System and method for locating solder bumps on semiconductor chips or chip carriers

• US 5,694,482 A
• Filed: 01/06/1997
• Issued: 12/02/1997
• Est. Priority Date: 11/08/1994
• Status: Expired due to Fees

• Current Assignee: Delaware Capital Formation Incorporated (Dover Corporation)
• Original Assignee: Universal Instruments Corporation (Delta Electronics Incorporated)
• Inventors: Maali, Ferevdoun, Poch, Leonard J., Hickok, George E.
• Primary Examiner(s): Bost, Dwayne
• Assistant Examiner(s): JOHNSON, BRIAN
• Application Number:
  US08/779,025
• Time in Patent Office: 330 Days
• Field of Search: 382/151, 382/125, 382/145, 382/294, 382/146, 348/94, 348/95, 348/87, 356/373
• US Class Current: 382/151
• CPC Class Codes: G06T 2207/30152 Solder
• CPC Class Codes: G06T 7/74 involving reference images or patches

Abstract

An automated vision system for identifying patterns, such as the solder bump patterns or leads of electronic components. The system is capable of correctly identifying the deviations of an imaged pattern from the nominal location of a reference pattern even if the imaged features are displaced more than half the minimum distance between any two adjacent features of the reference pattern. Further, the system is not limited to identifying patterns that have been merely translated relative to the reference pattern, but can be used if the imaged pattern has been rotated more than 45° relative to the reference pattern depending on the symmetry of the pattern.

Claims

1. A pattern recognition system comprising:

means for generating at least one reference pattern with a defined nominal location formed of individual features, wherein a minimum distance between any adjacent two individual features is defined as P_MIN within a number of individual non-colinear features greater than two,

means for storing a representation of said at least one reference pattern;

means for capturing an image of an actual pattern to obtain a representation of an imaged pattern; and

means for comparing said representation of said imaged pattern with said representation of said at least one reference pattern; and

means for identifying a deviation of said imaged pattern from said reference pattern, said deviation being determined even if the individual features in said imaged pattern have been translated by more than one half P_MIN.

2. A pattern recognition system as in claim 1, wherein said means for generating includes means for:

(I) representing a particular individual feature k of a blueprint as a multi-dimensional vector R_K that also represents position vectors of each other individual feature q with respect to a location of feature k in a complex coordinate system having its origin at individual feature k;

(ii) expressing each position vector by its two complex domain components to obtain position vector R_kq of features k and q; and

(iii) creating a vector model made of a plurality of prototype classes for the reference pattern by making a composite of representations for each individual feature.

3. A pattern recognition system as in claim 2, wherein said means for generating further includes means for reducing the number of prototype classes by applying uniqueness or global constraints on moduli or arguments of the position vectors in said model.

4. A pattern recognition system as in claim 2, wherein said means for generating further includes means for selecting two arbitrary anchor points and means for generating anchor vectors between each individual feature and said anchor points.

5. A pattern recognition system as in claim 4, wherein said means for comparing includes means for:

(i) representing a particular individual feature k of said imaged pattern as a multi-dimensional vector R_k that also represents the position vectors of each other individual feature q in said imaged pattern with respect to the location of feature k in a complex coordinate system having its origin at individual feature k;

(ii) expressing each position vector by its two complex domain components to obtain R_kq ;

(iii) creating a vector model of said imaged pattern by making a composite of representations for each individual feature; and

(iv) comparing the vector model of said imaged pattern with the vector models of every reference pattern.

6. A pattern recognition system as in claim 1, wherein said individual features are solder bumps on an electronic component.

7. A pattern recognition system as in claim 1, wherein said individual features are leads on an electronic component.

8. A pattern recognition system as in claim 1, further comprising a means for illuminating an actual pattern including an array of LED'"'"'s disposed oppositely of one another that project a shallow cone of light onto an electronic component.

9. A pattern recognition system as in claim 8, wherein said means for illuminating further includes foreground and background shields.

10. A pattern recognition system as in claim 1, wherein said means for capturing includes means for applying an image segmentation process to reduce the image of said imaged pattern to singular points in a complex reference frame.

11. A pattern recognition system as in claim 1, wherein said means for identifying a deviation determines said deviation, if the pattern is four-fold rotationally symmetric, even if the features have been rotated by as much as 45°

.

12. A pattern recognition system as in claim 1, wherein said means for identifying a deviation determines said deviation even if the features have been rotated more than 45°

depending on the symmetry of the pattern.

13. A pattern recognition system as in claim 1 wherein said means for generating includes means for generating representation vectors for individual features of said at least one reference pattern.

19. A method of recognizing a pattern as recited in claim 13 further comprising the step of illuminating an actual pattern formed of individual features so as to provide a sufficient degree of sharpness and contrast of each individual feature.

21. A method of recognizing a pattern as recited in claim 13 wherein said capturing step includes the step of applying an image segmentation process to reduce the image of said imaged pattern to singular points in a complex reference frame.

14. A method of recognizing a pattern, comprising the steps of:

generating at least one reference pattern formed of individual features, wherein a minimum distance between any adjacent two individual features is defined as P_MIN within a number of individual non-colinear features greater than two;

storing a representation of said at least one reference pattern;

capturing an image of an actual pattern to obtain a representation of an imaged pattern; and

comparing said representation of said imaged pattern with said representation of said at least one reference pattern;

identifying a deviation of said imaged pattern from said reference pattern, wherein said deviation can be determined even if the individual features in said imaged pattern have been translated by more than one half P_MIN.

15. A method as in claim 14, wherein said step of generating at least on reference pattern includes the steps of:

(I) representing a particular individual feature k of a blueprint as a multi-dimensional vector R_k that also represents position vectors of each other individual feature q with respect to a location of feature k in a complex coordinate system having its origin at individual feature k;

(ii) expressing each position vector by its two complex domain components to obtain position vector R_kq of features k and q; and

(iii) creating a vector model made of a plurality of prototype classes for the reference pattern by making a composite of representations for each individual feature.

16. A method as in claim 15, wherein said step of comparing includes the steps of:

(i) representing a particular individual feature k of said imaged pattern as a multi-dimensional vector R_k that also represents the position vectors of each other individual feature q in said imaged pattern with respect to the location of feature k in a complex coordinate system having its origin at individual feature k;

(ii) expressing each position vector by its two complex domain components to obtain position vector R_kq ;

(iii) creating a vector model of said imaged pattern by making a composite representations for each individual feature; and

(iv) comparing the vector model of said imaged pattern with the vector models of every reference pattern.

17. A method of recognizing a pattern as in claim 14, in said deviation identifying step, identifying the deviation, if the pattern is four-fold rotationally symmetric, even if the features of said pattern have been rotated by as much as 45°

.

18. A method of recognizing a pattern as in claim 14, in said deviation identifying step, identifying the deviation even if the features of said pattern have been rotated by more than 45°

depending on the symmetry of said pattern.

20. A method of recognizing a pattern as recited in claim 14 wherein said reference pattern generating step further includes the step of reducing the number of prototype classes by applying uniqueness or global constraints on the moduli or arguments of the position vectors in said model.

22. A pattern recognition system as in claim 14 wherein said means for generating includes means for generating representation vectors for individual features of said at least one reference pattern.