3D Point Cloud Camera Pose Estimation Using Virtual Image Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for estimating the location and orientation of a camera when it shot an image are time-consuming.
Innovation Solution
An information processing apparatus and method that generates virtual camera images based on three-dimensional point cloud data to estimate the location and orientation of a camera by comparing similarities between virtual and actual camera images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the technique from Japanese Unexamined Patent Application Publication No. 2018-200504 is used to estimate camera location and orientation, then the estimation can be performed using three-dimensional point cloud data and edge coordinates, but the process is relatively time-consuming
Solution Approach 1:
The patent pre-generates virtual camera images from three-dimensional point cloud data before actual camera image analysis. By preparing these reference images in advance, the system eliminates the need for real-time complex calculations when estimating camera location and orientation, thus reducing processing time while maintaining estimation accuracy through subsequent similarity comparison.
Solution Approach 2:
The patent creates virtual camera images that are synthetic copies or representations of what the actual camera should capture from different positions and orientations. These copied virtual images serve as reference patterns for comparison with actual camera images, enabling faster estimation by matching against pre-computed references rather than performing full 3D reconstruction and optimization in real-time.
2Measurement precision
If virtual camera images are generated for each of a plurality of virtual cameras in three-dimensional space, then camera location and orientation estimation accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the three-dimensional space into multiple discrete virtual camera positions and orientations, generating separate virtual images for each segment. This segmentation allows the system to systematically cover the entire 3D space without requiring continuous computation, making the complex task of 3D estimation manageable by breaking it into discrete, pre-computable segments that can be stored and compared later.
Data Source
AI summary
An information processing apparatus according to the present disclosure includes: at least one memory configured to store instructions; and at least one processor configured to execute the instructions to: generate, for each of a plurality of virtual cameras in a three-dimensional space, a virtual camera image shot by the respective one of the virtual cameras on the basis of point cloud data of the three-dimensional space; and estimate, on the basis of similarity between the virtual camera image generated by having the processor execute the generation instruction and a camera image shot by a camera, location and orientation of the camera when it shot the camera image.