and a non-building; and determining a localization pose of the terminal device based on location information of the terminal device, magnetometer angle deflection information of the terminal device, a satellite map, and the two-dimensional line feature information.