Pose Estimation Using Hierarchical Tree Nodes and Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pose estimation methods for articulated objects, such as human bodies, face instability due to variations in silhouette changes caused by differences in clothing, physique, link length, and thickness, leading to incorrect recognition even when the three-dimensional shape model and observed object have the same pose.
Innovation Solution
A pose estimating device and method utilizing a pose dictionary with a tree structure of layered nodes, where similarity among lower nodes is higher than upper nodes, incorporating a two-class discriminative function and mapping function to transform image feature information into pose information, and employing past pose estimating information and a motion model for prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If direct silhouette matching method is used for pose estimation, then the method is simple to implement, but the pose estimation becomes unstable due to variations in link length, thickness, and surface form
Solution Approach 1:
The patent segments the pose estimation problem into multiple hierarchical levels (tree structure with layered nodes). Instead of direct silhouette matching, it divides the search space into coarse-to-fine levels, where upper layers perform rough pose estimation and lower layers refine the results. This segmentation allows the system to handle variations in link properties while maintaining estimation stability.
Solution Approach 2:
The patent introduces a hierarchical tree structure as an additional dimension to the pose estimation process. By organizing pose parameters in a multi-level tree structure where nodes represent different levels of pose detail, the system transforms the single-step matching problem into a multi-stage refinement process, improving robustness against silhouette variations.
2Adaptability or versatility
If silhouette-based pose estimation is used, then no marker is required, but recognition accuracy decreases due to differences in clothing and physique
Solution Approach 1:
The patent changes the parameters used for pose estimation from direct silhouette matching to a hierarchical tree-structured pose representation. By representing poses as nodes in a tree structure with layered abstraction, the system can capture essential pose information while being invariant to surface variations like clothing and physique differences.
Solution Approach 2:
The patent performs preliminary pose estimation at upper hierarchical levels before refining at lower levels. The tree structure allows preliminary coarse-grained pose estimation that is robust to surface variations, followed by progressive refinement. This preliminary action at multiple levels ensures accurate pose recognition despite clothing and physique differences.
3Reliability
If tree structure with layered nodes is used for pose estimation, then pose estimation stability improves, but device complexity increases
Solution Approach 1:
The patent segments the complex pose space into manageable hierarchical levels. Each node in the tree structure represents a subset of pose parameters at a specific level of detail. This segmentation reduces the complexity at each individual level while maintaining overall pose estimation stability through the hierarchical structure.
Solution Approach 2:
The patent uses the hierarchical tree structure as an organizational dimension to manage pose data complexity. By arranging pose parameters in a multi-level hierarchy rather than a flat structure, the system makes the data more manageable and computationally tractable while preserving pose estimation stability.
Data Source
AI summary
A pose estimating device includes: a pose dictionary; an image feature extracting unit configured to extract observed image feature information; a past information storing unit configured to store past pose estimating information of the articulated object; a pose predicting unit configured to predict a present pose; a node predicting unit configured to calculate a prior probability as to whether each nodes includes a present pose; an identifying unit configured to calculate a likelihood of the observed image feature information for each node; a node probability calculating unit configured to calculate a probability in which the present pose belongs to the node in the upper layer; and a pose estimating unit configured to calculate pose information.