Laser Scan Registration via Rectified Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current laser scanning technologies face challenges in accurately registering multiple scans taken from different locations to generate a comprehensive three-dimensional model of a setting, as they struggle to determine precise transformations between scan data sets, leading to inconsistencies and incomplete reconstructions.
Innovation Solution
A method is introduced to determine registration between Scanworlds by generating rectified images based on viewpoints derived from 3D geometry data, using transformation tuples to align point and image data from different scans, facilitating the creation of a unified 3D model by selecting salient directions and employing algorithms like RANSAC for accurate transformation estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multiple laser scans are taken from different locations to obtain complete 3D model coverage, then the completeness of the 3D model improves, but the difficulty of accurately registering the scans increases
Solution Approach 1:
The patent introduces rectified images as an intermediary representation between raw point cloud data and registration algorithms. These rectified images are generated from point clouds using camera intrinsic parameters and extrinsic parameters, providing a standardized 2D projection that facilitates more reliable feature matching and registration between multiple scans taken from different locations.
2Ease of operation
If traditional point cloud registration methods are used, then the process is straightforward, but the registration precision and reliability deteriorate
Solution Approach 1:
The patent replaces direct point cloud-to-point cloud registration with an image-based registration approach. By projecting 3D point clouds into 2D rectified images using camera parameters, the system leverages成熟的image processing and feature matching algorithms to achieve more precise and reliable registration results while maintaining operational simplicity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of determining a registration between Scanworlds may include determining a first viewpoint of a setting (100) based on first point data of a first Scanworld. The first Scanworld may include information about the setting (100) as taken by a first laser scanner (102a) at a first location (104a). The method may further include determining a second viewpoint of the setting (100) based on second point data of a second Scanworld. The second Scanworld may include information about the setting (100) as taken by a second laser scanner (102b) at a second location (104b). The method may further include generating a first rectified image based on the first viewpoint and generating a second rectified image based on the second viewpoint. Additionally, the method may include determining a registration between the first Scanworld and the second Scanworld based on the first viewpoint and the second viewpoint.