Object recognition system for a robot

• US 4,920,572 A
• Filed: 02/01/1989
• Issued: 04/24/1990
• Est. Priority Date: 02/01/1988
• Status: Expired due to Fees

• Current Assignee: Toyoda Koki Kabushiki Kaisha
• Original Assignee: Toyoda Koki Kabushiki Kaisha
• Inventors: Sugita, Kazuhiko, Hotta, Shigeo, Matsuda, Osamu, Yamakawa, Yoichi
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): COUSO, YON JUNG
• Application Number:
  US07/304,563
• Time in Patent Office: 447 Days
• Field of Search: 382/48, 382/22, 382/30, 382/25, 382/53, 382/50, 382/8, 364/513
• US Class Current: 382/152
• CPC Class Codes: B25J 9/1697 Vision controlled systems
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors

Abstract

A method and a system for recognizing the location of a cylindrical shaft is provided for use in positioning an end effector of an industrial robot which handles said shaft. In the method and system, a multilevel gray image of the cylindrical shaft which is generated by a change coupled device (CCD) image sensor is first differentiated, and the absolute values of picture elements constituting the differentiated gray image are then binarized on the basis of a threshold value so as to extract an accentuated pattern image which is inherent to the shape of the cylindrical shaft. The accentuated pattern image thus extracted is then collated with a reference accentuated pattern image (template) so as to generate a normal curve signal which has its peak point at a position where the axis of said cylindrical shaft is located. The normal curve data is then binarized on the basis of another predetermined threshold value to generate a single strip image. The center of the single strip image is calculated to determine the location of the cylindrical shaft.

Claims

1. An object recognition system for an industrial robot, comprising:

an image sensor for taking up a gray level image of said object to be handled by said robot;

differentiation means responsive to said gray level image of said object for differentiating said gray level image;

binarization means for binarizing the gray levels of picture elements constituting said gray level image, on the basis of a predetermined threshold value so as to extract an accentuated pattern image inherent to the shape of said object;

reference image storage means for storing an accentuated reference pattern image which is inherent to the shape of a model object;

image collation means for collating said accentuated pattern image extracted by said binarization means with said reference accentuated pattern image stored in said reference image storage means; and

position calculation means for calculating the location of said object based on the result of collation obtained by said image collation means;

wherein said object takes the form of a substantially cylindrical object, and wherein said image sensor scans said object in a direction transverse to the axis of said cylindrical object so that said binarization means is able to output said accentuated pattern image wherein two parallel strips appear extending in the axial direction of said cylindrical shaft.

2. An object recognition system as set forth in claim 1, wherein:

said position calculation means is operable for differentiating said result of collation obtained by said collation means, on the basis of another predetermined threshold value so as to extract a single strip which extends in the axial direction of said cylindrical shaft at the center between said two strips extracted by said image binarization means, and for calculating the center position of said single strip so as to locate the position of said cylindrical shaft.

3. An object recognition system for an industrial robot, for recognizing each of plural cylindrical shafts of the same kind which lie contacted in parallel relation with each other, said system comprising:

an image sensor for taking up a gray level image of said cylindrical shafts to be handled by said robot one at a time;

differentiation means responsive to said gray level image of said cylindrical shaft for differentiating said gray level image;

binarization means for binarizing the gray levels of picture elements constituting a positive component of said differentiated gray level image, on the basis of a first predetermined threshold value so as to extract a first accentuated pattern image inherent to the shape of each said cylindrical shaft, and for binarizing the gray levels of picture elements constituting a negative component of said differentiated gray level image, on the basis of a second predetermined threshold value so as to extract a second accentuated pattern image inherent to the shape of each said cylindrical shaft;

reference image storage means for storing first and second accentuated reference pattern images which are inherent to the shape of a cylindrical model object;

image collation means for collating said first accentuated pattern image for said each cylindrical shaft with said first accentuated reference pattern image so as to extract a first output image which indicates the location of a first accentuated image for a portion of each said cylindrical shaft and for collating said second accentuated pattern image for each said cylindrical shaft with said second accentuated reference pattern image so as to extract a second output image which indicates the location of a second accentuated image for another portion of said cylindrical shaft;

image AND means for ANDing said first and second output images from said image collation means with each other so as to extract a single output image which indicates the location of a center portion of each said cylindrical shaft; and

position calculation means for calculating the location of the center portion of each said cylindrical shaft based of said single output image obtained by said image AND means.

4. An object recognition method of recognizing a cylindrical shaft, said method comprising the steps of:

(a) generating a gray level image of said cylindrical shaft;

(b) differentiating the gray levels of picture elements constituting said gray level image;

(c) binarizing said differentiated gray levels of said picture elements on the basis of a predetermined threshold value so as to extract an accentuated pattern image which is inherent to the shape of said cylindrical shaft;

(d) providing a reference accentuated pattern image which is inherent to the shape of a cylindrical model shaft which has same cross section as said cylindrical shaft to be recognized;

(e) collating said accentuated pattern image extracted in step (c) with said reference accentuated pattern image; and

(f) calculating the location of said cylindrical shaft based on the result of a collation executed at step (e);

wherein said collating step (e) generates a normal curve output having its peak point at a position where said reference accentuated pattern image completely overlaps with said accentuated pattern image, and wherein said calculating step (f) includes;

binarizing said normal curve output on the basis of another predetermined threshold value so as to extract a single strip image which indicates the center portion of said cylindrical shaft; and

calculating the center position of said single strip image so as to determine the location of said cylindrical shaft.