Oblique Aerial Image Geolocalization Using Candidate Pose Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aerial imagery captured from oblique angles often lacks complete pose data, making it challenging to geolocalize features accurately, and existing systems require complex and costly equipment to record this data.
Innovation Solution
An image geolocalizing pipeline that generates a geographic features file using machine learning models to select a best-matching candidate image from a limited search space, attributing candidate pose data to the actual image, even when partial or absent pose data is present.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex equipment is used to record complete pose data, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The system creates synthetic candidate images that copy the appearance characteristics of actual aerial images. These synthetic images are generated from 3D models and have associated pose data, serving as substitutes for actual images with missing pose information. The system matches actual images to their synthetic counterparts to infer pose data without requiring complex recording equipment.
Solution Approach 2:
Synthetic candidate images act as intermediaries between the actual aerial images and the required pose data. Instead of directly measuring pose data from the actual image (which would require complex equipment), the system uses synthetic images as a mediator that contains both the visual appearance and the corresponding pose information, enabling indirect acquisition of pose data.
2Manufacturing precision
If comprehensive pose data collection is implemented, then geolocalization accuracy is improved, but device complexity increases
Solution Approach 1:
The system generates synthetic candidate images that copy the visual characteristics of actual aerial images. These synthetic images are created from 3D terrain models and contain embedded pose data, serving as substitutes for actual images where pose data is missing or incomplete, thereby achieving accurate geolocalization without complex data collection systems.
Solution Approach 2:
The system performs preliminary generation of synthetic candidate images with known pose data before processing actual aerial images. By pre-preparing a library of synthetic images with accurate pose information, the system can later match actual images to these pre-generated candidates to infer pose data, avoiding the need for complex real-time measurement systems.
3Measurement precision
If pose data is made mandatory for image processing, then measurement precision is improved, but loss of information increases
Solution Approach 1:
The system creates synthetic candidate images that copy the appearance characteristics of actual aerial images. These synthetic images are generated from 3D models and include complete pose data, serving as substitutes for actual images that are missing pose information. This allows the system to maintain measurement precision by using the synthetic images with known pose data instead of discarding actual images with missing pose information.
Solution Approach 2:
Synthetic candidate images serve as an intermediary that bridges the gap between actual images with missing pose data and the requirement for complete pose information. The system uses these synthetic intermediaries to infer and restore pose data for actual images, preventing information loss while maintaining processing accuracy.
Data Source
AI summary
Methods, systems, and apparatus for receiving an image file recording an image and a set of metadata, determining a search space based on one or more of at least a portion of the set of metadata and auxiliary data, generating a set of candidate images based on the search space, identifying a candidate image in the set of candidate images as a best matching image relative to the image, the candidate image being associated with a set of candidate metadata, providing a set of augmented metadata for the image based on the set of metadata and the set of candidate metadata, the set of augmented metadata including at least a portion of the set of candidate metadata, and outputting a geographic features file that is generated using the set of augmented metadata, the geographic features file including data representing one or more geographic features represented in the image file.