Position recognizing method and apparatus therefor

• US 5,136,661 A
• Filed: 03/06/1991
• Issued: 08/04/1992
• Est. Priority Date: 03/08/1990
• Status: Expired due to Term

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Taira, Nobutaka, Kobayasi, Akira, Mizuoka, Seiji
• Primary Examiner(s): MOORE, DAVID K
• Application Number:
  US07/665,228
• Time in Patent Office: 517 Days
• Field of Search: 382/30, 382/38, 382/48, 382/44, 382/51, 382/22, 382/32, 382/50, 382/60, 382/61, 382/34, 382/9, 358/96, 358/125, 358/101, 358/105, 358/107, 340/707
• US Class Current: 382/288
• CPC Class Codes: G05B 2219/4703 View whole surface before edge detection, coarse scan then fine tracking
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 7/73 using feature-based methods

Abstract

In a position recognizing method and an apparatus, an edge is detected by scanning a one-dimensional model including an edge to be recognized and one-dimensional edge detecting data of an image obtained via a camera so as to determine the difference between the density of the pixel of the one-dimensional model and the density of the corresponding pixel of the one-dimensional edge detecting data and the geometric balance therebetween. Then, the amount of data for detecting the edge becomes smaller and the pattern or parts can be recognized without controlling the binary level and at a high speed.

Claims

1. A position recognizing method comprising the steps of:

taking an image of an object having a pattern to be recognized via a camera so as to get an image signal taken out from the image;

making an x-y-f space, where (f) is density of an x-y image plane, from the image signal, the space including a one-dimensional edge detecting data of the image;

making an edge model g(u) of a specified size including an edge of the pattern of the image on the basis of the one-dimensional edge detecting data in a direction unparallel with a line composed of a boundary between density in difference areas of the image;

taking out a one-dimensional edge detecting data fij as a partial template, where fij is density of a pixel (i, j) in the image plane;

scanning the edge model g(u) and then calculating a sum (D_1n) of difference in density between pixels corresponding to the one-dimensional edge detecting data fij(u) sequentially taken out as the partial template and the edge model g(u);

calculating a geometric unbalance degree (D_2n) between the edge model g(u) and the data fij(u) with respect to a target point of the edge model g(u);

calculating an edge evaluating exponent based on a mis-match degree D_1n) and the unbalance degree (D_2n) as follows;

Dn = m₁ D_1n + m₂ D_2n (where (m₁) and (m₂) are weight parameters) and mapping a point of an x-y-f space representing the partial template of the x-y-f space to an x-y-D space; and

detecting a minimum point of a curve of the x-y-D space composed of representative points of the template mapped from the x-y-f space to the x-y-D space by the one-dimensional edge detecting converting means.

2. The position recognizing method as claimed in claim 1, further comprising the steps of:

detecting density of a background or the pattern of the image;

setting as a window frame a range to be processed in the image;

calculating a separating level between the density of the background and the pattern of the image; and

emphasizing the image of the one-dimensional edge detecting data based on the detected density in accordance with the level.

3. The position recognizing method as claimed in claim 1, further comprising the steps of making histogram of the one-dimensional edge detecting data and detecting a separating level between density of a background and the pattern of the image based on the pattern of a specified size so as to perform image emphasis of the one-dimensional edge detecting data by the detected density.

4. The position recognizing method as claimed in claim 1, further comprising the steps of mapping only representative points of the template as candidate points when an edge evaluating exponent is within an allowable range and calculating a weighted average of a plurality of the candidate points (x, y, D) by an edge evaluating exponent (Dn) so as to detect an edge (x, y).

7. The position recognizing method as claimed in claim 4, further comprising the steps of, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of the allowable range when a number of candidate points becomes maximum during scanning of the edge model.

5. The position recognizing method as claimed in claim further comprising the steps of mapping only representative points of the template as candidate points when an edge evaluating exponent is within an allowable range in detecting a plurality of edges according to the one-dimensional edge detecting data so as to detect an edge and performing distance verification of the plurality of the candidate points based on a distance relationship between edges to be recognized.

6. The position recognizing method as claimed in claim 5, further comprising the steps of grouping points mapped as the candidate points according to the distance relationship between the mapped candidate points, calculating a weighted average of candidate points (x, y, D) of each group, and detecting the candidate points (x, y, D) as the representative value of each group so as to measure distance.

