Computer implemented method for recognizing an object based on a correspondence relationship between object feature points and pre-registered model feature points

• US 8,861,834 B2
• Filed: 03/05/2008
• Issued: 10/14/2014
• Est. Priority Date: 03/09/2007
• Status: Active Grant

• Current Assignee: Omron Corporation
• Original Assignee: Omron Corporation
• Inventors: Fujieda, Shiro
• Primary Examiner(s): Repko, Jason M
• Assistant Examiner(s): ROWLAND, JHERE M
• Application Number:
  US12/529,910
• Publication Number:
  US 20100098324A1
• Time in Patent Office: 2,414 Days
• Field of Search: None
• US Class Current: 382/154
• CPC Class Codes: B25J 9/1697 Vision controlled systems
• CPC Class Codes: G01B 11/002 for measuring two or more coordinates
• CPC Class Codes: G01B 11/272 using photoelectric detection means
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 2207/30164 Workpiece; Machine component
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G06V 10/757 Matching configurations of points or features

Abstract

A recognition processing method and an image processing device ends recognition of an object within a predetermined time while maintaining the recognition accuracy. The device extracts combinations of three points defining a triangle whose side length satisfy predetermined criterion values from feature points of the model of a recognition object, registers the extracted combinations as model triangles, and similarly extracts combinations of three points defining a triangle whose side lengths satisfy predetermined criterion values from feature points of the recognition object. The combinations are used as comparison object triangles and associated with the respective model triangles. The device calculates a transformation parameter representing the correspondence relation between each comparison object triangle and the corresponding model triangle using the coordinates of the corresponding points (A and A′, B and B′, and C and C′), determines the goodness of fit of the transformation parameters on the relation between the feature points of the model and those of the recognition object. The object is recognized by specifying the transformation parameters representing the correspondence relation between the feature points of the model and those of the recognition object according to the goodness of fit determined for each association.

Claims

1. A computer-implemented method for recognizing a recognition object based on a correspondence relationship between feature points of the recognition object and pre-registered feature points of a model, the feature points being distributed in a plane or a space, the method comprising the steps of:

(i) setting model combinations out of combinations each made up of a particular number n (n≧

2) of feature points of the feature points of the model, the distance between at least a pair of feature points of each model combination meeting a predefined criterion value, and registering in a memory information for specifying the set model combinations;

(ii) setting comparison object combinations out of combinations each made up of the particular number n of feature points of the feature points of the recognition object, the distance between at least a pair of feature points of each comparison object combination meeting the criterion value;

(iii) associating in turn the model combinations specified by the registered information in the memory with the comparison object combinations set in the (ii) step, obtaining per associated combinations a transformation parameter for a case of transforming the feature points of one of the combinations into the feature points of the other combination, and determining goodness of fit of the transformation parameter on a relationship between the feature points of the model and the feature points of the recognition object; and

(iv) based on each goodness of fit obtained in the (iii) step, specifying a transformation parameter indicating the correspondence relationship between the feature points of the model and the feature points of the recognition object, wherein in the (i) and (ii) steps, the particular number n is set to three, and such combinations of feature points are set as to constitute a triangle whose side lengths each have a value between a predetermined upper limit and a predetermined lower limit,

wherein in the (iii) step, after the transformation parameter is obtained for a model combination and a comparison object combination that are associated with each other, either three-dimensional coordinates of the feature points of the model or the feature points of the recognition object are transformed using the transformation parameter, each of the transformed three-dimensional coordinates of the transformed feature points is determined whether or not to match a corresponding one of each of the three-dimensional coordinates of the non-transformed feature points, and a numerical value indicating frequency of the transformed three-dimensional coordinates of the transformed feature points matching three-dimensional coordinates of the non-transformed feature points is set as the goodness of fit of the transformation parameter;

wherein in the (iv) step, specifying the transformation parameter includes comparing the numerical value indicating goodness of fit with a predetermined goodness of fit value.

2. The method according to claim 1, wherein the feature points to be processed are a set of feature points included in an image of the recognition object.

3. The method according to claim 1, wherein in the (i) and (ii) steps, such combinations of feature points are set as to constitute an equilateral triangle whose side length has a value between the upper and lower values.

4. The method according to claim 1, wherein the angles at both ends of the side of the triangle each fall within a predefined criterion angle range.

