Dynamic fuzzy logic process for identifying objects in three-dimensional data

• US 5,893,085 A
• Filed: 06/10/1997
• Issued: 04/06/1999
• Est. Priority Date: 06/10/1997
• Status: Expired due to Term

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Nettles, James L., Phillips, Ronald W.
• Primary Examiner(s): Davis, George B.
• Application Number:
  US08/872,280
• Time in Patent Office: 665 Days
• Field of Search: 706/50, 706/52, 706/900, 382/190
• US Class Current: 706/52
• CPC Class Codes: G01S 7/4802 using analysis of echo signal for target characterisation; Target signature; Target cross-section
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

A method employing fuzzy logic for identifying a detected object in LADAR data is disclosed. The method includes extracting a feature of the object; determining a confidence level in the extracted feature; comparing the extracted feature to a corresponding feature of a potential identification using a fuzzy logic process, and identifying the object. The comparison includes weighting at least one rule according to the confidence level in the extracted feature employed by the rule; shifting a membership function for the extracted feature responsive to the confidence level for the extracted feature; and executing the fuzzy logic process to obtain an indication of whether the segmented object corresponds to the potential identification.

Claims

1. A method employing fuzzy logic for identifying an object in LADAR data that has been detected and segmented, the method comprising:

extracting a feature of the object;

determining a confidence level in the extracted feature;

comparing the extracted feature to a corresponding feature of a potential identification using a fuzzy logic process including a rulebase, the comparison comprising;

weighting at least one rule in the rulebase according to the confidence level in the extracted feature employed by the rule;

shifting a membership function for the extracted feature responsive to the confidence level for the extracted feature; and

executing the fuzzy logic process to obtain an indication of whether the segmented object corresponds to the potential identification; and

identifying the object.

2. The method of claim 1, wherein extracting the feature of the object includes ascertaining a value for the extracted feature.

3. The method of claim 1, wherein the extracted feature includes at least one of the length, width, height, hull height and average height of the object.

4. The method of claim 1, wherein each rule is weighted according to n weight factors where n is an integer greater than zero.

5. The method of claim 4, wherein the weight factors include at least one of the orientation of the object with respect to the data acquisition, depression angle from the data acquisition to the object, and the resolution of the data.

6. The method of claim 4, wherein the act of weighting includes:

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

7. The method of claim 6, wherein the predetermined weight is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

8. The method of claim 1, wherein the membership functions include under, correct, and over.

9. The method of claim 1, wherein the indication of whether the segmented object corresponds to the potential identification is a likelihood.

10. The method of claim 1, wherein the object is identified by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihoods of correspondence.

11. A method for identifying an object from three-dimensional data, the method comprising:

detecting the object;

segmenting the detected object extracts environment;

extracting a feature from the segmented object;

determining a confidence level in the extracted feature; and

identifying the object using a fuzzy logic process including;

weighting at least one of a plurality of rules according to the confidence level in the extracted feature employed by the rule;

shifting a membership function for the extracted feature responsive to the confidence level in the extracted feature;

executing the fuzzy logic process to obtain an indication of the segmented object'"'"'s correspondence to a potential identification; and

determining the identity of the object from the indication of correspondence.

12. The method of claim 11, wherein extracting the feature includes ascertaining a value for the extracted feature.

13. The method of claim 11, wherein the extracted feature includes at least one of the length, width, height, hull height, and average height of the object.

14. The method of claim 11, wherein the each rule is weighted according to n weight factors where n is an integer greater than zero.

15. The method of claim 14, wherein the weight factors include at least one of the orientation of the object with respect to the data acquisition, depression angle from the data acquisition to the object, and the resolution of the data.

16. The method of claim 14, wherein the act of weighting includes:

measuring each of the n weight factors;

accessing for the extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

17. The method of claim 16, wherein the predetermined weight is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

18. The method of claim 11, wherein the membership functions include under, correct, and over.

19. The method of claim 11, wherein the indication of whether the segmented object corresponds to the potential identification is a likelihood.

20. The method of claim 11, wherein the object is identified by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihoods of correspondence.

