3D Point Cloud Object Extraction from Noisy Mobile Scans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D scanning technologies require specialized and costly hardware to generate high-quality 3D point clouds, which often contain noise, and non-specialized hardware like mobile devices produce even noisier clouds, making it difficult to process and extract objects of interest effectively.
Innovation Solution
A method for extracting an object-of-interest from a 3D point cloud using planar surface identification, distance and color thresholding, and cluster parameter calculation to remove noise and identify the object-of-interest, utilizing non-specialized hardware like mobile devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If non-specialized hardware like mobile devices is used to acquire 3D point clouds, then cost and accessibility are improved, but noise in the point cloud increases
Solution Approach 1:
The patent extracts and removes noise points from the 3D point cloud by identifying planar surfaces (such as ground planes) and filtering out data points that belong to these surfaces or are close to them. This extraction process separates the useful object data from the noisy background data, resolving the contradiction between using cheap hardware and maintaining data quality.
Solution Approach 2:
The patent converts the harmful noise (extraneous points on planar surfaces) into a beneficial filtering mechanism. By identifying that noise often lies on planar surfaces like the ground, the system uses this characteristic to its advantage, removing these surfaces to eliminate noise while preserving the actual objects of interest.
2Reliability
If planar surface filtering is applied to remove noise, then noise reduction is improved, but relevant data points may be removed
Solution Approach 1:
The patent applies different processing quality to different regions of the point cloud. It identifies planar surfaces and applies filtering specifically to points associated with these surfaces, while leaving other regions untouched. This localized approach ensures that noise reduction is applied where needed without affecting relevant data points in other areas.
Solution Approach 2:
The patent uses multiple parameters (distance threshold, color distance threshold) to control the filtering process. By adjusting these parameters, the system can fine-tune the filtering to remove noise while preserving relevant data, balancing noise reduction with information retention.
3Measurement precision
If multiple filtering criteria (distance and color) are used to identify noise, then filtering accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments the filtering process into distinct stages: first identifying planar surfaces, then filtering points based on distance to these surfaces, and finally applying color-based filtering. This segmentation breaks down the complex multi-criteria filtering into manageable steps, improving accuracy while keeping processing complexity manageable through structured organization.
Data Source
AI summary
Method and system for extracting an object-of-interest from a 3D point cloud prior to performing 3D operations onto the extracted object-of-interest are disclosed, the 3D point cloud including a plurality of data points. The method includes, accessing the 3D point cloud, in response to identifying a planar surface within the 3D point cloud, identifying first data points that define the planar surface, identifying second data points for which a distance to the planar surface is below a first distance threshold, identifying third data points for which a color distance to the planar surface is below a second distance threshold and removing the first, second and third data points from the 3D point cloud to create pre-curated 3D point cloud clusters. For each of the pre-curated 3D point cloud clusters, a corresponding cluster parameter. The object-of-interest is identified based on the calculated cluster parameters and 3D operations are performed thereon.