# US 9,519,843 B2 · Object recognition device

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 9,519,843 B2
- Document ID: US9519843B2
- Status: Granted
- Kind: B2
- Application number: 14/126,690
- Filed: 2012-06-14
- Published: 2014-05-15
- Granted: 2016-12-13
- Earliest claimed: 2011-06-17
- Assignees: Toyota Jidosha Kabushiki Kaisha; Eidgenossische Technische Hochschule Zurich
- Applicants: Nima Razavi; Juergen Gall; Luc Van Gool; Ryuji Funayama
- Inventors: Nima Razavi; Juergen Gall; Luc Van Gool; Ryuji Funayama
- Web page: https://patentgrep.com/patent/US9519843B2

## CPC

- G06V 20/58: Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06F 18/231: Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
- G06F 18/24: Classification techniques
- G06F 18/24323: {Tree-organised classifiers}
- G06V 10/48: by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
- G06V 10/7625: {Hierarchical techniques, i.e. dividing or merging patterns to obtain a tree-like representation; Dendograms}
- G06V 10/764: using classification, e.g. of video objects
- G06V 20/56: exterior to a vehicle by using sensors mounted on the vehicle
- G06V 2201/10: Recognition assisted with metadata
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A learning unit generates a function table indicating the relationship between the class number and position information of an object and the probability of appearance of the object for each small area image pattern of a code book, calculates a sharing matrix indicating the commonality of a feature amount between the classes, makes a tree diagram in which the classes with a similar feature amount are clustered, and calculates the weight of each node in the tree diagram for each small area image pattern. The recognition processing unit compares image data captured by a camera with the code book, selects the closest small area image pattern, extracts the class related to the node with the smallest weight among the nodes with a weight equal to or greater than a threshold value, and votes the position information of the small area image pattern for the class, thereby recognizing the object.

## Claims

### Claim 1

An object recognition device that recognizes an object on the basis of a captured image, comprising:
a function table acquiring unit configured to acquire a function table indicating a relationship between position information of the object and a probability of appearance of the object for a plurality of classes of the object for a plurality of image patterns indicating a portion of the object;
a pattern appearance frequency calculating unit configured to calculate the frequency of appearance of each image pattern for the class of the object, using the function table acquired by the function table acquiring unit;
a pattern selection unit configured to compare the captured image with the plurality of image patterns and to select an image pattern corresponding to the captured image;
a class extracting unit configured to extract the class at which the frequency of appearance of the image pattern selected by the pattern selection unit is equal to or greater than a predetermined value;
a voting unit configured to vote the position information of the image pattern selected by the pattern selection unit for the class extracted by the class extracting unit; and
a recognition unit configured to recognize the object on the basis of a voting result for the class extracted by the extracting unit;
wherein the class of the object indicates a kind of the object;
wherein each of the function table acquiring unit, the pattern appearance frequency calculating unit, the pattern selection unit, the class extracting unit, the voting unit and the recognition unit are implemented via a CPU (central processing unit);
wherein the pattern appearance frequency calculating unit calculates the commonality of a feature amount including the image pattern between the plurality of classes on the basis of the function table acquired by the function table acquiring unit, makes a tree diagram in which similar classes are clustered on the basis of the commonality of the feature amount and calculates a weight of each node in the tree diagram as the frequency of appearance of the image pattern for the class.

### Claim 2 (depends on 1)

The object recognition device according to claim 1, wherein each of the plurality of image patterns are individually derived directly from respective single source images.