5. An image processing device performing acquisition of a plurality of feature points of a recognition object that are distributed in a plane or a space through processing on an image of the recognition object, and recognition of the recognition object based on a correspondence relationship between the feature points and pre-registered feature points of a model, the device comprising:

input means for inputting a predetermined criterion value for the inter-feature-point distance of the feature points;

model setting means for setting model combinations out of combinations each made up of a particular number n (n≧

2) of feature points of the feature points of the model, the distance between at least a pair of feature points of each model combination meeting the criterion value;

registering means for registering in a memory information for specifying the model combinations set by the model setting means;

combination setting means for setting comparison object combinations out of combinations each made up of the particular number n of feature points of the feature points extracted from the input image, the distance between at least a pair of feature points of each comparison object combination meeting the criterion value;

matching means for associating in turn the model combinations specified by the registered information in the memory with the comparison object combinations set by the combination setting means, obtaining per associated combinations a transformation parameter for a case of transforming the feature points of one of the combinations into the feature points of the other combination, and determining goodness of fit of the transformation parameter on a relationship between the feature points of the model and the feature points of the recognition object; and

specifying means for specifying a transformation parameter indicating the correspondence relationship between the feature points of the model and the feature points of the recognition object, based on each goodness of fit determined by the matching means,

wherein the particular number is set to three, and such combinations of feature points are set as to constitute a triangle whose sides each have a value between a predetermined upper limit and a predetermined lower limit,

wherein after the transformation parameter is obtained for a model combination and a comparison object combination that are associated with each other, either three-dimensional coordinates of the feature points of the model or the feature points of the recognition object are transformed using the transformation parameter, each of the transformed three-dimensional coordinates of the transformed feature points is determined whether or not to match a corresponding one of each of the three-dimensional coordinates of the non-transformed feature points and a numerical value indicating frequency of the transformed three-dimensional coordinates of the transformed feature points matching three-dimensional coordinates of the non-transformed feature points is set as the goodness of fit of the transformation parameter;

wherein specifying the transformation parameter includes comparing the numerical value indicating goodness of fit with a predetermined goodness of fit value.

6. An image processing device performing input of images from a plurality of photographing devices installed to photograph a plurality of recognition objects having the same shape, acquisition of the space coordinates of a plurality of feature points of each recognition object that are distributed in a space through a three-dimensional measurement process using the input images, and individual recognition of the plurality of recognition objects based on a correspondence relationship between the feature points and pre-registered feature points of a model, the device comprising:

input means for inputting a predetermined criterion value for the inter-feature-point distance of the feature points;

model setting means for setting model combinations out of combinations each made up of a particular number n (n≧

3) of feature points of the feature points of the model, the distance between at least a pair of feature points of each model combination meeting the criterion value;

registering means for registering in a memory information for specifying the model combinations set by the model setting means;

combination setting means for setting comparison object combinations out of combinations each made up of the particular number n of feature points of the feature points acquired through the three-dimensional measurement, the distance between at least a pair of feature points of each comparison object combination meeting the criterion value;

matching means for associating in turn the model combinations specified by the registered information in the memory with the comparison object combinations set by the combination setting means, obtaining per associated combinations a transformation parameter for a case of transforming the feature points of one of the combinations into the feature points of the other combination, and determining goodness of fit of the transformation parameter on a relationship between the feature points of the model and the feature points of each recognition object; and

object recognizing means for specifying at least one of the goodnesses of fit determined by the matching means, the goodness of fit being above a predetermined criterion value, and recognizing one recognition object per specified goodness of fit using the transformation parameter corresponding to the goodness of fit, wherein the particular number is set to three, and such combinations of feature points are set as to constitute a triangle whose sides each have a value between a predetermined upper limit and a predetermined lower limit,

wherein after the transformation parameter is obtained for a model combination and a comparison object combination that are associated with each other, either three-dimensional coordinates of the feature points of the model or the feature points of the recognition object are transformed using the transformation parameter, each of the transformed three-dimensional coordinates of the transformed feature points is determined whether or not to match a corresponding one of each of the three-dimensional coordinates of the non-transformed feature points, and a numerical value indicating frequency of the transformed three-dimensional coordinates of the transformed feature points matching three-dimensional coordinates of the non-transformed feature points is set as the goodness of fit of the transformation parameter;

wherein obtaining the transformation parameter includes comparing the numerical value indicating goodness of fit with a predetermined goodness of fit value.