System and method for performing multi-image training for pattern recognition and registration

• US 8,315,457 B2
• Filed: 11/12/2008
• Issued: 11/20/2012
• Est. Priority Date: 12/21/2007
• Status: Active Grant

• Current Assignee: Cognex Corporation
• Original Assignee: Cognex Corporation
• Inventors: Bogan, Nathaniel, Wang, Xiaoguang, Wallack, Aaron S.
• Primary Examiner(s): Do, Anh Hong
• Application Number:
  US12/291,670
• Publication Number:
  US 20090290788A1
• Time in Patent Office: 1,469 Days
• Field of Search: 382/155, 382/159, 382/103, 382/118, 382/190, 382/224, 382/225, 382/179, 382/228, 382/154, 382/203, 600/408, 600/411, 707/780, 707/E17.039
• US Class Current: 382/155
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries

Abstract

A system and method for performing multi-image training for pattern recognition and registration is provided. A machine vision system first obtains N training images of the scene. Each of the N images is used as a baseline image and the N−1 images are registered to the baseline. Features that represent a set of corresponding image features are added to the model. The feature to be added to the model may comprise an average of the features from each of the images in which the feature appears. The process continues until every feature that meets a threshold requirement is accounted for. The model that results from the present invention represents those stable features that are found in at least the threshold number of the N training images. The model may then be used to train an alignment/inspection tool with the set of features.

Claims

1. A method for training a machine vision system, the method comprising the steps of:

obtaining N training images;

identifying one or more features that appear in at least M of the N training images;

selecting one of the N training images as a baseline image, substantially registering the N−

1 training images to the selected baseline image, and identifying corresponding features from the set of training images to the selected baseline image;

identifying one or more features that are supported by evidence from at least one of the N training images; and

generating, using a processor, a training model using the one or more identified features.

2. A method for training a machine vision system, the method comprising the steps of:

identifying a first feature from a first training image of a first instance;

identifying a second feature from a second training image of a second instance;

selecting one of the first training image and the second training image as a baseline image, substantially registering one of the first training image and the second training image to the selected baseline image, and identifying corresponding features from the set of training images to the selected baseline image;

identifying one or more features that are supported by evidence from at least one of the first training image and the second training image; and

generating, using a processor, a model comprising of the first feature and the second feature to train the machine vision system.

3. The method of claim 2 wherein the first feature is stable between the first and second training images.

4. The method of claim 2 wherein the first feature appears in at least M of N training images.

5. A system for training a machine vision system, the system comprising:

means for identifying a first feature from a first training image of a first instance;

means for identifying a second feature from a second training image of a second instance;

means for selecting one of the first training image and the second training image as a baseline image, substantially registering one of the first training image and the second training image to the selected baseline image, and identifying corresponding features from the set of training images to the selected baseline image;

means for identifying one or more features that are supported by evidence from at least one of the first training image and the second training image; and

means for generating a model comprising of the first feature and the second feature to train the machine vision system.

6. A method for identifying one or more stable features in a set of training images, the method comprising the steps of:

(a) selecting one of the set of training images as a baseline image;

(b) substantially registering the set of training images to the selected baseline image;

(c) determining corresponding features from the set of training images to the selected baseline image;

(d) repeating steps (a)-(c) with each image of the set of training images selected as the baseline image; and

(e) identifying, using a processor, the one or more stable features from the set of training images that represent a set of corresponding features.

7. The method of claim 6 further comprising the step of re-sampling the set of identified features to smooth spacing among the set of identified features.

8. The method of claim 6 further comprising the step of training a machine vision tool with the set of features.

9. The method of claim 6 wherein the step of identifying the one or more stable features further comprises identifying a set of features that appear in at least a predetermined number of the training images.

10. A non-transitory computer readable medium, including program instructions executing on a processor, for identifying one or more stable features in a set of training images, the computer readable medium including program instructions for:

(a) selecting one of the set of training images as a baseline image;

(b) substantially registering the set of training images to the selected baseline image;

(c) correspond features from the set of training images to the selected baseline image;

(d) repeating (a)-(c) with each image of the set of training images selected as the baseline image; and

(e) identifying, using a processor, the one or more stable features from the set of training images that represent a set of corresponding features.

11. A method for training a machine vision system, the method comprising the steps of:

obtaining N training images;

identifying a feature appearing in at least M of the N training images;

iteratively selecting each of the N training images and registering the other training images to the selected training image;

determining corresponding features from each of the other training images to the selected training image;

generating, using a processor, a model of the feature by identifying that the feature corresponds to at least M of the N training images; and

adding the identified feature to a model for use in training the machine vision system.

12. The method of claim 11 wherein the step of obtaining N training images further comprises the step of selecting, by a user, the N training images from a larger set of potential training images.