Using neural networks to examine objects
AssigneeAmazon Technologies, Inc.
InventorsWillem Hendrik Boshoff, Andrew Lee Allen, Joseph Benjamin Williams
Life of this patent
Estimating the term…
Abstract
Systems and methods are disclosed for examining objects (e.g., mobile storage units) using neural networks. Upon determining that the object is within an area of interest, the system uses multiple sensors positioned at various locations to capture the object from four or more sides at the same time. Using a neural network, the system identifies a first set of features of the object, which are then used to determine the location information of a second set of features, also identified by the neural network. The system evaluates whether this second set of features meets a series of criteria to determine if the object passes or fails the examination.
Claims20
A station, comprising:
- a plurality of cameras positioned at distinct locations and oriented to generate different perspectives of an object;
- one or more lighting elements to increase brightness of the object;
- one or more processors;
- and
- memory that stores computer-executable instructions that, if executed, cause the one or more processors to:
- determine that the object is positioned at a location within the station using at least one of the plurality of cameras;
- as a result of determining that the object is positioned at the station, use the plurality of cameras and one or more lighting elements to capture a set of images that include the object;
- identify, using one or more neural networks, a plurality of features of the object based, at least in part, on the set of images;
- identify, using one or more neural networks, a plurality of regions within the set of images that correspond to reference points to identify a plurality of coordinates of the plurality of features;
- determine whether the plurality of features of the object complied with a plurality of criteria based, at least in part, on the plurality of coordinates;
- and
- generate an indication on whether the object passed or failed an examination based, at least in part, on the determination.
A computer-implemented method, comprising:
- in response to determining that an object is located within an area of interest, causing a plurality of sensors located in different locations to generate a set of images that simultaneously capture the object from at least four sides to provide different viewpoints;
- identifying, using one or more neural networks, a first set of features of the object based, at least in part, on the set of images;
- identifying, using the one or more neural networks, a second set of features of the object usable to determine one or more locations of the first set of features;
- and
- determining whether the first set of features of the object complies with a set of criteria based, at least in part, on the one or more locations.
A non-transitory computer-readable storage medium storing thereon executable instructions that, as a result of being executed by one or more processors of a computer system, cause the computer system to at least:
- determine that an object is located within a station by at least using at least one of a plurality of sensors;
- cause the plurality of sensors placed at separate positions to concurrently generate a plurality of images capturing the object from four or more sides to obtain perspectives from multiple angles;
- identify, using a neural network, one or more coordinates corresponding to a first set of regions within the plurality of images based, at least in part, on a second set of regions within the plurality of images, the first set of regions and the second set of regions corresponding to different features of the object;
- and
- determine whether the object satisfies one or more criteria based, at least in part, on information generated using the one or more coordinates corresponding to the first set of regions.
Dependencies are inferred from the claim text.