3D Point Cloud Registration for Accurate Intraoral Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional image generation methods using intraoral scanners often result in poor image quality due to accumulated errors from the scanner's movement, leading to deformed images.
Innovation Solution
A method and system that capture multiple images, generate point clouds with initial pose information, adjust pose information through registration operations based on correlations among the point clouds, and combine them to generate a high-quality three-dimensional image using a detector, processor, and display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If the scanner moves continuously to capture multiple images in limited oral space, then the coverage of the three-dimensional image is improved, but the moving distance increases causing accumulated errors and image deformation
Solution Approach 1:
The patent segments the scanning process into multiple overlapping image captures, where each image is processed independently into a point cloud with pose information. This segmentation allows for localized processing and reduces the cumulative error effect across the entire scanning path.
Solution Approach 2:
The patent implements a feedback mechanism through pose graph optimization, where the system continuously refines the pose information of each point cloud based on correlations with neighboring point clouds. This feedback loop corrects accumulated errors by optimizing the global consistency of all point clouds simultaneously.
2Loss of information
If multiple images are captured to generate a complete three-dimensional image, then the completeness of the model is improved, but the combination of multiple point clouds introduces accumulated errors
Solution Approach 1:
The patent merges multiple point clouds into a single three-dimensional model through a unified pose graph optimization framework. By establishing correlations between all point clouds and optimizing their relative poses simultaneously, the system achieves accurate merging that preserves completeness while minimizing accumulated errors through global optimization.
3Area of stationary object
If the scanner moves longer distances to capture all oral surfaces, then the completeness of the three-dimensional image is improved, but the errors accumulate during data combination
Solution Approach 1:
The patent transitions from traditional sequential processing to a multi-dimensional approach by constructing a pose graph that connects all point clouds in a global optimization framework. This dimensional expansion allows simultaneous consideration of all point cloud relationships, enabling error correction across the entire scan while maintaining comprehensive coverage.
Data Source
AI summary
A three dimensional image generation method can include capturing a plurality of images of an object, generating a plurality of point clouds corresponding to a first sequence and a plurality pieces of first pose information according to the plurality of images, generating a second sequence according to a plurality of correlations among the plurality of images, performing a plurality of registration operations on the plurality of point clouds in the second sequence to adjust the plurality pieces of first pose information to generate a plurality pieces of second pose information, and combining the plurality of point clouds according to the plurality pieces of second pose information to generate a three dimensional image of the object.