Platform Pose Estimation Using Synthetic Image Homography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation methods, such as satellite navigation, are unreliable in certain environments, leading to inaccurate pose determination due to spurious sensor data and errors in self-localization using observations of an aircraft's surroundings.
Innovation Solution
A method involving the generation of a synthetic image from a georeferenced reference representation, determination of a homography matrix between the synthetic and input images, and use of algorithms like RANSAC and neural networks to accurately determine the pose of a platform, independent of sensor data errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If self-localization using observation of surroundings is employed, then navigation capability is provided in GPS-denied environments, but accuracy deteriorates due to spurious sensor data effects
Solution Approach 1:
The patent generates a synthetic image from a georeferenced reference representation (such as a 3D map or aerial photograph) that serves as a virtual copy of the real-world scene. This synthetic image is then compared with the actual sensor image to determine pose. By using this synthetic copy instead of relying solely on direct sensor observations, the system avoids spurious effects in sensor data while maintaining the ability to navigate in GPS-denied environments.
Solution Approach 2:
The patent introduces a homography matrix as an intermediary computational element that relates points between the synthetic image and the sensor image. This homography matrix serves as a mediator that transforms coordinates from one reference frame to another, enabling accurate pose determination by mathematically bridging the synthetic and real images without directly comparing raw sensor data that may contain spurious effects.
2Measurement precision
If homography matrix determination is used to relate input image to reference representation, then pose determination accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the complex pose determination problem into distinct computational stages: first generating the synthetic image from reference data, then determining the homography matrix between synthetic and sensor images, and finally extracting pose information from the homography. This segmentation allows each step to be optimized independently and facilitates the use of efficient algorithms like RANSAC for robust homography estimation.
Solution Approach 2:
The patent performs preliminary generation of the synthetic image from the georeferenced reference representation before comparing it with the sensor image. This preliminary action prepares the synthetic image in advance with known geometric relationships, so that when the homography matrix is determined, the computational burden is reduced because the reference frame is already established and structured.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method of determining a pose of a platform, preferably an aircraft, the method comprising: receiving an input image of a portion of a surface, preferably the surface of the earth, the input image captured from a position of the platform; receiving a georeferenced reference representation of the portion; based on the reference representation, generating a synthetic image of the portion; determining a homography matrix between the synthetic image and the input image; and determining the pose of the platform based on the homography matrix.