Cross-World Visual Positioning Using Scene Graph Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional visual positioning techniques are limited to cameras in the real world and do not consider associating positions between the real and virtual worlds, leading to inaccuracies due to differences in fine structures between the two.
Innovation Solution
A visual positioning system that extracts feature amounts from images by abstracting them into scene graphs, securing consistency between real and virtual worlds by matching feature amounts at a higher level, allowing accurate association of camera viewpoints across both domains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature amounts are extracted directly from images without abstraction, then fine structural details are preserved, but differences between real and virtual worlds cause matching failures
Solution Approach 1:
The patent introduces scene graphs as an intermediary representation between raw images and feature amounts. The scene graph abstracts images into structured representations containing objects, their attributes, and spatial relationships, serving as a mediator that bridges real and virtual world images while filtering out fine structural differences that cause matching failures.
Solution Approach 2:
The patent transforms the parameter representation from pixel-level details to scene graph-level abstractions. By changing the parameter space from raw image pixels to structured scene elements (objects, attributes, relationships), the system achieves both preservation of essential features and elimination of world-specific variations.
2Measurement precision
If visual positioning is performed without scene graph abstraction, then processing speed is faster, but positioning accuracy between real and virtual worlds deteriorates
Solution Approach 1:
The patent segments the image processing task into distinct stages: scene graph generation (abstraction) and feature amount extraction (matching). This segmentation allows the system to perform abstraction once on real images to build scene graphs, then efficiently match virtual images against these structured representations, improving both accuracy and overall processing efficiency.
Solution Approach 2:
The patent performs preliminary action by generating scene graphs from real images in advance before the actual positioning task. This pre-processing creates a structured reference framework that enables faster and more accurate matching during the actual positioning operation, separating the computationally intensive abstraction phase from the matching phase.
Data Source
AI summary
A first world is one of a real world and a virtual world simulating the real world, and a second world is another of them. A first image is captured by a first camera in the first world, and a second image is captured by a second camera in the second world. A visual positioning system executes common processing that generates a scene graph representing a positional relationship between objects included in the image and extracts a feature amount of the scene graph. The visual positioning system performs matching between a first feature amount extracted by the common processing on the first image and a second feature amount extracted by the common processing on the second image, and then associates the first camera position in the first world and the second camera position in the second world with each other based on a result of the matching.