Fluorescent Marker Training Data for Flexible Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods face difficulties in estimating the three-dimensional shape of flexible objects like clothing based on image information, which hinders effective robotic recognition and manipulation.
Innovation Solution
A method involving a robotic controller that uses fluorescent markers invisible under visible light, capturing images under different illumination conditions to generate training data, and employing machine learning to estimate picking positions without the need for the markers during operation, allowing robots to recognize and manipulate flexible objects accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fluorescent markers are used to enable robotic recognition of flexible objects, then recognition accuracy is improved, but the complexity of the system increases due to requiring multiple illumination conditions and marker management
Solution Approach 1:
The patent applies preliminary action by pre-attaching fluorescent markers to the flexible object before robotic manipulation. These markers are trained in advance under specific illumination conditions (UV light) to create distinctive visual features. During actual robotic operation, the pre-trained markers enable accurate recognition and 3D shape estimation without requiring real-time complex processing, thus improving recognition accuracy while managing system complexity through advance preparation.
Solution Approach 2:
The fluorescent markers serve as an intermediary between the flexible object and the robotic vision system. The markers absorb UV light and emit visible fluorescence, creating a mediating visual signal that the robot can easily detect and process. This intermediary mechanism translates the complex 3D shape information of flexible objects into simplified fluorescent patterns that are easier for the robotic system to recognize and interpret, thereby improving measurement precision.
2Measurement precision
If multiple illumination conditions are used to capture training data, then model accuracy is improved, but the time required for data acquisition increases
Solution Approach 1:
The patent employs periodic action by capturing images under different illumination conditions (visible light and UV light) in alternating sequences during the training data acquisition process. The imaging device switches between illumination modes periodically, capturing both the fluorescent marker patterns under UV light and the object appearance under visible light. This periodic switching enables comprehensive training data collection with multiple illumination conditions while maintaining efficient data acquisition throughput, balancing model accuracy with time consumption.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate robotic recognition and manipulation of flexible objects by generating a trained model that can infer picking positions from visible light images alone, improving robotic accuracy and versatility in handling complex shapes.
Implementation Method 1
a marker unit recognizable under a first illumination condition is provided
Data Source
AI summary
One aspect of the present disclosure relates to a generation method for a training dataset, comprising: capturing, by one or more processors, a target object to which a marker unit recognizable under a first illumination condition is provided; and acquiring, by the one or more processors, a first image where the marker unit is recognizable and a second image obtained by capturing the target object under a second illumination condition.