SYSTEM AND METHOD FOR VISUAL RECOGNITION

• US 20110298939A1
• Filed: 06/03/2010
• Published: 12/08/2011
• Est. Priority Date: 06/03/2010
• Status: Active Grant

• Current Assignee: Visual Mathematics, LLC
• Original Assignee: Recognition Robotics, Inc.
• Inventors: Melikian, Simon
• Granted Patent: US 8,842,191 B2
• Time in Patent Office: Days
• US Class Current: 348/222.1
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]

Abstract

A method for generating a signal based on a visual image includes photographing a target object with a digital camera to obtain a target image; receiving the target image into a processor that is in communication with the camera; cross-correlating the target image with a structure having a variety of scales across the target image; and based on cross-correlating the target image, generating a signal for output on a device associated with the camera. A visual recognition system is also disclosed.

Claims

1. A method for generating a signal based on a visual image, the method comprising:

photographing a target object with a digital camera to obtain a target image;

receiving the target image into a processor that is in communication with the camera;

cross-correlating the target image with a structure having a variety of scales across the target image;

based on cross-correlating the target image, generating a signal for output on a device associated with the camera.

2. The method of claim 1, further comprising:

extracting icons from a coordinate generated by cross-correlating the target image;

normalizing each icon to a predetermined constant size;

comparing at least one normalized icon to stored icons associated with learned images to determine a match.

3. The method of claim 2, wherein signal includes an audible signal indicating the learned image that matches the target image.

4. The method of claim 2, wherein the camera includes a lens having a predetermined focal length and a sensor having a predetermined size, the method further comprising:

computing a distance between the target object and the camera based on comparing the at least one normalized icon to at least one stored icon from a learned image that matches the target image and based on the focal length and size of the sensor, wherein the signal includes an audible signal indicating the distance between the target object and the camera or a direction that the camera is located with respect to the target object.

5. The method of claim 1, further comprising:

identifying RGB data for at least one location within the target, wherein the signal includes an indication of at least one color found on the target object based on the RGB data.

6. The method of claim 5, further comprising:

determining whether the target object is associated with a particular color;

extracting icons from a coordinate generated by cross-correlating the target image;

normalizing each icon to a predetermined constant size;

comparing at least one normalized icon to stored icons associated only with learned images that include the particular color to determine a match.

7. The method of claim 5, further comprising:

identifying RGB data for at least two locations within the extracted icon, a first location being adjacent a centroid of the icon and a second location being adjacent a peripheral edge of the icon, wherein the signal includes an indication of at least one color found on the target object based on the RGB data.

8. The method of claim 1, further comprising:

determining a pattern on the target object based on cross-correlating the target image, wherein the signal includes an indication of the pattern on the target object.

9. The method of claim 1, wherein cross-correlating the target image with a structure having a variety of scales across the target image generates at least one coordinate corresponding to an extremum in the target image.

10. The method of claim 9, further comprising:

extracting icons from the at least one coordinate generated by cross-correlating the target image, wherein a size of the extracted icons is based on the scales of the structures used when cross-correlating the target image.

11. A recognition system comprising:

a camera;

an interface configured for receiving commands from an operator of the recognition system;

a speaker;

a speaker driver in communication with the speaker;

a memory;

a processor in electrical communication with the camera, the interface, the speaker driver and the memory, wherein the processor is programmed so as to receive a target image of a target object from the camera, to cross-correlate the target image with a structure, and to extract icons from the target image.

12. The system of claim 11, further comprising a housing, wherein the camera, the speaker, the speaker driver, the memory and the processor are disposed within the housing and the interface mounts to the housing so as to be accessible from an exterior of the housing.

13. The system of claim 11, wherein the processor is programmed to cross-correlate the target image with a structure locating extremum points on the target image and to extract icons from the target image based on the extremum points.

14. The system of claim 13, wherein the processor is programmed to normalize each extracted icon to a fixed constant size.

15. The system of claim 11, wherein the camera includes a lens having a predetermined focal length and a sensor, wherein the processor is programmed to compute a distance between the target object and the camera based on comparing the normalized icons to stored icons from at least one learned image that matches the target image.

16. The system of claim 15, wherein the processor is programmed to output a signal to the speaker driver, wherein the signal includes an audible signal indicating the distance between the target object and the camera.

17. The system of claim 11, wherein the processor is programmed to identify RGB data for at least one location within the target image and to output a signal to the speaker driver, wherein the signal includes an indication of at least one color found on the target object based on the RGB data.

18. The system of claim 11, wherein the processor is programmed to determine a pattern on the target object based on cross-correlating the target image and to output a signal to the speaker driver, wherein the signal includes an indication of the pattern on the target object.

19. A recognition system comprising:

a camera;

an interface configured for receiving commands from an operator of the recognition system;

a speaker;

a speaker driver in communication with the speaker;

a memory;

a processor in electrical communication with the camera, the interface, the speaker driver and the memory, wherein the processor is programmed so as to receive a target image of a target object from the camera, to cross-correlate the target image with a structure, and to output a signal to the speaker driver, wherein the signal includes an indication of the pattern on the target object based on cross-correlating the target image with the structure.

20. The system of claim 19, wherein the processor is programmed to identify RGB data for at least one location based on cross-correlating the target image and to output a signal to the speaker driver, wherein the signal includes an indication of at least one color found on the target object based on the RGB data.