Hand-Eye Vision Calibration for Accurate Robot Pose Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot systems face challenges in accurately calibrating the coordinate system for precise positioning and posture estimation of objects using vision sensors, particularly in hand-eye robot systems where the vision sensor is fixed on the robot's hand-tip part.
Innovation Solution
An arithmetic apparatus that includes a control part for outputting control signals to position an imaging part relative to a learning target object, generating a model through learning using imaging results, and performing arithmetic processing to calculate the position and posture of a processing target object based on learned parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If coordinate system calibration is performed using traditional vision sensor methods, then the robot system can perform basic positioning tasks, but the positioning precision and posture estimation accuracy are insufficient for high-precision operations
Solution Approach 1:
The system performs preliminary coordinate system calibration and creates a parameter determination model before actual robot operations. By pre-calibrating the vision sensor coordinate system relative to the robot coordinate system and establishing mathematical models for position and posture calculation, the system ensures high measurement precision from the outset rather than attempting to achieve it during operation.
Solution Approach 2:
The patent replaces traditional mechanical positioning and measurement methods with vision-based measurement. The vision sensor captures images of calibration patterns and objects, and mathematical models convert these optical measurements into precise position and posture data, eliminating the need for complex mechanical measurement systems and improving both precision and reliability.
2Adaptability or versatility
If the vision sensor is fixed on the robot hand-tip for hand-eye calibration, then the system can perform object manipulation tasks, but the coordinate system calibration becomes complex and accuracy is compromised
Solution Approach 1:
The patent introduces a calibration pattern as an intermediary object between the vision sensor and the robot coordinate system. This calibration pattern serves as a reference medium that facilitates the establishment of mathematical relationships between different coordinate systems, simplifying the calibration process while maintaining high accuracy for hand-eye operations.
Solution Approach 2:
The system determines optimal parameters for position and posture calculation through the learned model, adjusting calculation parameters based on the specific geometric relationships and measurement conditions. This parameter optimization simplifies the calibration procedure while ensuring accurate coordinate transformation for versatile robot operations.
3Measurement precision
If traditional arithmetic processing is used without learned models, then the system operates with simpler processing, but the accuracy of position and posture calculation is insufficient
Solution Approach 1:
The system performs preliminary learning to determine optimal arithmetic processing parameters before actual position and posture calculation. By pre-processing calibration data and establishing mathematical models in advance, the system creates optimized calculation frameworks that deliver high measurement precision during operation without requiring complex real-time processing.
Solution Approach 2:
The patent creates mathematical models that copy and represent the geometric relationships observed during calibration. These models serve as simplified representations of complex spatial relationships, enabling accurate position and posture estimation through straightforward arithmetic operations rather than complex real-time calculations.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
An arithmetic apparatus includes: a control part that outputs a control signal for controlling an imaging part and a robot equipped with the imaging part; and a learning part that generates a model for determining a parameter of arithmetic processing in the control part, by learning using an imaging result of a learning target object by the imaging part. The control part outputs a first control signal. The learning part generates the model, by learning using learning image data generated by the imaging part imaging the learning target object in the predetermined positional relation, by control based on the first control signal. The control part performs the arithmetic processing, by using the parameter determined by the model and processing target image data generated by the imaging part imaging a processing target object, and calculates at least one of a position and posture of the processing target object.