Constructing contours from imagery
AssigneeGoogle Inc.
InventorIlya Dmitriyevich Zharkov
Life of this patent
Estimating the term…
Abstract
Systems and methods for identifying contours of objects depicted in imagery are provided. A contour of an occluded object can be reconstructed based on a source contour extracted from an image and/or other geographic data. The source contour can be analyzed to identify a main contour direction for one or more points on the source contour. A plurality of rays can be extended from each of the one or more points based on the main contour direction associated with the point. A graph model can be constructed from the plurality of rays extended from the plurality of points. A path can be determined through the graph model and the contour can be constructed based at least in part on the determined path.
Claims13
A computer-implemented method of identifying a contour for an object depicted in an image, the method comprising:
- determining, by one or more computing devices, a contour direction for a point on a source contour for the object;
- determining, by the one or more computing devices, a plurality of rays extending from the point on the source contour based at least in part on the contour direction;
- constructing, by the one or more computing devices, a graph model based at least in part on the plurality of rays;
- and
- constructing, by the one or more computing devices, a contour based at least in part on a path through the graph model;
- wherein the contour is constructed notwithstanding one or more occlusions of the object depicted in the image;
- wherein constructing, by the one or more computing devices, a graph model based at least in part on the plurality of rays comprises:
- constructing, by the one or more computing devices, a plurality of nodes of the graph model, each of the plurality of nodes representing one of the plurality of rays;
- and
- constructing, by the one or more computing devices, a plurality of edges interconnecting the nodes in the graph model, each of the plurality of edges representing an intersection between the plurality of rays;
- and
- wherein constructing, by the one or more computing devices, the graph model comprises determining, by the one or more computing devices, an edge weight for each of the plurality of edges in the graph model based at least in part on an area between the rays associated with the nodes interconnected by the edge and the source contour.
A computing system, comprising:
- one or more processors;
- and
- one or more memory devices storing computer-readable instructions that when executed by the one or more processors cause the one or more processors to perform operations for constructing a contour for an object depicted in an image, the operations comprising:
- identifying a source contour for the object from the image;
- determining a contour direction for a point on the source contour for the object;
- determining a plurality of rays extending from the point on the source contour based at least in part on the contour direction;
- constructing a graph model based at least in part on the plurality of rays, the graph model comprising a plurality of nodes, each node associated with one of the plurality of rays, the graph model further comprising at least one edge interconnecting the plurality of nodes, the at least one edge associated with an intersection between the plurality of rays;
- and
- constructing a contour based at least in part on a shortest path through the graph model;
- wherein the contour is reconstructed notwithstanding one or more occlusions of the object depicted in the image;
- wherein the operation of constructing the graph model comprises determining an edge weight for each of the plurality of edges in the graph model based at least in part on an area between the rays associated with the nodes interconnected by the edge and the source contour.
One or more tangible, non-transitory computer-readable media storing computer-readable instructions that when executed by one or more processors cause the one or more processors to perform operations for constructing a contour of a building from an aerial image, the operations comprising:
- identifying a source contour for the building from the image;
- determining a contour direction for a point on a source contour for the building;
- determining a plurality of rays extending from the point on the source contour based at least in part on the contour direction;
- constructing a graph model based at least in part on the plurality of rays;
- constructing a contour for the building based at least in part on a shortest path through the graph model;
- and
- providing the contour for the building for display on a display device;
- wherein the contour is constructed notwithstanding one or more occlusions of the object depicted in the image;
- wherein constructing a graph model based at least in part on the plurality of rays comprises:
- constructing a plurality of nodes of the graph model, each of the plurality of nodes representing one of the plurality of rays;
- and
- constructing a plurality of edges interconnecting the nodes in the graph model, each of the plurality of edges representing an intersection between the plurality of rays;
- and
- wherein constructing the graph model comprises determining an edge weight for each of the plurality of edges in the graph model based at least in part on an area between the rays associated with the nodes interconnected by the edge and the source contour.
Dependencies are inferred from the claim text.