3D Information Acquisition via 2D Key Point Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for acquiring 3D information of objects, especially dynamic objects, face challenges due to high computational costs, low accuracy, and susceptibility to ambient light and other factors, resulting in poor robustness and accuracy of the acquired 3D information.
Innovation Solution
A method that involves extracting 2D key points from an image, matching them with corresponding 3D key points in a preset 3D model, determining reference attitudes and positions, calculating reprojection error values, and selecting a target attitude and position to acquire accurate 3D information, which includes depth information and attitude, using a computer device with a processor and memory executing a computer program.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to acquire 3D information of dynamic objects, then computational cost is high, but accuracy and robustness are poor
Solution Approach 1:
The patent segments the 3D information acquisition process into distinct modules: 2D key point detection, candidate 3D model selection, reprojection error calculation, and optimization. This segmentation allows each module to be optimized independently, reducing overall computational complexity while maintaining accuracy through specialized processing at each stage.
Solution Approach 2:
The patent performs preliminary actions by pre-selecting candidate 3D models that match the detected 2D key points before full 3D reconstruction. This preliminary filtering reduces the search space and computational load for the subsequent optimization steps, while still achieving high accuracy through the reprojection error minimization on the reduced candidate set.
2Reliability
If traditional methods are used to acquire 3D information, then computational cost is high, but robustness to ambient light and other factors is poor
Solution Approach 1:
The patent implements feedback through the reprojection error calculation, where the 3D model is repeatedly projected onto the 2D image plane and the error between projected and actual key points is used to refine the 3D parameters. This iterative feedback mechanism enhances robustness by continuously correcting for disturbances like ambient light variations, while the feedback loop operates on a computationally efficient formulation.
Solution Approach 2:
The patent changes parameters by optimizing 3D model parameters (position, orientation, scale) based on reprojection error minimization rather than directly processing raw image data. This parameter transformation approach improves robustness by working in a more stable parameter space that is less sensitive to ambient light and other environmental factors, while reducing computational complexity through dimensionality reduction.
3Measurement precision
If complex algorithms are used to solve 3D information, then accuracy may improve, but efficiency decreases
Solution Approach 1:
The patent segments the complex 3D reconstruction problem into manageable stages: 2D key point detection, candidate model selection, and optimized parameter estimation. Each segment uses algorithms tailored to its specific requirements, achieving high overall precision without the computational burden of applying complex algorithms throughout the entire pipeline.
Solution Approach 2:
The patent transforms the complex 3D reconstruction problem into a parameter optimization problem where only a few key parameters (position, orientation, scale) need to be estimated. This parameter reduction maintains precision by focusing computational effort on the most critical parameters while significantly improving efficiency by reducing the dimensionality of the optimization space.
Data Source
AI summary
The present disclosure provides a method and device for acquiring 3D information of an object. The method includes: extracting two-dimensional (2D) key points of the object based on the image, and determining a candidate 3D model set matching the image; determining a plurality of first reference attitudes and positions of each candidate 3D model according to the 3D key points and the 2D key points; acquiring a plurality of reprojection error values between each candidate 3D model and the object at the plurality of first reference attitudes and positions; determining a first target attitude and position and a first target 3D model corresponding to a minimum reprojection error value in the first reprojection error value set; and acquiring the 3D information of the object based on the first target attitude and position and the first target 3D model.