Adaptive Gravity and Scale Estimation via Pose Difference Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems that combine computer vision and inertial sensors for augmented reality and navigation struggle to accurately coordinate scale and gravity estimation, leading to potential drift and inaccuracies in pose determination.
Innovation Solution
A method and apparatus that compute an image-based pose and an inertia-based pose using a scaling factor and gravity vector estimation, and then form a difference between the two poses to refine the estimation of the gravity vector or scaling factor, enabling accurate and adaptive estimation of scale and gravity in mobile devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scale and gravity are estimated separately using computer vision and inertial sensors independently, then each estimation can be computed using dedicated algorithms, but coordination between the two estimations becomes difficult leading to drift and inaccuracies
Solution Approach 1:
The patent merges the separate scale estimation from computer vision with gravity estimation from inertial sensors into a unified framework. The sensor fusion module combines image-based pose data with accelerometer data to jointly estimate both scale and gravity, allowing them to coordinate their estimations and eliminate drift through mutual constraint rather than operating independently
Solution Approach 2:
The system implements feedback by using the computed difference between image-based pose and inertia-based pose to refine both scale and gravity estimates. The coordinate estimation module uses this difference feedback to iteratively improve the accuracy of both estimations, ensuring they remain synchronized and reducing cumulative drift over time
2Measurement precision
If accelerometer measurements are adjusted by gravity estimation to determine linear acceleration, then movement detection accuracy improves, but errors in gravity estimation cause cumulative drift in position determination
Solution Approach 1:
The system uses feedback from image-based pose measurements to continuously correct gravity estimates. By comparing the inertia-based pose (which depends on gravity-adjusted acceleration) with the image-based pose over time, the system detects drift and refines the gravity estimate, extending the valid duration of the estimation
Solution Approach 2:
The patent maintains continuous refinement of gravity and scale estimates by constantly comparing inertial and visual data streams. This continuous feedback loop ensures that the gravity estimation remains accurate over extended periods, preventing cumulative drift from propagating through the navigation solution
3Measurement precision
If computer vision algorithms are used to determine scale estimation, then scale can be obtained from image data, but the estimation lacks metric accuracy without additional constraints
Solution Approach 1:
The patent combines computer vision scale estimation with inertial sensor data in a unified coordinate estimation framework. The image-based pose provides scale information while the accelerometer provides gravity-direction information, and their integration through the sensor fusion module creates mutually constraining estimates that achieve metric accuracy without requiring complex external constraints
Solution Approach 2:
The system uses the difference between image-based pose and inertia-based pose as feedback to refine scale estimates. This feedback mechanism allows the scale estimation to be continuously improved by the constraints provided by the inertial measurements, achieving accurate metric scale without adding complex external reference systems
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, apparatus and methods for estimating gravity and/or scale in a mobile device are presented. A difference between an image-based pose and an inertia-based pose is using to update the estimations of gravity and/or scale. The image-based pose is computed from two poses and is scaled with the estimation of scale prior to the difference. The inertia-based pose is computed from accelerometer measurements, which are adjusted by the estimation for gravity.