3D Point Cloud Segmentation Using Bounding Volumes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Segmenting multiple 3D objects represented by a large number of points is technically challenging, especially when they are adjacent or connected, as existing methods fail to efficiently distinguish individual components within a scanned object.
Innovation Solution
A 3D segmentation system that groups 3D points into sets by using a client device or server to create and transmit 3D volumes, reducing data transmission by limiting the number of points and using volume coordinates, thereby facilitating efficient segmentation of individual objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If segmentation is performed on multiple 3D objects represented by a large number of points, then individual objects can be separated, but processing time and computational resources increase significantly
Solution Approach 1:
The patent applies segmentation by dividing the point cloud data into multiple subsets, where each subset corresponds to a different 3D object. The system creates multiple 3D volumes (bounding boxes) that partition the spatial space, and assigns points to volumes based on their spatial coordinates. This segmentation approach enables parallel processing of different object subsets, significantly reducing overall processing time while maintaining segmentation accuracy.
Solution Approach 2:
The patent introduces a volumetric dimensionality approach by creating 3D bounding volumes that encompass groups of points. Instead of processing points individually in 3D space, the system organizes them into volumetric containers defined by coordinate boundaries. This dimensional organization allows for more efficient spatial queries and reduces the computational complexity of point-to-point comparisons.
2Loss of information
If all 3D points are transmitted for processing, then complete object representation is achieved, but data transmission volume increases
Solution Approach 1:
The patent extracts and transmits only the essential volumetric boundary information (minimum bounding boxes) rather than all individual point coordinates. The system determines 3D volumes that encompass groups of points and transmits these volume definitions along with representative point information. This extraction approach maintains complete object representation while dramatically reducing the volume of data that needs to be transmitted across the network.
Solution Approach 2:
The system performs preliminary processing to organize points into volumetric groups before transmission. By pre-segmenting the point cloud data into spatial volumes and identifying representative points for each volume, the system prepares the data in an optimized format that reduces redundant information. This preliminary organization allows the transmission of compressed representations that can be efficiently reconstructed and processed at the receiving end.
Data Source
AI summary
Embodiments are described for a graphic segmentation device comprising a memory and at least one processor coupled to the memory. The at least one processor is configured to receive point cloud data that represents a plurality of 3D objects and select one or more 3D volumes that segment a 3D object out of the plurality of 3D objects. Each of the one or more 3D volumes includes coordinates forming a 3D shape. The at least one processor is further configured to transmit a volume report to a graphic segmentation server, wherein the volume report includes the one or more 3D volume.