Cluster-Based Point Cloud Comparison for Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D scanning technologies face challenges in accurately detecting displacements and defects in point clouds due to noise in measurement data, leading to frequent false-positive detections and missed true defects, especially on large or flat surfaces.
Innovation Solution
The method employs cluster-based cloud-to-cloud comparison using multi-radii cluster matching and region growing segmentation to identify displaced points, reducing noise effects and avoiding data smoothing, which allows for the detection of small feature displacements undetectable by human inspection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional point-by-point cloud-to-cloud comparison is used, then the detection process is simple, but measurement precision deteriorates due to noise causing false-positive detections and missed defects
Solution Approach 1:
The patent segments the point cloud data into multiple clusters based on spatial proximity and similarity metrics. Instead of comparing individual points, the system groups points into clusters and performs comparison at the cluster level, which reduces the impact of noise on individual points and improves both measurement precision and reliability in defect detection
Solution Approach 2:
The patent merges multiple comparison results from different clusters to form a comprehensive defect detection outcome. By combining evidence from multiple clusters and using voting mechanisms or aggregated metrics, the system achieves more reliable defect detection while reducing false positives that would occur with isolated point comparisons
2Reliability
If data smoothing is applied to reduce noise, then false-positive detections decrease, but manufacturing precision deteriorates due to loss of small feature details
Solution Approach 1:
The patent segments point cloud data into clusters that preserve local geometric features while reducing noise impact. By operating at the cluster level rather than individual point level, the system maintains sensitivity to small features without requiring smoothing operations that would blur or eliminate fine details
Solution Approach 2:
The patent applies different comparison strategies and noise tolerance thresholds to different regions of the point cloud based on local feature characteristics. Critical small features are evaluated with higher precision settings while less critical areas use more aggressive noise filtering, preserving manufacturing precision where needed while reducing false positives elsewhere
3Speed
If feature-based registration is used for handheld scanning, then registration speed is fast, but measurement precision deteriorates on large flat surfaces with few features
Solution Approach 1:
The patent implements a multi-functional registration approach that can operate in different modes depending on the scanning scenario. For handheld scanning, it uses feature-based registration for speed, while for large flat surfaces, it automatically switches to cluster-based comparison methods that provide higher precision without sacrificing excessive registration speed
Data Source
AI summary
Examples described herein provide a method that includes performing cluster matching with one or more cluster sizes for each of a plurality of points of a measurement point cloud. The method further includes determining, based on results of the multi-radii cluster matching, whether an object is displaced or whether the object includes a defect.