21. A method implementing an automatic target recognition system identifying an object within a surrounding environment as one of a plurality of potential identifications, the object and the potential identifications having a plurality of predetermined features, the method comprising:

acquiring three-dimensional data regarding the object and the surrounding environment;

detecting the object from the three-dimensional data;

segmenting the detected object from the surrounding environment;

extracting a plurality of features from the segmented object by;

ascertaining values for the extracted features, and

determining a confidence level in each of the ascertained values,

comparing the object from the extracted features and a plurality of corresponding features of a plurality of potential identifications using a fuzzy logic process, the comparison including;

weighting at least one rule according to the confidence level in the value of the extracted feature employed by the rule;

shifting a membership function for at least one of the extracted features responsive to the confidence level for the each of ascertained values; and

executing the fuzzy logic process for each potential identification to obtain an indication of the segmented object'"'"'s correspondence to that potential identification;

identifying the object from the correspondence indications; and

acting on the object identification.

22. The method of claim 21, wherein the three-dimensional data is acquired from a platform that is at least one of in the air and on the ground.

23. The method of claim 21, wherein the three-dimensional data comprises pixels.

24. The method of claim 21, wherein the object that is at least one of in the air and on the ground.

25. The method of claim 21, wherein the data is ranging data.

26. The method of claim 25, wherein the ranging data is LADAR data.

27. The method of claim 21, wherein the object is detected by examining the data for indications of an identifying feature present in all potential identifications.

28. The method of claim 21, wherein the object is segmented by identifying which pixels in the LADAR data represent the object and discarding all other pixels.

29. The method of claim 21, wherein the values for the extracted features are ascertained by measuring the object'"'"'s features.

30. The method of claim 21, wherein the extracted features include at least one of the length, width, height, hull height, and average height of the object.

31. The method of claim 21, wherein the each rule is weighted according to n weight factors where n is an integer greater than zero.

32. The method of claim 31, wherein the weight factors include at least one of the orientation of the object with respect to the data acquisition, depression angle from the data acquisition to the object, and the resolution of the data.

33. The method of claim 31, wherein the act of weighting includes:

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

34. The method of claim 33, wherein the predetermined weight is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

35. The method of claim 21, wherein the membership functions include under, correct, and over.

36. The method of claim 21, wherein the indication of whether the segmented object corresponds to the potential identification is a likelihood.

37. The method of claim 21, wherein the object is identified by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihoods of correspondence.

38. The method of claim 21, wherein the object identification is acted upon by at least one of targeting the object, reporting the identification of the object, and recording the identification of the object.

39. A method implementing an automatic target recognition system, the method comprising:

acquiring three-dimensional LADAR data regarding a scanned field of view from a platform that is at least one of in the air and on the ground, wherein;

the LADAR data comprises pixels, and

the field of view includes an object that is at least one of in the air and on the ground;

detecting the object by examining the LADAR data for indications of an identifying feature present in a plurality of potential identifications;

segmenting the object from the surrounding environment within the field of view by;

identifying which pixels in the LADAR data represent the object, and

discarding all other pixels;

extracting predetermined features from the segmented object by;

measuring the predetermined features, and

determining the confidence in the accuracy of the measurements,

wherein the predetermined features include the length, width, height, hull height, and average height of the object;

identifying the object by comparing the extracted features with previously stored features of potential identifications using a fuzzy logic process, the fuzzy logic process including;

weighting each rule according to the confidence in the accuracy of the measurement of the extracted feature employed by the rule and according to n weight factors, the weight factors including object orientation with respect to the platform, depression angle from the LADAR acquisition to the object, and the pixel resolution of the LADAR data, the act of weighting including;

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight by at least one of the following;

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight; and

assigning the predetermined weight to the respective rule;

shifting a membership function for at least one of the features responsive to the confidence in the feature measurements, the membership functions including under, correct, and over;

executing the fuzzy logic process for each potential identification to obtain an indication of the likelihood of whether the segmented object corresponds to that potential identification; and

identifying the object by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a predetermined number of potential identifications with the highest likelihoods of correspondence; and

acting on the object identification by at least one of targeting the object, reporting the identification of the object, and recording the identification of the object.

