Method of and apparatus for pattern recognition

• US 4,567,610 A
• Filed: 07/22/1982
• Issued: 01/28/1986
• Est. Priority Date: 07/22/1982
• Status: Expired due to Term

• Current Assignee: Wayland Research Incorporated
• Original Assignee: Wayland Research Incorporated
• Inventors: McConnell, Robert K.
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/400,948
• Time in Patent Office: 1,286 Days
• Field of Search: 382/18, 382/30, 358/107
• US Class Current: 382/170
• 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

Abstract

A method of and an apparatus for analysis of patterns both in static and dynamic modes. A test histogram is described in terms of the optimum code described for a reference histogram.

Claims

1. A method of recognizing a pattern in a test object, said method comprising the steps of:

specifying properties characteristic of said pattern;

specifying discrete ranges of values of said properties;

measuring the values of said properties in said test object;

arranging the measured values in at least one test histogram;

determining a reference set of values of said properties and arranging said set as at least a first reference histogram; and

comparing said test and reference histograms by determination of the value of a function which provides a measure of the amount of information necessary to express said at least one test histogram in terms of the optimum code for describing at least said first reference histogram.

2. A method as defined in claim 1 including the steps of digitizing said reference set of values and said measured values prior to aranging same in said respective reference and test histograms.

3. A method as defined in claim 1 wherein

said step of determining said reference set of values includes arranging said set in a plurality of reference histograms,

said step of comparing includes comparing said plurality of reference histograms with said test histogram by determination of, for each such comparison, the value of a function which provides a measure of the amount of information necessary to express said test histogram in terms of each of said reference histograms;

and

said method further including the step of selecting that one of said plurality of reference histograms which provides for said function a value bearing a predetermined relationship to a desired value.

4. A method as defined in claim 1 wherein

said test object is one of a plurality of test objects,

said step of measuring includes measuring the values of said properties for each of said test objects,

said step of arranging includes arranging the measured values of said properties in a plurality of test histograms,

said step of comparing includes comparing said plurality of test histograms with said first reference histogram by determination, for each such comparison, of the value of a function which provides a measure of the amount of information necessary to express each of said test histograms in terms of said first reference histogram;

and

said method further including the step of selecting that one of said plurality of test histograms which provides for said function a value bearing a predetermined relationship to a desired value.

9. A method as defined in claim 1 including the step of

generating a signal, based upon the result of the comparison of histograms.

10. A method as defined in claim 1 wherein said properties comprise (a) positions in a field of view, and (b) the nature of a signal arising from a unit area of said field at each of said positions.

11. A method as defined in claim 10 wherein each of said positions in said field of view is defined in terms of location within a rectangular array of positions.

12. A method as defined in claim 10 wherein each of said positions in said field of view is defined in terms of location in one of a series of concentric rings.

13. A method as defined in claim 10 wherein each of said positions in said field of view is defined in terms of location within an angular sector around a center within said field.

14. A method as defined in claim 10 wherein said nature of said signal is the intensity thereof.

15. A method as defined in claim 10 wherein the nature of said signal is the wavelength range thereof.

16. A method as defined in claim 10 wherein each of said positions in said field of view is defined in terms of location in adjacent rows.

17. A method as defined in claim 10 wherein each of said positions is defined in terms of adjacent columns.

18. A method as defined in claim 10 wherein said properties comprise the number of selected points in said field of view having a value falling within each of said discrete ranges.

19. A method as defined in claim 1 wherein said properties comprise (a) edges of said test object and (b) the slope of said edges at selected points.

20. A method as defined in claim 1 wherein the values in at least one of said histograms are sums of weighted values of a given sequence of values.

21. A method as defined in claim 20 wherein said at least one histogram is said test histogram.

22. A method as defined in claim 20 wherein said at least one of said histograms is said reference histogram.

23. A method as defined in claim 1 wherein the values in both said refence and test histograms are sums of weighted values of given sequences of values.

24. A method as defined in claim 1 wherein said function is I(T:

R) and is defined as ##EQU11## where T is a test histogram;

R is a reference histogram;

I(T;

R) is the information of T expressed in terms of R;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i is the probability of the ith symbol occurring in R; and

Q is the number of different symbols in T.

25. A method as defined in claim 1 wherein said function is I(T:

T) and is defined as ##EQU12## where T is a test histogram;

I(T;

T) is the information of T expressed in terms of T;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i '"'"' is the probability of the ith symbol occurring in T; and

Q is the number of different symbols in T.

26. A method as defined in claim 1 wherein said function is H(T:

R) and is defined as ##EQU13## where T is a test histogram;

R is a reference histogram;

H(T;

R) is the entropy of T expressed in terms of R;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i is the probability of the ith symbol occurring in R;

Q is the number of different symbols in T; and

##EQU14##

27. A method as defined in claim 1 wherein said function is H(T:

T) and is defined as ##EQU15## where T is a test histogram;

H(T;

T) is the entropy of T expressed in terms of T;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

P_i '"'"' is the probability of the ith symbol occurring in T;

Q is the number of different symbols in T; and

##EQU16##

28. A method as defined in claim 1 wherein said function is

29. A method as defined in claim 1 wherein said function is

30. A method as defined in claim 1 wherein said function is

31. A method as defined in claim 1 wherein said function is

32. A method as defined in claim 9 wherein said signal is generated when said value of said function is found to be below a predetermined value.

