3D Building Reconstruction Using Missing Point Cloud Completion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional reconstruction methods face challenges with disorderliness, sparsity, and limited information in point cloud data, particularly in scenarios like autonomous driving, leading to incomplete and biased models due to missing data.
Innovation Solution
A method involving panoramic data acquisition, multi-view image correction, dense matching, and a neural network-based approach to assess and complete missing point cloud data, using encoding-decoding networks and the ShapeNet dataset for accurate reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional point cloud collection methods are used for three-dimensional reconstruction, then the reconstruction process is simpler, but the point cloud data is sparse and contains missing elements
Solution Approach 1:
The patent applies preliminary action by performing point cloud completion before the actual three-dimensional reconstruction process. The system identifies and fills missing point cloud data elements in advance using a completion model, ensuring that the reconstruction process receives complete point cloud data. This preliminary completion step prevents information loss that would otherwise occur during reconstruction.
Solution Approach 2:
The patent introduces an intermediary component - the point cloud completion model - that acts as a mediator between the raw point cloud data and the reconstruction process. This intermediary analyzes the point cloud data, identifies missing elements, and generates completion results that are then integrated into the original data. This intermediary layer resolves the contradiction by enhancing data completeness without requiring changes to the fundamental reconstruction process.
2Measurement precision
If point cloud data is collected with higher density to reduce sparsity, then the geometric information is more complete, but the data processing complexity increases
Solution Approach 1:
The patent replaces the mechanical approach of collecting denser point cloud data with a computational approach. Instead of physically acquiring more data points through additional sensors or scans, the system uses a completion model that computationally infers and generates missing geometric information. This substitution maintains measurement precision while avoiding the increased processing complexity that would result from collecting and handling larger volumes of raw data.
3Productivity
If the point cloud data is processed without completion to maintain speed, then the processing time is shorter, but the three-dimensional model accuracy is reduced
Solution Approach 1:
The patent applies partial action by implementing point cloud completion only for the missing or sparse portions of the data, rather than processing the entire point cloud uniformly. The completion model selectively identifies and fills only the necessary missing elements, maintaining reconstruction speed while improving accuracy. This partial completion approach avoids the time penalty that would result from comprehensive reprocessing of all data points.
Data Source
AI summary
The invention provides a method, apparatus, and storage medium for reconstructing three-dimensional models of buildings based on missing point cloud data. The method includes integrating image-based point cloud generation, neural network techniques, and skeleton line extraction methods, offering a novel approach to handling missing point cloud data. The generation of point cloud data is achieved using principles of Structure from Motion based on video or panoramic image data. The point cloud is sampled and segmented using a region growing algorithm. A neural network based on PointNet is constructed, utilizing cross-entropy loss functions to assess the missing points in the point cloud. For mapping high-confidence point clouds from sampled points, Truth Points is employed to complete the entire process of real-world three-dimensional reconstruction. The integration of images into the three-dimensional scene is achieved with strict geometric relationships.