Point Cloud Spatial Alignment via Geometric Partitioning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Aligning point clouds representing complex objects, such as a tube, into a single coherent representation is computationally expensive due to the lack of a known spatial relationship between scans performed from varying positions and orientations, making existing methods like automated tracking systems or marker-based alignment impractical.

Innovation Solution

The method involves segmenting point clouds and datasets into partitions based on natural geometric features, generating feature descriptors, and comparing these descriptors to identify corresponding partitions, allowing for spatial alignment through translation and rotation adjustments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If point clouds are aligned using traditional methods (comparing individual points or manual marker-based approaches), then alignment accuracy can be achieved, but computational complexity and processing time increase significantly

Engineering Contradiction:
Improvealignment accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the point cloud into multiple partitions based on geometric features (planes, cylinders, spheres). Instead of comparing all points across entire point clouds, the method segments both point clouds into comparable geometric parts, dramatically reducing the computational search space while maintaining alignment accuracy through feature-based matching.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts geometric feature descriptors (normal vectors, curvature, radius, plane equations) from partitioned point clouds. By taking out only the essential geometric characteristics rather than processing all point coordinates, the method achieves accurate alignment with reduced computational complexity.

Inventive Principle:
Principle #2Taking out (Extraction)

2Ease of operation

If automated tracking systems or robotic arms are used to obtain scanner position and orientation information, then point cloud alignment can be achieved, but system cost and complexity increase

Engineering Contradiction:
Improvealignment process automationVSAvoidsystem complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent enables point clouds to self-align by automatically extracting and comparing geometric feature descriptors from the point cloud data itself. The system uses intrinsic geometric properties (normals, curvature, shape descriptors) to determine spatial relationships without requiring external tracking systems, robotic arms, or manual marker placement, making the process self-sufficient and cost-effective.

Inventive Principle:
Principle #25Self-service

3Ease of operation

If optical markers are projected onto the object surface for alignment, then spatial relationship can be established, but the method fails when the object or holder blocks the projection

Engineering Contradiction:
Improvealignment processVSAvoidalignment reliability
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent uses the inherent geometric features of the object surface itself (planes, cylinders, spheres) as alignment references. By extracting feature descriptors directly from the point cloud data representing the object's natural geometry, the method eliminates the need for external markers that could be blocked, ensuring reliable alignment regardless of object shape or scanning conditions.

Inventive Principle:
Principle #25Self-service

4Manufacturing precision

If all point clouds are assembled and aligned as complete datasets, then complete surface representation is achieved, but processing time and computational resources increase

Engineering Contradiction:
Improvesurface representation completenessVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent partitions point clouds into geometric feature-based segments (planes, cylinders, spheres) before alignment. By working with these smaller, feature-specific partitions rather than complete point clouds, the method reduces computational complexity and processing time while still achieving complete and accurate surface representation through subsequent assembly of aligned partitions.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4231238A1Method for spatially aligning a point cloud
Publication Date: 2023.08.23 SIEMENS AG

AI summary

A method for aligning a point cloud that represents a three-dimensional object or a part of such an object, to a second dataset that represents said object or a part of said object comprises the steps of segmenting the point cloud and the second dataset into partitions, each partition comprising a subset of the points of the point cloud or a subset of the dataset, respectively, generating feature descriptors for some or all of the partitions of the point cloud and for some or all of the partitions of the second dataset, wherein the feature descriptors are derived from characteristics of the points of the partition, comparing feature descriptors of the point cloud and the second dataset to identify partitions of the point cloud and second dataset that correspond to each other and performing a spatial alignment using the identified partitions.