Method for identifying an object image

• US 20020064299A1
• Filed: 11/29/2001
• Published: 05/30/2002
• Est. Priority Date: 11/29/2000
• Status: Active Grant

• Current Assignee: Giken Trastem Co., Ltd. (OPTEX GROUP Co., Ltd.)
• Original Assignee: Giken Trastem Co., Ltd. (OPTEX GROUP Co., Ltd.)
• Inventors: Shibuya, Satoru
• Granted Patent: US 6,873,712 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06V 10/752 Contour matching

Abstract

A method for identifying an object image shown in a picture using density differences between an object image and a background. A screen taken by a camera is first divided into blocks; for each block, a Standard Object image is positioned based upon an arbitrary point on each picture; from the density differences in the picture, a normal vector of the outline portion of the Standard Object image is determined; then related vector data, that includes a position information up to a normal vector group and an angle information, is determined; the related vector data is stored as standard data in a block from which the normal vector is detected; a normal vector for a picture that shows an object image to be recognized is determined; from the normal vector, Answer points are determined based on standard data of the Standard Object image; and a focus point region comprised of Answer points is evaluated.

Claims

1. A method for identifying an object image that uses a density difference between a background image and an object image in a picture, said method comprising the steps of:

dividing a screen taken by a camera into a plurality of blocks;

placing a Standard Object image, which corresponds to said blocks, as an insert picture for each block in said screen, said placing being made based upon an arrangement point that is an arbitrary point in said blocks on a picture showing a background image and is used as a reference point;

determining, for said inserted picture, a standard normal vector group for outline portions of said Standard Object image based upon density differences in said Standard Object image and a background image;

determining related vector data which consists of a position information and an angle information, said position information being from an arrangement point of said Standard Object image to respective normal vectors of a respective said standard normal vector group of said Standard Object image, and said angle information being of a respective normal vector;

storing said related vector data as standard data for said Standard Object image in a block in which each normal vector of said Standard normal vector group is detected;

storing standard data based on said Standard object image for all of said divided blocks;

determining a normal vector group for outline portions of an object image based upon density differences in an object image and a background image for an inputted picture in a screen that shows an object to be identified and is taken by a camera;

determining Answer point groups, which are the same as arrangement points of respective said Standard Object image, from a normal vector group based upon standard data stored in blocks where a normal vector of a normal vector group appears; and

evaluating a focus point region formed by said Answer point groups.

2. The method for identifying an object image according to claim 1, wherein a tangent line group is used instead of said normal vector group, and a tangent line is used instead of said normal vector.