Reference-Guided Point Cloud Upsampling Under Noisy Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating dense point clouds using high-performance hardware are costly, and existing computer vision techniques to increase point cloud density often result in inaccurate representations due to noise, especially in non-laboratory environments.
Innovation Solution
Utilize a reference point cloud of a reference object to learn features of a target object's point cloud, transforming it into a denser representation using deep learning to enhance accuracy and reduce hardware costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-performance hardware devices are used to directly generate dense point clouds, then the point cloud density and 3D representation quality are improved, but the hardware cost increases excessively
Solution Approach 1:
The patent uses a reference point cloud from a reference object as a template to generate the target point cloud, rather than directly capturing with expensive high-performance hardware. This copying approach allows dense point cloud generation using lower-cost devices by transferring structure from the reference object.
Solution Approach 2:
The patent introduces a reference point cloud as an intermediary between the sparse target point cloud and the final dense representation. The reference point cloud serves as a mediator that provides geometric guidance to transform sparse points into dense point clouds through feature matching and coordinate transformation.
2Ease of manufacture
If computer vision technology is used to increase point cloud density from sparse recordings, then hardware cost is reduced, but measurement accuracy deteriorates due to noise
Solution Approach 1:
The patent employs feedback through iterative optimization where the generated dense point cloud is compared against the reference point cloud, and the transformation parameters are refined based on this comparison. This feedback mechanism reduces noise accumulation and improves accuracy despite using lower-cost hardware.
Solution Approach 2:
The patent transforms the problem from direct sparse-to-dense generation to a parameter optimization problem, where transformation parameters (rotation, translation, scaling) are adjusted to minimize the difference between generated and reference point clouds. This parameter-based approach reduces noise sensitivity.
Data Source
AI summary
Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for processing point clouds. The method includes performing upsampling on a first feature of a first point cloud of a target object. The method further includes determining a reference feature of a second point cloud of a reference object. The method further includes determining a second feature based on the first feature subjected to upsampling and the reference feature. The method further includes generating a third point cloud of the target object based on the second feature and the second point cloud of the reference object, where the third point cloud has a larger number of points than the first point cloud. Through embodiments of the present disclosure, a point cloud of the target object can be made denser, with increased accuracy, thereby providing a more detailed description of the target object.