Visual Localization via 3D Model Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current localization methods, such as GPS, suffer from accuracy issues in complex urban environments due to signal reflection or obstruction, and visual-based localization methods face scalability challenges with large databases and computational constraints.
Innovation Solution
The method involves capturing an image with an image capturing device and comparing it to a plurality of artificial images generated from a three-dimensional model. By analyzing attributes such as color, texture, and feature classes, the method identifies matching artificial images to determine the location and pose of a mobile apparatus.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS is used for localization, then real-time location data is provided, but accuracy deteriorates in complex urban environments due to signal reflection or obstruction
Solution Approach 1:
The patent introduces visual features (landmarks, buildings, roads) as an intermediary between GPS and the mobile device for localization. Instead of relying solely on GPS signals that are blocked or reflected by buildings, the system uses visual features that can be captured by the device's camera and matched against a pre-built visual map, providing an alternative localization path that works in urban canyons where GPS fails
Solution Approach 2:
The patent replaces the radio-based GPS localization mechanism with a visual-based localization mechanism using the device's camera and image processing algorithms. This substitution allows the system to leverage the device's existing visual sensors and processing capabilities to achieve accurate localization without depending on satellite signals that are vulnerable to urban obstruction
2Measurement precision
If visual-based localization methods are implemented, then localization accuracy is improved, but scalability deteriorates due to large database requirements and computational constraints
Solution Approach 1:
The patent segments the visual map database into spatially organized regions or tiles that can be loaded on-demand based on the device's current location and movement patterns. Instead of storing and processing all visual features globally, the system divides the map into manageable segments, reducing the database size and computational burden on mobile devices while maintaining comprehensive coverage for accurate localization
Solution Approach 2:
The patent performs preliminary processing of visual features during map creation, pre-computing and storing extracted visual descriptors, feature hierarchies, and spatial relationships in an optimized format. This preliminary action reduces the computational complexity of real-time matching by preparing the data in advance, allowing the mobile device to perform faster comparisons without requiring excessive processing power or memory
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
An apparatus receives a first image captured by an image capture device of a mobile apparatus. The first image corresponds to surroundings of the mobile apparatus. Each artificial image of a first plurality of artificial images respectively is a two-dimensional projection of a three-dimensional model from a perspective of an image position and image pose. The apparatus determines one or more first image attributes respectively for one or more sections of the first image; identifies at least one artificial image of the first plurality of artificial images that has one or more artificial image attributes that substantially match the one or more first image attributes for corresponding sections of the at least one artificial image and the first image; and determines a location and/or pose of the mobile apparatus based at least in part on the image position and/or the image pose associated with the at least one artificial image.