8. The position recognizing method as claimed in claim 5, further comprising the steps of, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of the allowable range when a number of candidate points becomes maximum during scanning of the edge model.

9. The position recognizing method as claimed in claim 6, further comprising the steps of, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of an allowable range when a number of candidate points becomes maximum during scanning of the edge model.

10. A position recognizing apparatus comprising:

a camera for taking an image of an object having a pattern to be recognized so as to get an image signal taken out from the image;

a one-dimensional edge detecting data making means for making an x-y-f space, where (f) is density of an x-y image plane, from the image signal, the space including a one-dimensional edge detecting data of the image;

a one-dimensional edge model making means for making an edge model g(u) of a specified size including an edge of the pattern of the image on the basis of the one-dimensional edge detecting data in a direction unparallel with a line composed of a boundary between density in difference areas of the image;

a one-dimensional edge detecting template takingout means for taking out a one-dimensional edge detecting data fij as a partial template, where fij is density of a pixel (i, j) in the image plane;

a mis-match degree calculating means for scanning the edge model g(u) and then calculating a sum (D_1n) of difference in density between pixels corresponding to the one-dimensional edge detecting data fij(u) sequentially taken out as the partial template and the edge model g(u);

an unbalance degree calculating means for calculating a geometric unbalance degree (D_2n) between the edge model g(u) and the data fij(u) with respect to a target point of the edge model g(u);

a one-dimensional edge detecting converting means for calculating an edge evaluating exponent based on the mis-match degree (D_1n) and the unbalance degree (D_2n) as follows Dn = m₁ D_1n + m₂ D_2n (where (m₁) and (m₂) are weight parameters) and mapping a point of an x-y-f space representing the partial template of the x-y-f space to an x-y-D space; and

a minimum value detecting means for detecting a minimum point of a curve of the x-y-D space composed of representative points of the template mapped from the x-y-f space to the x-y-D space by the one-dimensional edge detecting converting means.

11. The position recognizing apparatus as claimed in claim 10, wherein the detecting means detects density of a background or the pattern of the image;

the apparatus further comprising a window frame setting means for setting as a window frame a range to be processed in the image, and an image emphasis means for calculating a separating level between the density of the background and the pattern of the image and emphasizing the image of the one-dimensional edge detecting data based on the detected density in accordance with the level.

12. The position recognizing apparatus as claimed in claim 10, further comprising an image emphasis means for making histogram of the one-dimensional edge detecting data and detecting a separating level between density of a background and the pattern of the image based on the pattern of a specified size so as to perform image emphasis of the one-dimensional edge detecting data by the detected density.

13. The position recognizing apparatus as claimed in claim 10, wherein the one-dimensional edge detecting converting means maps only representative points of the template as candidate points when an edge evaluating exponent is within an allowable range and calculates a weighted average of a plurality of the candidate points (x, y, D) by an edge evaluating exponent (Dn) so as to detect an edge (x, y).

14. The position recognizing apparatus as claimed in claim 10, wherein the one-dimensional edge detecting converting means maps only representative points of the template as candidate points when an edge evaluating exponent is within an allowable range in detecting a plurality of edges according to the one-dimensional edge detecting data so as to detect an edge and performs distance verification of the plurality of the candidate points based on a distance relationship between edges to be recognized.

15. The position recognizing apparatus as claimed in claim 14, further comprising means for grouping points mapped as the candidate points according to the distance relationship between the mapped candidate points, calculating a weighted average of candidate points (x, y, D) of each group, and detecting the candidate points (x, y, D) as the representative value of each group so as to measure distance.

16. The position recognizing apparatus as claimed in claim 13, further comprising means for, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of the allowable range when a number of candidate points becomes maximum during scanning of the edge model.

17. The position recognizing apparatus as claimed in claim 14, further comprising means for, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of the allowable range when a number of candidate points becomes maximum during scanning of the edge model.

18. The position recognizing apparatus as claimed in claim 15, further comprising means for, in detecting the candidate points, setting a maximum number of a plurality of candidate points and updating a value of an allowable range when a number of candidate points becomes maximum during scanning of the edge model.