Roof Vector Projection for 3D Structure Height Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to efficiently extract three-dimensional structure information from overhead imagery, particularly for large-scale geospatial applications, due to varying camera perspectives and lighting conditions.
Innovation Solution
A scalable approach that leverages two-dimensional vector maps derived from overhead imagery, combined with additional imagery from alternate viewpoints, using vector data projection and feature matching processes to determine the height of structures, employing machine learning techniques for feature matching, including supervised and self-supervised training methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional photogrammetry and stereoscopic techniques are used to extract three-dimensional structure from overhead imagery, then three-dimensional information can be obtained, but the process becomes computationally intensive and difficult to scale to large geographic areas
Solution Approach 1:
The patent segments the three-dimensional structure extraction process into two distinct phases: (1) generating two-dimensional vector maps from overhead imagery using conventional photogrammetry, and (2) projecting these vector maps onto three-dimensional surfaces using digital surface models. This segmentation allows the computationally intensive 3D processing to be avoided for large areas, while only requiring lightweight 2D vector map generation that can be easily scaled across large geographic regions.
Solution Approach 2:
The patent introduces two-dimensional vector maps as an intermediary representation between overhead imagery and three-dimensional structure extraction. These vector maps serve as a simplified intermediate form that captures essential geometric features (buildings, roads, water bodies) in 2D space, which can then be efficiently projected onto 3D surfaces using DSMs, avoiding the need for direct computationally intensive 3D processing of raw imagery across large areas.
2Measurement precision
If additional imagery from multiple viewpoints is collected to improve three-dimensional structure determination, then height information accuracy improves, but data collection complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by generating two-dimensional vector maps from overhead imagery before the three-dimensional structure extraction process. These pre-generated vector maps contain accurate geometric representations of structures in 2D space, which can then be directly projected onto 3D surfaces using DSMs. This preliminary 2D mapping step eliminates the need for complex multi-viewpoint imagery collection and processing, as all necessary geometric information is already captured in the vector maps.
3Measurement precision
If conventional structure-from-motion methods are applied to overhead imagery, then three-dimensional models can be generated, but the methods struggle with varying lighting conditions and camera perspectives
Solution Approach 1:
The patent extracts only the essential two-dimensional geometric information from overhead imagery into vector map representations, separating this 2D geometric content from the problematic elements (lighting conditions, camera perspectives, image quality variations). By taking out only the necessary geometric features and representing them as simplified vector maps with precise coordinates and shapes, the method eliminates sensitivity to lighting and perspective variations that plague conventional structure-from-motion approaches.
Data Source
AI summary
Methods and systems for determining the height of structures based on imagery of the structures and associated two-dimensional vector data are provided. An example method involves projecting two-dimensional vector data outlining a roof of a structure into images of the structure captured from different perspectives and feature matching the vector data across the imagery to determine a best-matching three-dimensional position for the roof situated at the height of the structure.