Building Facade Point Cloud Denoising for Accurate Structure Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building facade extraction methods using three-dimensional laser scanning face inaccuracies due to noise points in point cloud data, leading to flawed structure construction.
Innovation Solution
A method involving supervoxel clustering based on voxel similarity and normal vector differences to identify and remove noise points, followed by plane fitting and Constrained Planar Cuts algorithm to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If point cloud data is acquired using three-dimensional laser scanning technology, then building facade structure can be extracted, but noise points in the data lead to inaccuracies in structure construction
Solution Approach 1:
The patent divides the point cloud data into multiple voxels (volumetric pixels) and performs supervoxel clustering to segment the data into meaningful regions. This segmentation allows noise points to be identified and removed while preserving the underlying facade structure, directly addressing the accuracy issue caused by noise contamination
Solution Approach 2:
The patent extracts and removes noise points from the point cloud data through supervoxel clustering and normal vector analysis. By identifying points that deviate from the dominant surface orientation and extracting them as noise, the method cleans the data while maintaining the essential facade structure information
2Measurement precision
If supervoxel clustering is performed based on multiple similarity factors, then noise filtering accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary supervoxel clustering based on positional similarity before conducting the more computationally intensive multi-factor similarity analysis. This preliminary segmentation reduces the search space and allows subsequent normal vector and color similarity calculations to be performed more efficiently on already-grouped voxels
Solution Approach 2:
The patent employs multiple similarity parameters (positional, color, and normal vector similarity) to comprehensively evaluate voxel relationships. By changing from single-parameter to multi-parameter evaluation, the method achieves more accurate noise identification while the weighted combination approach balances computational load
Data Source
AI summary
Disclosed is a method, device, electronic equipment, and medium for extracting building facade structure, the method comprises acquiring first point cloud data of a building facade and establishing a three-dimensional space based on the first point cloud data; dividing the three-dimensional space into multiple voxels and performing supervoxel clustering on the multiple voxels to obtain at least one first supervoxel; determining the difference between the normal vectors of adjacent points within the first supervoxel for each first supervoxel; identifying points with differences greater than a preset difference threshold as noise points and removing them to obtain second point cloud data; and determining the facade structure of the building. By performing clustering and then screening out noise points in the point cloud based on the normal vectors of the points within the first supervoxel obtained from the clustering, this disclosure can improve the validity of the point cloud data.