Object recognition
AssigneeGoogle Inc.
InventorsAlexander Toshkov Toshev, King Hong Thomas Leung, Jiwoong Jack Sim
Life of this patent
Estimating the term…
Abstract
Techniques for a shape descriptor used for object recognition are described. Tokens of an object in digital image data are captured, where tokens can be edges, interest points or even parts. Geometric configurations of the tokens are captured by describing portions of the shape of the object. The shape of such configurations is finely quantized and each configuration from the image is assigned to a quantization bin. Objects are recognized by utilizing a number of quantization bins as features. This Abstract is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims.
Claims26
One or more computing devices configured to recognize objects in digital image data, the one or more computing devices comprising:
- an image input module configured to input images of objects;
- an image processing module configured to receive the input images from the image input module, the image processing module further configured to:
- detect tokens of an object in digital image data, wherein the tokens are detected from at least two selected from the group consisting of edges, contours, interest points, and parts of the object,
- capture a geometric configuration of the tokens by describing a plurality of portions of the shape of the object,
- quantize the geometric configuration of the tokens;
- assigning the geometric configuration of the tokens to a quantization bin, and
- recognize the object by utilizing a randomly selected subset of a plurality of quantization bins as features;
- and
- a memory in communication with the image processing module, the memory storing the recognized object.
A method implemented by one or more computing devices configured to recognize objects in digital image data, the method comprising:
- detecting tokens of an object in digital image data, wherein the tokens are detected from at least two selected from the group consisting of edges, contours, interest points, and parts of the object;
- capturing a geometric configuration of the tokens by describing a plurality of portions of the shape of the object;
- quantizing the geometric configuration of the tokens;
- assigning the geometric configuration of the tokens to a quantization bin;
- recognizing the object by utilizing a randomly selected subset of a plurality of quantization bins as features;
- and
- storing the recognized object in a memory.
One or more non-transitory computer-readable media storing processor-executable instructions that, when executed, cause one or more processors to perform operations that recognize objects in digital image data, the operations comprising:
- detecting tokens of an object in digital image data, wherein the tokens are detected from at least two selected from the group consisting of edges, contours, interest points, and parts of the object;
- capturing geometric configuration of the tokens by describing a plurality of portions of the shape of the object;
- quantizing the geometric configuration of the token;
- assigning the geometric configuration of the tokens to a quantization bin;
- and
- recognizing the object by utilizing a randomly selected subset of a plurality of quantization bins as features.
Dependencies are inferred from the claim text.