3D AR Scene Registration Using Plane-Based Rotation Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality (AR) systems struggle to accurately register virtual elements in real-world scenes, requiring precise alignment of AR scenes to real-world structures, which is computationally complex and prone to drifting due to changes in viewpoint or location.
Innovation Solution
A two-phased registration method involving a principal rotation angle selection followed by a fine-tune rotation angle to align a geometrical dataset with a 3D model, using translation vectors to minimize distance between corresponding planes, and dynamic adjustment to maintain registration accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional AR registration methods are used to align virtual elements with real-world structures, then registration accuracy can be maintained, but computational complexity increases and the system becomes prone to drifting due to viewpoint changes
Solution Approach 1:
The patent segments the registration process into two distinct phases: a first phase that computes a preliminary transformation matrix using a subset of features, and a second phase that refines this matrix using all features. This segmentation reduces computational complexity by dividing the complex registration task into manageable stages, while maintaining accuracy through progressive refinement.
Solution Approach 2:
The patent performs preliminary computation of a transformation matrix using a subset of features before the final registration step. This preliminary action provides an initial alignment that reduces the computational burden of the subsequent refinement phase, preventing system drift while reducing overall computational complexity.
2Reliability
If traditional AR registration methods are used to ensure accurate alignment, then registration reliability can be maintained, but the system requires extensive computational resources and time
Solution Approach 1:
The registration process is divided into two phases: an initial phase using a subset of features for quick alignment, and a refinement phase using all features for final precision. This segmentation improves processing speed by avoiding the need to process all features simultaneously, while maintaining reliability through the refinement step.
Solution Approach 2:
The patent uses a subset of features in the first phase rather than processing all features immediately. This partial action provides sufficient initial alignment to establish a reliable transformation matrix, which is then refined in the second phase, improving overall processing efficiency without sacrificing registration reliability.
Data Source
AI summary
Disclosed herein are methods and systems for registering an Augmented Reality (AR) scene to a 3D model of a certain area, comprising accumulating a geometrical dataset of a certain area based on visual data captured by imaging sensors during an AR session, obtaining a 3D model of the certain area, computing a similarity score for the geometrical dataset compared to the 3D model rotated in several candidate principal rotation angles based on parallelism of corresponding planes identified in the geometrical dataset and in the rotated 3D model, selecting a highest score principal rotation angle, computing a translation vector for each of a plurality of fine-tune rotation angles with respect to the selected principal rotation angle based on a distance between the corresponding planes, selecting a lowest distance fine-tune rotation angle, and registering the geometrical dataset to the 3D model according to the selected fine-tune rotation angle and its translation vector.