Adaptive Camera Pose Estimation for Moving Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vision aided Inertial Navigation Systems (VINS) perform poorly under certain conditions and degrade unpredictably in Augmented Reality applications, particularly when the target moves, leading to inaccurate pose determination between a camera and a target.
Innovation Solution
A mobile device adapts by reducing or eliminating the contribution of inertial sensor measurements when target movement is detected, switching to vision-only tracking to maintain accurate pose determination, using a combination of vision-based and inertial measurements to track the relative pose between a camera and a target.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If VINS approach is used to improve pose determination accuracy, then measurement precision is improved, but reliability deteriorates under certain conditions
Solution Approach 1:
The system dynamically adapts the tracking approach by detecting target motion and switching between VINS and vision-only methods. The processor monitors whether the target is stationary or moving and adjusts the contribution of inertial sensor measurements accordingly, making the system flexible rather than static.
Solution Approach 2:
The system changes the parameter of inertial sensor contribution based on target motion state. When the target is detected as moving, the contribution of inertial sensor measurements is reduced or eliminated; when stationary, full VINS processing is used. This parameter adaptation resolves the contradiction between precision and reliability.
2Productivity
If inertial sensor measurements are used to improve pose tracking, then productivity is improved, but measurement precision deteriorates when target moves
Solution Approach 1:
The system dynamically adjusts the tracking methodology based on real-time detection of target motion. The processor continuously monitors target state and switches between fast inertial-based tracking and accurate vision-based tracking, optimizing both speed and accuracy for different operational conditions.
Solution Approach 2:
The system uses an intermediary approach by combining both vision-based and inertial-based tracking methods. The processor acts as a mediator, selecting or combining methods based on target motion state, thus resolving the contradiction between the speed advantages of inertial tracking and the accuracy advantages of vision-based tracking.
3Reliability
If VINS approach is used to enhance robustness, then reliability is improved, but adaptability deteriorates in dynamic environments
Solution Approach 1:
The system transforms from a static VINS approach to a dynamic adaptive system that detects target motion and adjusts its processing mode accordingly. This dynamic adaptation enables the system to maintain robustness in stationary conditions while becoming versatile in dynamic environments.
Solution Approach 2:
The system achieves multi-functionality by implementing both VINS and vision-only tracking capabilities within a single processing framework. The processor can selectively apply either method or combine them, making the system universally applicable to both stationary and moving target scenarios.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A mobile device tracks a relative pose between a camera and a target using Vision aided Inertial Navigation System (VINS), that includes a contribution from inertial sensor measurements and a contribution from vision based measurements. When the mobile device detects movement of the target, the contribution from the inertial sensor measurements to track the relative pose between the camera and the target is reduced or eliminated. Movement of the target may be detected by comparing vision only measurements from captured images and inertia based measurements to determine if a discrepancy exists indicating that the target has moved. Additionally or alternatively, movement of the target may be detected using projections of feature vectors extracted from captured images.