Position and Orientation Measurement Using Coarse Image Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring the position and orientation of objects with complex shapes and textures in robotic assembly tasks are not robust when the assumed range of camera positions and orientations is not satisfied, leading to inaccurate measurements.
Innovation Solution
A system that acquires coarse positions and orientations from images, generates candidate positions and orientations using a Gaussian distribution, and performs fitting using three-dimensional model information to derive the precise position and orientation of objects, enhancing robustness through iterative correction and edge detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a predetermined sampling interval is used to generate initial positions and orientations, then the measurement process is simplified, but the robustness deteriorates when the assumed range is not satisfied
Solution Approach 1:
The system performs preliminary action by acquiring coarse position and orientation information before the detailed fitting process. This preliminary coarse information serves as a foundation for generating candidate initial values, making the subsequent measurement process more robust without requiring complex predetermined sampling intervals.
Solution Approach 2:
The system dynamically adjusts the range of candidate initial values based on the acquired coarse position and orientation. Instead of using a fixed predetermined sampling interval, the candidate values are generated within a range centered on the coarse information, allowing the system to adapt to different measurement scenarios and maintain robustness.
2Device complexity
If a fixed range of camera positions and orientations is assumed, then the calculation is simplified, but the adaptability deteriorates when the actual position falls outside the assumed range
Solution Approach 1:
The system makes the measurement range dynamic by centering candidate initial values on the acquired coarse position and orientation. This dynamic approach allows the measurement range to adapt to the actual object position while keeping the calculation process simple, eliminating the need for fixed predetermined ranges.
Solution Approach 2:
The system changes the parameter range dynamically based on the coarse position and orientation information. By adjusting the candidate initial values within a range centered on the measured coarse information, the system adapts to different object positions and orientations without requiring complex fixed-range assumptions.
3Reliability
If multiple candidate positions and orientations are generated based on coarse information, then the measurement robustness is improved, but the computational load increases
Solution Approach 1:
The system applies partial action by generating multiple candidate initial values within a focused range around the coarse position and orientation, rather than exhaustively searching the entire possible space. This provides sufficient robustness through multiple candidates while avoiding the excessive computational load of a complete search.
Solution Approach 2:
The system concentrates computational resources locally around the coarse position and orientation by generating candidate values within a centered range. This local quality approach improves measurement robustness through multiple candidates in the relevant region while minimizing overall computational load by not searching irrelevant areas.
Data Source
AI summary
There is provided a position and orientation measurement apparatus, information processing apparatus, and an information processing method, capable of performing robust measurement of a position and orientation. In order to achieve the apparatuses and method, at least one coarse position and orientation of a target object is acquired from an image including the target object, at least one candidate position and orientation is newly generated as an initial value used for deriving a position and orientation of the target object based on the acquired coarse position and orientation, and the position and orientation of the target object in the image is derived by using model information of the target object and by performing at least once of fitting processing of the candidate position and orientation generated as the initial value with the target object in the image.