Shape and pose estimation for object placement
AssigneeAmazon Technologies, Inc.
InventorsAzarakhsh Keipour, Dhruv Kool Rajamani, Sicong Zhao, Tyler W Garaas, Avnish Sachar
Life of this patent
Estimating the term…
Abstract
Systems and methods are described relating to pose and shape estimation of objects. In some examples, a camera generates images depicting an object from different viewpoints. Within the images, a system identifies corners of the object, and the system uses these corners to generate lines that are projected from different viewpoints through the corners. Points may be identified by intersecting lines such that the points may be used to generate estimations of the object. The estimation with the highest score may be used to place the object in a location different from where the object was located.
Claims20
A computer-implemented method, comprising:
- obtaining a set of images that is captured from different viewpoints, wherein the set of images depicts an object;
- identifying a plurality of corners of the object in the set of images as a result of performing image segmentation on the set of images;
- generating a plurality of lines in 3-dimensional (3D) space that project from the different viewpoints through the plurality of corners of the object in the set of images;
- identifying a first set of intersection points using the plurality of lines in 3D space;
- selecting a second set of intersection points from the first set of intersection points in response to determining that the second set of intersection points is within a region of interest (ROI);
- generating a plurality of estimations of the object using the second set of intersection points, wherein the plurality of estimations indicates shape and pose of the object;
- selecting, from the plurality of estimations, an estimation that corresponds to a highest score;
- and
- identifying a location in a container to place the object in using the selected estimation of the object.
A system, comprising:
- one or more processors;
- memory that stores computer-executable instructions that, if executed, cause the one or more processors to:
- receive a plurality of images that is captured from a plurality of viewpoints, where the plurality of images includes an object;
- identify a plurality of corners of the object based, at least in part, on performing image segmentation of the plurality of images;
- determine a plurality of 3-dimensional (3D) lines that project from at least the plurality of corners through center of the plurality of viewpoints;
- identify a plurality of points based, at least in part, on intersecting at least two of the plurality of 3D lines;
- generate a plurality of representations of the object based, at least in part, on the plurality of points;
- select a representation from the plurality of representations based, at least in part, on a threshold;
- and
- cause one or more robots to move the object to a location in a container based, at least in part, on the selected representation.
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:
- identify an object that is located in a first location;
- receive two or more images depicting the object from different viewpoints;
- identify sets of boundaries of the object in the two or more images;
- identify two or more corners of the object using the sets of boundaries;
- generate a set of lines that are projected from the different viewpoints through the two or more corners of the object;
- identify two or more intersections based, at least in part, on the set of lines;
- identify two or more estimations of the object using the two or more intersections of at least two of the set of lines;
- select a highest score from the two or more estimations;
- and
- place the object in a second location based, at least in part, on an estimation associated with the highest score.
Dependencies are inferred from the claim text.