AR 3D Map Generation Using ML to Exclude Movable Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Augmented reality (AR) systems face challenges in maintaining accurate localization due to changes in physical environments, such as moving objects, which can lead to failed comparisons between digital representations and 3D maps, resulting in incorrect localization of AR environments across different devices.
Innovation Solution
A method utilizing machine learning models to identify regions with visual data likely to change in the digital representation of an AR environment, removing these areas to generate a reduced digital representation, which is then used to create a 3D map that excludes movable objects, thereby improving the accuracy of AR localization by focusing on stationary features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the complete digital representation including all visual objects is used to create the 3D map, then the map contains comprehensive environmental information, but the localization accuracy deteriorates when movable objects are present
Solution Approach 1:
The patent segments the digital representation into static and dynamic regions using machine learning classification. Static regions containing stable visual features are extracted for 3D map creation, while dynamic regions with movable objects are excluded. This segmentation allows the system to maintain comprehensive environmental information in the full digital representation while using only reliable static features for localization comparisons.
Solution Approach 2:
The patent extracts and removes dynamic regions identified as containing movable objects from the digital representation before generating the 3D map. This extraction process isolates the problematic movable objects while preserving the static environmental features needed for accurate localization, resolving the contradiction between comprehensive information and localization reliability.
2Reliability
If machine learning models are used to identify and remove dynamic regions, then localization accuracy improves, but system complexity increases
Solution Approach 1:
The patent introduces machine learning models as intermediary components that automatically classify regions as static or dynamic. These models act as mediators between the raw digital representation and the 3D map generation process, enabling the system to handle complex classification tasks while maintaining a relatively simple overall architecture. The intermediary ML models manage the complexity by encapsulating the classification logic in specialized modules.
3Measurement precision
If dynamic regions are removed from the digital representation, then comparison accuracy improves, but the digital representation becomes less comprehensive
Solution Approach 1:
The patent applies local quality by treating different regions of the digital representation differently based on their characteristics. Static regions are processed with high fidelity for inclusion in the 3D map, while dynamic regions are identified and excluded. This local differentiation ensures that comparison accuracy is maximized in the relevant static areas without unnecessarily reducing the overall information content of the complete digital representation.
Data Source
AI summary
According to an aspect, a method for creating a three-dimensional map for augmented reality (AR) localization includes obtaining a digital representation of a scene of an AR environment, where the digital representation has been captured by a computing device. The method includes identifying, using a machine learning (ML) model, a region of the digital representation having visual data identified as likely to change, and removing a portion of the digital representation that corresponds to the region of the digital representation to obtain a reduced digital representation, where the reduced digital representation is used to generate a three-dimensional (3D) map for the AR environment.