Mosaic Image Generation via 3D Model Reprojection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for creating aerial imagery of geographic areas often result in inconsistent depictions, such as visible seams and artifacts, due to the stitching of overlapping aerial images, leading to unrealistic and inaccurate representations.
Innovation Solution
A computer-implemented method that generates a two-dimensional mosaic image by obtaining a three-dimensional model of the geographic area, reprojecting each pixel to the 3D model, and selecting source images from a plurality of aerial images based on camera ray consistency and pose alignment to minimize artifacts and ensure visual continuity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional stitching techniques are used to create mosaic images from overlapping aerial images, then the images can be assembled to cover the geographic area, but visible seams and artifacts appear resulting in inconsistent depictions
Solution Approach 1:
The patent transitions from two-dimensional image stitching to three-dimensional modeling by creating a textured 3D model of the geographic area. This dimensional change allows consistent depiction across the entire geographic area without seams or artifacts, as the 3D model provides a unified geometric framework that eliminates the discontinuities inherent in 2D stitching approaches.
Solution Approach 2:
The patent creates a virtual copy of the geographic area in three-dimensional space through photogrammetry and 3D reconstruction. This virtual 3D model serves as an accurate replica that can be viewed from any angle without the artifacts present in conventional 2D mosaic stitching, providing consistent and realistic depictions.
2Area of stationary object
If aerial images are stitched together to create mosaic images, then coverage of the geographic area is achieved, but buildings and terrain appear to lean or distort inconsistently
Solution Approach 1:
By constructing a three-dimensional model with accurate geometric representation of buildings and terrain, the patent preserves true shapes and orientations regardless of viewing angle. The 3D structure inherently maintains geometric precision that cannot be achieved through 2D stitching of aerial images.
Solution Approach 2:
The patent transforms the representation parameters from 2D image coordinates to 3D spatial coordinates, allowing accurate depiction of vertical and horizontal features. This parameter transformation enables buildings and terrain to maintain their true geometric properties across the entire covered area without distortion or leaning artifacts.
3Device complexity
If conventional 2D mosaic image creation is used, then the process is computationally simpler, but the visual representation lacks realism and accuracy
Solution Approach 1:
The patent justifies the increased computational complexity of 3D modeling by the significant improvement in realism and accuracy. The additional computational requirements are necessary to create the unified 3D geometric framework and texture mapping that produce artifact-free, consistent, and realistic depictions of the geographic area.
Data Source
AI summary
Systems and methods for determining two-dimensional (2D) images are presented. For instance, data indicative of a three-dimensional (3D) model of a geographic area can be obtained. A 2D output image can be generated depicting at least a portion of the geographic area based at least in part on the 3D model. Each pixel in the output image can then be reprojected to the 3D model. A plurality of aerial images depicting the geographic area can be obtained. A source image can then be determined for each pixel in the output image from the plurality of aerial images. The source image can be determined based at least in part on the reprojection of the pixel in the output image to the three-dimensional model.