Technique for object orientation detection using a feed-forward neural network

• US 5,060,276 A
• Filed: 05/31/1989
• Issued: 10/22/1991
• Est. Priority Date: 05/31/1989
• Status: Expired due to Term

• Current Assignee: American Telephone & Telegraph Company (AT&T, Inc.), Bell Telephone Laboratories, Inc. (Nokia Corporation)
• Original Assignee: AT&T, Inc.
• Inventors: Morris, Robert J. T., Rubin, Lawrence D.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Johns, Andrew W.
• Application Number:
  US07/359,296
• Time in Patent Office: 874 Days
• Field of Search: 382/61, 382/8, 382/46, 382/48
• US Class Current: 382/151
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G06V 10/242 by image rotation, e.g. by 90 degrees

Abstract

The present invention relates to a technique in the form of an exemplary computer vision system for detecting the orientation of text or features on an object of manufacture. In the present system, an image of the features or text is used to extract lines using horizontal bitmap sums, and then individual symbols using vertical bitmap sums, using thresholds with each of the sums. The separated symbols are then appropriately trimmed and sealed to provide individual normalized symbols. A Decision Module comprising a Feed-Forward Neural network and a sequential decision arrangement determines the "up", "down" or "indeterminate" orientation of the text after a variable number of symbols have been processed. The system can then compare the determined orientation with a database to further determine if the object is in the "right-side up" "upside down" or "indeterminate" orientation.

Claims

1. A method of determining the orientation of features or text on an object, the method comprising the steps of:

(a) extracting a predetermined sized image of each of at least one of the features, or at least one symbol of the text, disposed on the object;

(b) normalizing each of the at least one feature or symbol obtained in step (a) within a second predetermined sized image;

(c) processing the predetermined sized image for each of the at least one feature or symbol obtained in step (a) or (b) for determining a similarity measure between the second sized image and a reference image and determining therefrom first and second conditional probabilities indicative of the at least one feature or symbol having an "up" or "down" orientation, respectively;

(d) determining from the first and second conditional probabilities for each of the at least one feature or symbol obtained in step (c) whether the feature or text is in the "up" or "down" orientation or is in an "indeterminate" orientation; and

(e) determining from the indication of the orientation of each one feature or symbol obtained in step (d) whether the object is in the "right-side up", "upside-down" or indeterminate orientation.

2. The method according to claim 1 wherein in performing step (a), performing the substeps of:

(a1) summing the binary values for each of the rows of a bitmap of the features or text and applying a threshold value to the sums of the rows for determining the location of lines of features or text; and

(a2) summing the binary values for each of the columns of a bitmap of the features or text and applying a second threshold value to the sums of the columns for determining, in conjunction with the result of step (a1), the location of the individual features or symbols of the text.

3. The method according to claim 1 or 2 wherein in performing step (c), performing the substeps of:

(c1) determining, from a bitmap of each normalized extracted feature or symbol of the text obtained in step (a) or (b), the probability that an input vector is observed, by measuring the Hamming distance using bitwise "exclusive OR" and table lookup operations between the input vector and each of a set of reference features or symbols in a particular "up" or "down" orientation, where the input vector comprises a distortion of a similar feature or symbol from the reference set; and

(c2) in response to step (c1), determining the conditional probabilities of obtaining the extracted symbol based on the extracted symbol having an up or down orientation.

4. The method according to claim 1 or 2 wherein in performing step (d):

multiplying the ratio of the first and second conditional probabilities determined for each symbol processed in step (c) with the product of the ratio obtained for each prior processed symbol until a certain threshold is attained indicating that the feature or text is in the "up", "down" or "indeterminate" orientation.

5. The method according to claim 1 or 2 wherein in performing step (d), performing the substeps of:

(d1) providing a determination of whether each symbol of the text is in the "up", "down" or "indeterminate" orientation; and

(d2) from a consensus of the determinations made in step (d1), generating a determination of whether the text is in the "up", "down" or "indeterminate" orientation.

6. Apparatus for determining the orientation of a feature or text on an object comprising:

means for extracting a feature or symbol of the text from an image, and generating an output signal representative of a normalized sized bitmap for each extracted symbol; and

a decision arrangement comprising

first means responsive to the output signal from the extracting and generating means for determining a similarity measure between each such output signal for each extracted feature or symbol and each of a set of reference images, and generating an output signal for each extracted feature or symbol indicating the probability that each such extracted feature or symbol is in an "up", "down" or "indeterminate" orientation, and

second means for generating an output signal indicating that the object is in the "up", "down" or "indeterminate" orientation in response to an accumulation of each output signal of the first means.

7. Apparatus according to claim 6 wherein the decision arrangement further comprises:

third means for determining the orientation of the object from the output signal of the second means.

8. Apparatus according to claims 6 or 7 wherein the extracting and generating means comprises:

means for generating a sum of the binary values for each of the rows of a bitmap of the image of each feature or text and applying a threshold value to the sums of the rows for generating an output signal indicating the location of lines within the feature or text;

means for generating a sum of each of the columns of a bitmap of the image of the feature or text and applying a second threshold value to the sums of the columns for determining, from both the output signal from the first means and the sums of the columns, the location of each feature or each symbol within the text.

9. Apparatus according to claim 6 or 7 wherein the second means of the decision apparatus comprises:

means responsive to each said output signal from the first means of the decision apparatus for multiplying a ratio of the probabilities determined for each said output signal with a product formed from the ratio of probabilities of each prior processed feature or symbol of the text until a certain threshold is attained indicating that each feature or text is in the "up", "down" or "indeterminate" orientation and generating an output signal representative of such orientation determination.

10. Apparatus according to claim 6 or 7 wherein the second means of the decision apparatus comprises:

means for generating an output signal indicating whether each feature or symbol of the text is in the "up", "down" or "indeterminate" orientation, and

means for forming a consensus from the output signals from the generating means of the second means for determining whether the features or text are in the "right-side up", "upside-down" or "indeterminate" orientation.