Rolling Sequential Bundle Adjustment for GNSS Position Smoothing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating camera motion from video frames are computationally expensive and not suitable for real-time implementation, while existing solutions like Kalman filters fail to effectively connect current measurements to previous position and orientation states, leading to drift issues.
Innovation Solution
The rolling sequential bundle adjustment algorithm efficiently estimates camera motion by using previous frames and sequentially adding new measurements, removing the oldest frame, allowing for real-time determination of camera coordinates and orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bundle adjustment is used to estimate camera motion, then measurement precision is improved, but computational time increases significantly
Solution Approach 1:
The patent segments the bundle adjustment process into a rolling sequential framework where only a subset of recent frames is processed at a time. The most recent frame is added to the adjustment while the oldest frame is removed, creating a sliding window approach that maintains computational efficiency while preserving measurement precision for current camera motion estimation.
Solution Approach 2:
The patent implements a dynamic rolling sequential bundle adjustment that adapts the set of frames used for computation. Instead of using all available frames statically, the system dynamically updates the frame set by adding the newest frame and removing the oldest frame, allowing real-time computation with maintained accuracy.
2Loss of time
If Kalman filter is used to estimate camera motion, then computational time is reduced, but connection between current measurements and previous states is lost
Solution Approach 1:
The patent incorporates feedback mechanisms where the rolling sequential bundle adjustment uses previously adjusted camera positions and orientations to inform current estimates. The connection between current measurements and previous states is maintained through the sequential adjustment process, which propagates information across time steps unlike Kalman filters that may lose this temporal connection.
3Measurement precision
If all frames are used in bundle adjustment, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the complete frame sequence into a manageable rolling window of recent frames. This segmentation reduces the computational complexity by limiting the number of frames processed in each adjustment cycle while maintaining sufficient measurement precision through the inclusion of relevant recent measurements.
Solution Approach 2:
The patent discards the oldest frame from the adjustment set while recovering and incorporating the newest frame. This cyclic process maintains an optimal subset of frames for computation, balancing measurement precision with device complexity by continuously updating the frame set rather than processing all historical frames simultaneously.
Data Source
AI summary
A method of improving position determination of a device using locally measured movement. A first position fix of a Global Navigation Satellite System (GNSS) receiver system of a device is accessed. A second position fix of the GNSS receiver system is accessed at a time subsequent to the first position fix. Locally measured device movement information is obtained from at least one sensor, that is in a known physical relationship to the device, for a time period after the first position fix and no later than the second position fix, wherein the at least one sensor comprises an image capture device. The quality of measurement of the second position fix is improved by disciplining the second position fix based on the locally measured device movement information.