Visual Localization via Integrated Street View and Aerial Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current visual localization methods face challenges in generating accurate maps for precise positioning, especially on sidewalks, due to limitations in street view images and aerial photo-based 3D models, such as missing sidewalk views and distorted low-rise building textures.
Innovation Solution
A method and system that generate a 3-dimensional feature point map by integrating street view images and 3D model data from aerial photos, compensating for viewpoint differences to create robust map data for localization, allowing precise estimation of 3D positions and poses using feature point matching and error optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a map is generated using only street view images, then the map covers road perspectives, but sidewalk views are missing and low-rise building textures are distorted
Solution Approach 1:
The patent merges street view images and aerial photos into a single integrated map data structure. The map generation unit combines these different data sources, allowing the system to leverage both the ground-level perspective of street views and the top-down coverage of aerial imagery, thereby eliminating sidewalk view gaps while maintaining building texture fidelity
Solution Approach 2:
The patent introduces map data as an intermediary that bridges street view images and aerial photos. This intermediate representation integrates information from both sources, enabling the localization unit to access comprehensive spatial information including sidewalk areas that would be invisible in street view images alone
2Area of stationary object
If a map is generated using only aerial photos, then complete area coverage is achieved, but sidewalk-level detail and texture accuracy are lost
Solution Approach 1:
The patent merges aerial photos and street view images to achieve both complete area coverage and high texture accuracy. The integration process combines the broad spatial coverage of aerial imagery with the detailed ground-level textures from street views, ensuring that building facades and sidewalk features are rendered with appropriate detail
Solution Approach 2:
The patent applies local quality by using different data sources for different spatial regions. Aerial photos provide accurate coverage for open areas and rooftops, while street view images provide enhanced texture detail for building facades and sidewalk-level features, creating a map with locally optimized quality
3Reliability
If feature point maps from different viewpoints are integrated, then localization robustness improves, but position difference compensation complexity increases
Solution Approach 1:
The patent introduces a map generation unit as an intermediary that automatically handles the complex task of integrating feature point maps from different viewpoints. This unit performs position difference compensation and data fusion, shielding the localization process from integration complexity while maintaining robustness benefits
Solution Approach 2:
The patent creates a unified map data structure that copies and integrates information from multiple viewpoint-specific feature point maps. By maintaining standardized data formats across different viewpoints and automatically reconciling position differences, the system achieves localization robustness without exposing complexity to the localization process
Data Source
AI summary
A visual localization method includes generating a first feature point map by using first map data calculated on the basis of a first viewpoint; generating a second feature point map by using second map data calculated on the basis of a second viewpoint different from the first viewpoint; constructing map data for localization having the first and second feature point maps integrated with each other, by compensating for a position difference between a point of the first feature point map and a point of the second feature point map; and performing visual localization by using the map data for localization.