33. A method as defined in claim 9 wherein said signal is generated when said value is found to be at a minimum.

34. A method as defined in claim 9 wherein said signal is generated when said value is found to be at a maximum.

35. A method as defined in claim 9 wherein said signal is generated when said value is found to be above a predetermined value.

36. A method as defined in claim 1 wherein the step of measuring includes establishing test object measurement coordinates to define said values of properties in said test object, and further wherein the value of said function is assigned to locations among said test object measurement coordinates so as to produce an enhanced test object.

37. A method as defined in claim 36 including the steps of forming a digitized image of said test object;

and wherein said function is H(T;

T) and is defined as ##EQU26## wherein T is a test histogram;

H(T;

T) is the entropy of T expressed in terms of T;

b is the base of the logarithm;

m_i is the number of occurrences is the ith symbol in T;

p_i '"'"' is the probability of the ith symbol occurring in T;

Q is the number of different symbols in T;

##EQU27## and wherein said properties are (a) position in a field of view of said test object, and (b) the nature of the signals arising from said position, said comparison values are assigned to a position at the geometric center of said field of view, and said field is scanned over the original digitized image to produce an enhanced digitized image.

5. A method of recognizing a pattern in a test object, said method comprising the steps of:

specifying properties characteristic of said pattern;

specifying discrete ranges of values of said properties,

measuring the values of said properties in said test object;

determining the number of such measured values falling in each said discrete range and arranging such number in each said range in at least one test histogram;

determining a reference set of values of said properties and arranging said set as at least a first reference histogram; and

comparing said test and reference histograms by determination of the value of a function which provides a measure of the amount of information necessary to express said at least one test histogram in terms of the optimum code for describing at least said first reference histogram.

6. A method as defined in claim 5 wherein said step of determining said reference set of values comprises taking a plurality of measurements of one or more of said objects and summing said plurality of measurements to derive said reference set of values.

7. A method as defined in claim 5 wherein said measured values are arranged into a plurality of said test histograms, and said step of comparing determines which of said plurality of test histograms when compared to said at least one reference histogram provides a minimum value of said function.

8. A method as defined in claim 5 wherein said reference set of values is arranged into a plurality of said reference histograms, and said step of comparing determines which of said plurality of reference histograms when compared with said at least one test histogram provides a minimum value of said function.

38. Apparatus for recognizing a pattern in a test object, said apparatus comprising, in combination:

means for measuring in said test object the values of specified properties characteristic of said pattern,

means for arranging the measured values of said properties in at least one test histogram;

means for storing a reference set of values of said specified properties as at least a first reference histogram;

means for comparing said test and reference histograms by determination of the value of a function which provides a measure of the amount of information necessary to express said at least one test histogram in terms of the optimum code for describing at least said first reference histogram.

39. Apparatus as defined in claim 38 including means for digitizing said reference set of values and said measured values prior to aranging same in said respective reference and test histograms.

40. Apparatus according to claim 38 wherein said means for measuring comprises means for scanning said test object.

41. Apparatus according to claim 40 wherein said means for scanning is an optical scanner.

42. Apparatus according to claim 39 wherein said means for arranging said measured values comprises digital counter means.

43. Apparatus according to claim 39 wherein said means for storing said reference set comprises digital memory means.

44. Apparatus according to claim 39 wherein said means for comparing comprises digital computer means programmed to determine the value of said function.

45. Apparatus according to claim 38 wherein said function is I(T:

R) and is defined as ##EQU28## where T is a test histogram;

R is a reference histogram;

I(T;

R) is the information of T expressed in terms of R;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i is the probability of the ith symbol occurring in R; and

Q is the number of different symbols in T.

46. Apparatus according to claim 38 wherein said function is I(T:

T) and is defined as ##EQU29## where T is a test histogram;

I(T;

T) is the information of T expressed in terms of T;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i '"'"' is the probability of the ith symbol occurring in T; and

Q is the number of different symbols in T.

47. Apparatus according to claim 38 wherein said function is H(T:

R) and is defined as ##EQU30## where T is a test histogram;

R is a reference histogram;

H(T;

R) is the entropy of T expressed in terms of R;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i is the probability of the ith symbol occurring in R;

Q is the number of different symbols in T; and

##EQU31##

48. Apparatus according to claim 38 wherein said function is H(T:

T) and is defined as ##EQU32## where T is a test histogram;

H(T;

T) is the entropy of T expressed in terms of T;

b is the base of the logarithm;

m_i is the number of occurrences of the ith symbol in T;

p_i '"'"' is the probability of the ith symbol occurring in T;

Q is the number of different symbols in T; and

##EQU33##

49. Apparatus according to claim 38 wherein said function is

50. Apparatus according to claim 38 wherein said function is

51. Apparatus according to claim 38 wherein said function is

52. Apparatus according to claim 38 wherein said function is