3D Point Cloud Segmentation for Thin-Walled Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional object segmentation algorithms face challenges in accurately segmenting thin-walled objects with hollowed-out structures due to poor differentiation between hollowed-out structures and small depressions, leading to increased computation and excessive sub-object segmentation, which is inefficient for simulation environments.
Innovation Solution
A method involving scaling a three-dimensional object's point cloud to the same scale in all directions, using three-dimensional space ellipse fitting to determine scaling directions, and segmenting the object based on a scaled point cloud to improve segmentation accuracy and reduce excessive sub-object segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional approximate convex set segmentation algorithms are used, then segmentation speed is improved, but segmentation accuracy deteriorates for thin-walled objects with hollowed-out structures
Solution Approach 1:
The patent applies preliminary action by performing scaling transformation on the point cloud before segmentation. The method determines three scaling directions, scales the point cloud along these directions to obtain a scaled point cloud with uniform scale, and then performs segmentation. This preliminary scaling action prevents hollowed-out structures from being misrecognized as depressions, thereby improving segmentation accuracy while maintaining efficient processing speed.
2Measurement precision
If precise convex set segmentation algorithms are used, then segmentation accuracy is improved, but computation amount increases exponentially
Solution Approach 1:
The patent applies parameter changes by transforming the point cloud's scale parameters before segmentation. By determining three scaling directions and scaling the point cloud along these directions, the method changes the scale parameters of the point cloud to make hollowed-out structures distinguishable from depressions. This parameter transformation enables accurate segmentation using efficient algorithms without requiring computationally expensive precise convex set segmentation methods.
Data Source
AI summary
The present disclosure describes a three-dimensional object segmentation method and device, and medium, and relates to the field of computer vision (CV) technologies of an artificial intelligence (AI) technology. The method includes obtaining, by a device, a point cloud of a three-dimensional object. The device includes a memory storing instructions and a processor in communication with the memory. The method includes determining, by the device, three scaling directions of the point cloud; scaling, by the device, positions of points in the point cloud along the three scaling directions to obtain a scaled point cloud, so that the scaled point cloud has the same scale in the three scaling directions; and segmenting, by the device, the three-dimensional object based on the scaled point cloud to obtain a segmentation result.