Point Cloud Cleaning via Cell Voxel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile scanning technologies face issues such as varying scan density due to speed variations, noise from reflective surfaces and electronic interference, ghosting, and processing inefficiencies due to the large volume of points in point clouds, leading to inaccurate representations of indoor environments.
Innovation Solution
A method for cleaning point clouds using an electronic computing device, involving decomposition into cell voxels, removal of outlier voxels and points, subsampling, and elimination of double surfaces, which includes steps like initializing a connected components grid, using a floor plan to identify outliers, and estimating normals to remove artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mobile scanning is performed to capture complete 3D environment data, then coverage completeness is improved, but processing time and computational load increase due to millions of points generated
Solution Approach 1:
The point cloud is divided into multiple subsets or chunks, allowing parallel processing of different portions of the data. This segmentation enables the system to handle millions of points more efficiently by distributing the computational load across multiple processing units or time steps, thereby reducing overall processing time while maintaining complete coverage.
2Measurement precision
If point cloud density is increased to improve accuracy, then measurement precision is improved, but processing complexity increases due to larger data volume
Solution Approach 1:
The high-density point cloud is segmented into manageable subsets that can be processed independently. This allows the system to maintain the accuracy benefits of high-density scanning while reducing the computational complexity by handling smaller data portions in parallel or sequentially.
Solution Approach 2:
Different processing strategies are applied to different regions of the point cloud based on local characteristics. Areas requiring higher precision maintain full density, while less critical regions undergo downsampling or simplified processing, optimizing the balance between accuracy and computational complexity.
3Reliability
If multiple lasers are used to provide full coverage scan, then coverage completeness is improved, but processing time increases due to merging multiple point clouds
Solution Approach 1:
Point clouds from multiple lasers are processed as separate segments before being merged. This allows parallel processing of data from different laser sources, reducing the time required to handle the combined dataset while ensuring complete coverage from all directional sources.
Solution Approach 2:
Pre-processing steps such as filtering, downsampling, or feature extraction are performed on individual laser point clouds before merging. This preliminary action reduces the computational burden of the merging operation and improves overall processing efficiency while maintaining the comprehensive coverage provided by multiple lasers.
Data Source
AI summary
Methods for cleaning a point cloud using an electronic computing device are presented, the method including: causing the electronic computing device to decompose the point cloud into a number of cell voxels; removing a number of outlier cell voxels; subsampling the point cloud; removing a first number of outlier points; removing a number of double surfaces; and removing a second number of outlier points. In some embodiments, method further include outputting a clean point cloud to file. In some embodiments, the causing the electronic computing device to decompose the point cloud further includes: reading into memory a number of points corresponding with the point cloud; specifying a cell voxel size; creating a cell grid to chunk the point cloud into a number of smaller sets; sorting the number of points into the number of cell voxels; and merging all point clouds.