Pose Estimation for Planar Models Using Coarse-Fine Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pose estimation algorithms face challenges with pose redundancy when dealing with planar models, as the number of estimated poses increases exponentially, limiting their effectiveness in simultaneous pose and correspondences determination.
Innovation Solution
A method involving coarse pose estimation and the iterative extended TsPose algorithm to obtain two pose candidates, with convergence criteria based on uncertainty parameters and objective function values, selecting the final pose based on smaller objective function values, and utilizing the extended TsPose algorithm for accurate pose and correspondences estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If iterative algorithms are used for simultaneous pose and correspondences determination with a planar model, then the algorithm can handle unknown correspondences, but the number of estimated poses increases exponentially causing pose redundancy
Solution Approach 1:
The patent applies preliminary action by performing coarse pose estimation before fine pose estimation. The coarse pose estimation algorithm generates initial pose estimates that serve as starting points for subsequent refinement, preventing the exponential growth of pose candidates by establishing a limited set of initial hypotheses early in the process.
Solution Approach 2:
The patent segments the pose estimation process into two distinct stages: coarse pose estimation and fine pose estimation. The coarse stage generates initial pose candidates while the fine stage refines these candidates through iterative optimization, allowing the system to manage computational complexity by handling different aspects of pose determination in separate phases.
2Device complexity
If coarse pose estimation is performed to restrict search space, then pose redundancy is reduced, but the initial pose estimates may lack precision
Solution Approach 1:
The patent applies dynamics by transitioning from static coarse pose estimates to dynamic refined pose estimates through iterative optimization. The fine pose estimation algorithm dynamically adjusts pose parameters based on image data and model constraints, allowing the system to achieve high precision while maintaining a restricted search space defined by the coarse estimates.
Solution Approach 2:
The coarse pose estimation serves as preliminary action that establishes a restricted search space. This preliminary step provides initial guesses that guide the subsequent fine pose estimation, ensuring that the iterative refinement process starts from reasonable hypotheses and converges to accurate solutions without exploring the entire pose space.
3Reliability
If multiple pose candidates are generated, then the chance of finding the correct pose increases, but the computational cost increases
Solution Approach 1:
The patent applies partial action by generating a limited number of coarse pose candidates that are sufficient to guarantee finding the correct pose, rather than exhaustively searching all possible poses. The coarse estimation algorithm produces enough candidates to ensure the correct pose is included, while the subsequent fine estimation refines only these selected candidates, avoiding unnecessary computational expenditure.
Solution Approach 2:
The patent segments the candidate generation and refinement processes into separate stages. The coarse pose estimation generates a manageable number of initial candidates, and the fine pose estimation refines only these candidates through iterative optimization. This segmentation allows the system to maintain high reliability by considering multiple candidates while controlling computational cost by limiting the refinement process to a small subset.
Data Source
AI summary
A method and software for the simultaneous pose and points-correspondences determination from a planar model are disclosed. The method includes using a coarse pose estimation algorithm to obtain two possible coarse poses, and using each one of the two coarse poses as the initialization of the extended TsPose algorithm to obtain two candidate estimated poses; selecting one from the two candidate estimated poses based on the cost function value. Thus, The method solves the problem of pose redundancy in the simultaneous pose and points-correspondences determination from a planar model, i.e., the problem that the numbers of estimated poses increase exponentially as the iterations go. The disclosed embodiment is based on the coplanar points, and does not place restriction on the shape of a planar model. It performs well in a cluttered and occluded environment, and is noise-resilient in the presence of different levels of noise.