Back to results
US 12,731,275 B2Granted
Image processing device, component gripping system, image processing method and component gripping method
AssigneeYamaha Hatsudoki Kabushiki Kaisha
InventorAtsushi Yamamoto
Life of this patent
Estimating the term…
Abstract
A composite image is generated by combining a gray scale image, referred to as a luminance image, and a depth image respectively representing the plurality of components. In the generated composite image, the shape of the component at a relatively high position among the plurality of components easily remains, and the composite image is useful in confirming such a component.
Claims19
An image processing device, comprising:
- an image acquirer configured to acquire a luminance image representing a plurality of components stored in a container and a depth image representing the plurality of components;
- and
- an image compositor configured to generate a composite image by combining the luminance image and the depth image acquired by the image acquirer,
- wherein:
- the luminance image represents a luminance Vg for each pixel,
- the depth image represents a depth Vd for each pixel,
- the composite image represents a composite value Vc for each pixel, and
- the image compositor is configured to generate the composite image by calculating a composite value Vc obtained by weighting the luminance Vg by the depth Vd.
5 more elements
An image processing method, comprising:
- acquiring a luminance image representing a plurality of components stored in a container and a depth image representing the plurality of components;
- and
- generating a composite image by combining the luminance image and the depth image,
- wherein:
- the luminance image represents a luminance Vg for each pixel,
- the depth image represents a depth Vd for each pixel,
- the composite image represents a composite value Vc for each pixel, and
- generating the composite image includes generating the composite image by calculating a composite value Vc obtained by weighting the luminance Vg by the depth Vd.
5 more elements
A component gripping method, comprising:
- acquiring a luminance image representing a plurality of components stored in a container and a depth image representing the plurality of components;
- generating a composite image by combining the luminance image and the depth image;
- generating a plurality of target component images respectively representing different components each other by setting a hand range representing a range of a robot hand for the component and generating the target component image representing the component included in the hand range based on the composite image and calculating a grip success probability when the robot hand tries to grip the component in a positional relationship of the robot hand and the component represented by the target component image for each of the plurality of target component images;
- selecting one target component image, out of the plurality of target component images, based on the grip success probabilities respectively calculated for the plurality of target component images and determining to grip the component represented by the one target component image by the robot hand in a posture represented by the one target component image;
- and
- gripping the component by the robot hand based on a determination to select one target component image out of the plurality of target component images and grip the component represented by the one target component image by the robot hand in a posture represented by the one target component image.
3 more elements
An image processing device, comprising:
- an image acquirer configured to acquire a luminance image representing a plurality of components stored in a container and a depth image representing the plurality of components;
- an image compositor configured to generate a composite image by combining the luminance image and the depth image acquired by the image acquirer;
- and
- a success probability calculator configured to
- generate a plurality of target component images representing different components that are different from each other by setting a hand range representing a range of a robot hand for the component and generating the target component image representing the component included in the hand range based on the composite image, and
- calculate a grip success probability when the robot hand tries to grip the component in a positional relationship of the robot hand and the component represented by the target component image for each of the plurality of target component images,
- wherein:
- the success probability calculator is configured to select one target component image, out of the plurality of target component images, based on the grip success probabilities respectively calculated for the plurality of target component images, and determine to grip the component represented by the one target component image by the robot hand in a posture represented by the one target component image,
- the success probability calculator is configured to select the target component image having a highest grip success probability, out of the target component images not satisfying a predetermined exclusion condition out of the plurality of target component images, as the one target component image, and
- the success probability calculator includes a neural network having learned a relationship of the target component image and the grip success probability using a success/failure of grip when the robot hand tries to grip the component in a positional relationship of the robot hand and the component represented by the target component image as training data, and the grip success probability is calculated for each of the plurality of target component images by the neural network.
7 more elements
Dependencies are inferred from the claim text.