Dynamic Model Switching for 3D Position Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for determining the position and orientation of a target object with a three-dimensional shape, such as Model-Based Vision, face challenges in balancing processing speed and accuracy, particularly when using CAD data or polygon models, as they require complex calculations and may not accurately represent the object's shape in three-dimensional space.
Innovation Solution
An information processing apparatus and method that acquires images of a target object, uses a combination of polygon and piecewise parametric curved surface models to estimate position and orientation by switching between models based on residual differences, allowing for high-speed and high-accuracy calculations by selecting appropriate model information for each three-dimensional point.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CAD data with parametric curved surfaces is used for high-accuracy position and orientation estimation, then measurement precision is improved, but device complexity and processing time increase due to complex geometrical calculations
Solution Approach 1:
The patent segments the model into two types: polygon models for initial coarse estimation and parametric curved surface models for final precise estimation. This segmentation allows the system to use simple calculations first, then apply complex calculations only where needed, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent implements a dynamic switching mechanism between polygon models and parametric curved surface models based on processing requirements. The system dynamically selects which model type to use at each stage of the estimation process, allowing it to adapt calculation complexity to the specific needs of each estimation step.
2Productivity
If polygon models are used for low calculation load and high processing speed, then productivity is improved, but measurement precision deteriorates due to polygonal approximation differences from the actual object shape
Solution Approach 1:
The patent divides the estimation process into two segments: a first estimation phase using polygon models for speed, and a second refinement phase using parametric curved surface models for accuracy. This segmentation allows the system to benefit from both model types without sacrificing overall processing efficiency.
Solution Approach 2:
The patent performs preliminary estimation using polygon models before conducting the final precise estimation with parametric models. This preliminary action using simpler models reduces the overall computational burden while maintaining the ability to achieve high accuracy in the final result.
3Measurement precision
If the number of polygons is increased to reduce approximation error, then measurement precision is improved, but device complexity and processing time increase due to the exponential increase in total polygon count
Solution Approach 1:
The patent segments the shape representation into two levels: polygon models for general shape approximation and parametric curved surface models for precise shape definition. This segmentation eliminates the need to increase polygon count exponentially, as the parametric models provide the necessary precision without the computational burden of excessive polygons.
Solution Approach 2:
The patent changes the representation parameters from a high-count polygon mesh to parametric equations defining curved surfaces. This parameter change allows the system to represent complex shapes with fewer elements while maintaining or improving accuracy, thus preserving processing speed.
Data Source
AI summary
There is provided with an information processing apparatus. An image including a target object is acquired. A coarse position and orientation of the target object is acquired. Information of a plurality of models which indicate a shape of the target object with different accuracy is held. A geometrical feature of the target object in the acquired image is associated with a geometrical feature indicated by at least one of the plurality of models placed at the coarse position and orientation. A position and orientation of the target object is estimated based on the result of association.