Manufacturing part identification using computer vision and machine learning
AssigneeFord Motor Company
InventorsAlberto Rivera, Kiril Spasovski, Shounak Athavale, Randy Nunez
Life of this patent
Estimating the term…
Abstract
Systems, methods, and devices for object detection, including manufacturing part detection, are disclosed. A method for identifying a manufacturing part includes receiving an image comprising a manufacturing part from an augmented reality device and determining a bounding perimeter encapsulating the manufacturing part. The method includes receiving a prediction label and a confidence value for the manufacturing part from a neural network. The method includes generating a message comprising the image, the prediction label, and the confidence value, and providing the message to the augmented reality device.
Claims20
A method for identifying a manufacturing part, the method comprising:
- receiving an image comprising a manufacturing part from an augmented reality device;
- determining a bounding perimeter encapsulating the manufacturing part;
- receiving a prediction label and a confidence value for the manufacturing part from a neural network;
- generating a message comprising the image, the prediction label, and the confidence value;
- and
- providing the message to the augmented reality device.
A manufacturing part identification system comprising:
- an augmented reality device comprising a camera;
- a server in communication with the augmented reality device comprising computer readable storage media storing instructions that, when executed by one or more processors cause the one or more processors to:
- receive an image comprising a manufacturing part from the augmented reality device;
- determine a contour outline of the manufacturing part based on the image;
- and
- determine a bounding perimeter encapsulating the manufacturing part;
- and
- a neural network in communication with the server, wherein the neural network is configured to:
- determine a prediction label for the manufacturing part;
- and
- determine a confidence value comprising a statistical likelihood that the prediction label is correct.
Non-transitory computer readable storage media storing instructions that, when executed by one or more processors, cause the one or more processors to:
- receive an image comprising a manufacturing part from an augmented reality device;
- apply a color and magnitude of gradients threshold to the image such that a background of the image is mostly eliminated, and a contour of the manufacturing part remains;
- determine a contour outline of the manufacturing part;
- determine a bounding perimeter encapsulating the contour outline;
- receive a prediction label and a confidence value for the manufacturing part from a convolutional deep neural network;
- generate a message comprising the image, the prediction label, and the confidence value;
- and
- provide the message to the augmented reality device.
Dependencies are inferred from the claim text.