40. A program storage device readable by a computer, the program storage device encoding statements for performing a method employing fuzzy logic for identifying an object in LADAR data that has been detected and segmented, the method comprising:

extracting a feature of the object;

determining a confidence level in the extracted feature;

comparing the extracted feature to a corresponding feature of a potential identification using a fuzzy logic process including a rulebase, the comparison comprising;

weighting at least one rule in the rulebase according to the confidence level in the extracted feature employed by the rule;

shifting a membership function for the extracted feature responsive to the confidence level for the extracted feature; and

executing the fuzzy logic process to obtain an indication of whether the segmented object corresponds to the potential identification; and

identifying the object.

41. The program storage device claim 40, wherein extracting the feature of the object in the programmed method includes ascertaining a value for the extracted feature.

42. The program storage device of claim 40, wherein the extracted feature in the programmed method includes at least one of the length, width, height, hull height, and average height of the object.

43. The program storage device of claim 40, wherein each rule in the programmed method is weighted according to n weight factors where n is an integer greater than one.

44. The program storage device of claim 43, wherein the weight factors in the programmed method include at least one of the orientation of the object with respect to the data acquisition, depression angle from the data acquisition to the object, and the resolution of the data.

45. The program storage device of claim 43, wherein the act of weighting in the programmed method includes:

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

46. The program storage device of claim 45, wherein the predetermined weight in the programmed method is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

47. The program storage device of claim 40, wherein the membership functions in the programmed method include under, correct, and over.

48. The program storage device of claim 40, wherein the indication of whether the segmented object corresponds to the potential identification in the programmed method is a likelihood.

49. The program storage device of claim 40, wherein the object is identified in the programmed method by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihoods of correspondence.

50. A program storage device readable by a computer, the program storage device encoding statements for performing a method for identifying an object from three-dimensional data, the method comprising:

detecting the object;

segmenting the detected object from its environment;

extracting a feature from the segmented object;

determining a confidence level in the extracted feature; and

identifying the object using a fuzzy logic process including;

weighting at least one of a plurality of rules according to the confidence level in the extracted feature employed by the rule;

shifting a membership function for the extracted feature responsive to the confidence level for the extracted feature;

executing the fuzzy logic process to obtain an indication of the segmented object'"'"'s correspondence to potential identification; and

determining the identity of the object from the indication of correspondence.

51. The program storage device of claim 50, wherein extracting the feature in the programmed method includes ascertaining a value for the extracted feature.

52. The program storage device of claim 50, wherein the extracted feature in the programmed method includes at least one of the length, width, height, hull height, and average height of the object.

53. The program storage device of claim 50, wherein each rule in the programmed method is weighted according to n weight factors where n is an integer greater than zero.

54. The program storage device of claim 53, wherein the weight factors in the programmed method include at least one of the orientation of the data acquisition with respect to the object, depression angle of the data acquisition with respect to the object, and the resolution of the data.

55. The program storage device of claim 53, wherein the act of weighting in the programmed method includes:

measuring each of the n weight factors;

accessing for the extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

56. The program storage device of claim 55, wherein the predetermined weight in the programmed method is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

57. The program storage device of claim 50, wherein the membership functions in the programmed method include under, correct, and over.

58. The program storage device of claim 50, wherein the indication of whether the segmented object corresponds to the potential identification in the programmed method is a likelihood.

59. The program storage device of claim 50, wherein the object is identified in the programmed method by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihoods of correspondence.

60. A program storage device readable by a computer, the program storage device encoding statements for performing a method implementing an automatic target recognition system identifying an object within a surrounding environment as one of a plurality of potential identifications, the object and the potential identifications having a plurality of predetermined features, the program storage device readable by a computer, the program storage device encoding statements for performing a method comprising:

acquiring three-dimensional data regarding the object and the surrounding environment;

detecting the object from the three-dimensional data;

segmenting the detected object from the surrounding environment;

extracting a plurality of features from the segmented object by;

ascertaining values for the extracted features, and

determining a confidence level in each of the ascertained values,

comparing the object from the extracted features and a plurality of corresponding features of a plurality of potential identifications using a fuzzy logic process, the comparison including;

weighting at least one rule according to the confidence level in the value of the extracted feature employed by the rule;

shifting a membership function for at least one of the extracted features responsive to the confidence level for each of the ascertained values; and

executing the fuzzy logic process for each potential identification an indication of the segmented object'"'"'s correspondence to that potential identification;

identifying the object from the correspondence indications; and

acting on the object identification.

61. The program storage device of claim 60, wherein the three-dimensional data in the programmed method is acquired from a platform that is at least one of in the air and on the ground.

