3D Reconstruction Using Line Features and Quadtree Splitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current head-mounted displays, such as VR, AR, and MR devices, face inefficiencies and inaccuracies in three-dimensional reconstruction due to excessive redundant feature information and poor feature quality, limiting their application and user experience.
Innovation Solution
A three-dimensional reconstruction method that involves sliding calculation on image frames, point and line feature extraction, iterative quadtree splitting for circular regions of interest, feature screening, and using target features for reconstruction, thereby improving efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional visual SLAM technology is used for three-dimensional reconstruction, then the reconstruction can be performed on complicated scenes, but the process is time-consuming and has poor reconstruction effect due to excessive redundant feature information and poor feature quality
Solution Approach 1:
The patent divides the image processing into fixed-size windows and further segments features within each window using iterative quadtree splitting. This segmentation allows the system to focus computational resources on specific regions of interest rather than processing the entire image, reducing redundant calculations while maintaining reconstruction accuracy.
Solution Approach 2:
The patent extracts and screens features within each window to identify target features that meet quality criteria. By extracting only the necessary feature information and discarding redundant features through screening, the system reduces the amount of data processed while preserving the quality features needed for accurate reconstruction.
2Reliability
If all features in the image are used for three-dimensional reconstruction, then more feature information is available, but the reconstruction efficiency decreases due to excessive redundant feature information
Solution Approach 1:
The patent applies different processing strategies to different regions by dividing the image into windows and performing iterative quadtree splitting within each window. This allows the system to adaptively allocate computational resources to regions with higher feature quality potential, improving overall reconstruction reliability while maintaining efficiency.
Solution Approach 2:
The patent performs feature screening on all features within each window to identify target features, which may be fewer than the total features extracted. This partial action approach ensures that only high-quality features are used for reconstruction, improving reliability without requiring processing of all extracted features.
3Measurement precision
If feature screening is performed on all features in the image, then feature quality improves, but the processing time increases
Solution Approach 1:
The patent divides the image into fixed-size windows and performs feature screening independently within each window. This segmentation reduces the total processing time by allowing parallel processing of multiple windows and focusing screening efforts on localized regions rather than the entire image at once.
Solution Approach 2:
The patent performs point feature extraction and line feature extraction before iterative quadtree splitting and feature screening. This preliminary action organizes and pre-processes features in advance, making the subsequent screening process more efficient by having features ready in an organized structure.
Data Source
AI summary
A three-dimensional reconstruction method, comprising: performing slide calculation on an acquired current image frame to obtain to-be-processed windows; performing point feature extraction and line feature extraction on the to-be-processed windows, and determining corresponding features in each to-be-processed window; performing iterative quadtree splitting for circles on each target to-be-processed window to obtain circular regions of interest corresponding to each target to-be-processed window; performing feature screening on features in the target circular regions of interest to obtain target features corresponding to each target circular region of interest; and performing three-dimensional reconstruction corresponding to the current image frame by using the target features. By introducing line features in the feature extraction stage of three-dimensional reconstruction, the line segment features in the three-dimensional reconstruction scenes can be better perceived, and the reconstruction efficiency of simple regular three-dimensional models can be accelerated.