Satellite Image Geolocation Correction Using Mobile Mapping Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Satellite imagery is affected by deviations between actual and reported satellite ephemeris and sensor pointing direction, leading to inaccuracies in geolocation of image coordinates.
Innovation Solution
A method involving a sensor model function with input parameters, including rational polynomial coefficients, azimuth-elevation functions, and satellite ephemeris, is used to map ground coordinates to satellite images, defining trajectories and applying weighting functions to improve geolocation accuracy by adjusting pitch and roll parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If satellite ephemeris and sensor pointing data are used directly from metadata, then processing is simple and fast, but geolocation accuracy deteriorates due to deviations between actual and reported values
Solution Approach 1:
The system uses mobile mapping data as a reference standard to measure the deviation between actual and reported satellite pointing positions. This feedback loop enables continuous correction of sensor model parameters (pitch and roll) to compensate for ephemeris and pointing deviations, thereby improving geolocation accuracy without requiring complete system redesign
Solution Approach 2:
The invention adjusts specific parameters in the sensor model function (particularly pitch and roll angles) based on the measured deviations from mobile mapping trajectories. By changing these parameters iteratively to minimize the difference between satellite-derived and mobile mapping coordinates, the system achieves sub-meter geolocation accuracy while maintaining a relatively simple processing framework
2Measurement precision
If sensor model parameters are adjusted to improve geolocation, then accuracy improves, but computational time and processing complexity increase
Solution Approach 1:
Rather than adjusting all sensor model parameters or processing entire satellite images, the system focuses on correcting only the pitch and roll parameters using selected trajectory points from mobile mapping data. This partial action approach achieves the necessary geolocation correction without the computational burden of comprehensive parameter optimization across the entire image dataset
Solution Approach 2:
The system performs preliminary alignment using the satellite metadata before applying corrections based on mobile mapping data. By establishing an initial geometric model and then iteratively refining only the necessary parameters (pitch and roll) against known accurate trajectory points, the processing time is reduced while still achieving high geolocation accuracy
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Satellite images have inherent geo-positional errors of orders a few meters. Corrections are achieved by adjusting a sensor model which maps ground coordinates of control features into image coordinates and establishing a correspondence between the ground and image features, in this case a road network. The ground coordinates are obtained from mobile pose points. To adjust the sensor model we rely on the fact that the roads are typically much more uniform than surrounding features, and therefore have smaller entropy. The sensor model is adjusted so that the image pixels, obtained from projecting ground coordinates of the mobile pose points onto the image, minimize the entropy of the pixels that represent the road network.