62. The program storage device of claim 60, wherein the three-dimensional data in the programmed method comprises pixels.

63. The program storage device of claim 60, wherein the object in the programmed method that is at least one of in the air and on the ground.

64. The program storage device of claim 60, wherein the data in the programmed method is ranging data.

65. The program storage device of claim 64, wherein the ranging data in the programmed method is LADAR data.

66. The program storage device of claim 60, wherein the object is detected in the programmed method by examining the data for indications of an identifying feature present in all potential identifications.

67. The program storage device of claim 60, wherein the object is segmented in the programmed method by identifying which pixels in the LADAR data represent the object and discarding all other pixels.

68. The program storage device of claim 60, wherein the values for the extracted features in the programmed method are ascertained by measuring the object'"'"'s features.

69. The program storage device of claim 60, wherein the extracted features in the programmed method include at least one of the length, width, height, hull height, and average height of the object.

70. The program storage device of claim 60, wherein each rule in the programmed method is weighted according to n weight factors where n is an integer greater than zero.

71. The program storage device of claim 70, wherein the weight factors in the programmed method include at least one of the orientation of the object with respect to the data acquisition, depression angle from the data acquisition to the object, and the resolution of the data.

72. The program storage device of claim 70, wherein the act of weighting in the programmed method includes:

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight from the look-up table; and

assigning the predetermined weight to the respective rule.

73. The program storage device of claim 72, wherein the predetermined weight in the programmed method is ascertained by at least one of the following:

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight.

74. The program storage device of claim 60, wherein the membership functions in the programmed method include under, correct, and over.

75. The program storage device of claim 60, wherein the indication of whether the segmented object corresponds to the potential identification in the programmed method is a likelihood.

76. The program storage device of claim 60, wherein the object is identified in the programmed method by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a plurality of potential identifications with the highest likelihood of correspondence.

77. The program storage device of claim 60, wherein the object identification in the programmed method is acted upon by at least one of targeting the object, reporting the identification of the object, and recording the identification of the object.

78. A program storage device readable by a computer, the program storage device encoding statements for performing a method implementing an automatic target recognition system, the program storage device readable by a computer, the program storage device encoding statements for performing a method comprising:

acquiring three-dimensional LADAR data regarding a scanned field of view from a platform that is at least one of in the air and on the ground, wherein;

the LADAR data comprises pixels, and

the field of view includes an object that is at least one of in the air and on the ground;

detecting the object by examining the LADAR data for indications of an identifying feature present in a plurality of potential identifications;

segmenting the object from the surrounding environment within the field of view by;

identifying which pixels in the LADAR data represent the object, and

discarding all other pixels;

extracting predetermined features from the segmented object by;

measuring the predetermined features, and

determining the confidence in the accuracy of the measurements,

wherein the predetermined features include the length, width, height, hull height, and average height of the object;

identifying the object by comparing the extracted features with previously stored features of potential identifications using a fuzzy logic process, the fuzzy logic process including;

weighting each rule according to the confidence in the accuracy of the measurements of the extracted feature employed by the rule and according to n weight factors, the weight factors including object orientation with respect to the platform, depression angle from the LADAR acquisition to the object, and the pixel resolution of the LADAR data, the act of weighting including;

measuring each of the n weight factors;

accessing for the respective extracted feature an n-dimensional look-up table indexed by each of the weight factors;

ascertaining a predetermined weight by at least one of the following;

truncating each of the measured weight factors to the nearest table index for the respective weight factor and retrieving the predetermined weight from the corresponding location in the n-dimensional look-up table; and

interpolating from the table indices corresponding to the measured weight factors and the corresponding weight factors in the look-up table to obtain the predetermined weight; and

assigning the predetermined weight to the respective rule;

shifting a membership function for at least one of the features responsive to the confidence in the feature measurements, the membership functions including under, correct, and over;

executing the fuzzy logic process for each potential identification to obtain an indication of the likelihood of whether the segmented object corresponds to that potential identification; and

identifying the object by at least one of selecting the potential identification with the highest likelihood of correspondence and performing three-dimensional model matching on a predetermined number of potential identifications with the highest likelihoods of correspondence; and

acting on the object identification by at least one of targeting the object, reporting the identification of the object, and recording the identification of the object.