Direct Plane Fitting for Low-Texture, Occluded Surface Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating 3D models from 2D images, such as indirect plane fitting, are slow and inaccurate, particularly for low-texture structures and scenes with occlusions, due to intensive computation requirements and the inability to identify planar surfaces effectively.
Innovation Solution
A direct plane fitting method that utilizes two images captured from different known positions to identify common regions, compute gradients, and project rays to intersect at a world-point, using a similarity metric to determine the orientation of planar surfaces without intermediate feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If indirect plane fitting methods are used to generate 3D models from 2D images, then the ability to extract features and derive planar surfaces is improved, but the computation time and complexity increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for plane fitting directly from image pixels, bypassing the need for complex intermediate feature extraction steps. By taking out only the necessary pixel data and directly computing plane parameters through algebraic methods, the system achieves accurate results without the time-consuming intermediate processing of point clouds, line segments, or deep network features that characterize traditional indirect methods
Solution Approach 2:
Instead of following the traditional indirect approach of extracting features first and then fitting planes, the patent inverts the process by directly fitting planes to raw pixel data. This inversion eliminates the intermediate feature extraction step entirely, using direct algebraic methods to compute plane parameters from image coordinates, thereby dramatically reducing computation time while maintaining accuracy
2Difficulty of detecting and measuring
If traditional indirect plane fitting methods are used, then feature extraction capability is enhanced, but the method becomes inaccurate for low-texture structures and occluded scenes
Solution Approach 1:
The patent extracts plane parameters directly from raw pixel coordinates without relying on intermediate feature extraction. By taking out only the essential correspondences between points in different images and directly computing plane parameters through algebraic methods, the system avoids the pitfalls of traditional methods that fail when features are unavailable or unreliable in low-texture and occluded regions
Solution Approach 2:
The patent introduces an algebraic plane fitting model as an intermediary that directly relates image pixel coordinates to plane parameters. This intermediary model bypasses the need for complex feature extraction and processing, providing a robust mathematical framework that works reliably even when traditional feature-based methods fail in low-texture and occluded scenarios
3Device complexity
If indirect plane fitting with intensive computation is used, then comprehensive feature processing is achieved, but the overall processing speed decreases
Solution Approach 1:
The patent extracts only the minimal necessary information from images—specifically, pixel correspondences between multiple views—and directly computes plane parameters using efficient algebraic methods. By taking out only this essential data and avoiding complex intermediate processing steps, the system achieves high processing speed while maintaining the capability to handle comprehensive scene analysis
Solution Approach 2:
The patent replaces the mechanical, multi-step feature extraction and processing pipeline with a direct algebraic computation system. Instead of mechanically extracting features, clustering them, and then fitting planes, the system uses algebraic methods to directly compute plane parameters from pixel coordinates, substituting a computationally efficient mathematical approach for the traditional mechanical processing chain
Data Source
AI summary
Embodiments of the disclosed technology are directed to using pixel data from two or images of the same real-world region to determine the orientation of substantially planar surfaces in those images. The disclosed technology may be utilized to generate positions and/or orientations of planar surfaces based on multiple 2D images of the planar surface. Implementations can include modeling a planar surface, which can include capturing at least two images (from corresponding different, known positions), determining, from the images, regions that correspond to substantially the same portion of the real-world planar surface, computing a similarity metric for the two regions, and then determining the orientation of the planar surface based on maximizing the similarity metric over various different regions from the images. The described embodiments provide improvements in speed and accuracy compared to existing procedures and are able to find planar surfaces in both low-texture environments and occluded environments.