Method and apparatus for classifying targets

• US 5,337,053 A
• Filed: 10/22/1993
• Issued: 08/09/1994
• Est. Priority Date: 10/22/1993
• Status: Expired due to Fees

• Current Assignee: the united states of america as represented by the secretary of the navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Dwyer, Roger F.
• Primary Examiner(s): Sotomayor, John B.
• Application Number:
  US08/147,238
• Time in Patent Office: 291 Days
• Field of Search: 342/90, 342/22, 342/52, 342/192, 342/193, 342/195, 342/194, 367/99, 367/904
• US Class Current: 342/90
• CPC Class Codes: G01S 7/415 Identification of targets based on measurements of movement associated with the target
• CPC Class Codes: G01S 7/539 using analysis of echo signal for target characterisation; Target signature; Target cross-section
• CPC Class Codes: G06V 20/64 Three-dimensional objects
• CPC Class Codes: Y10S 367/904 Doppler compensation systems

Abstract

A method and apparatus for classifying objects based upon information conned in the fourth order cumulant derived from energy from the object. Successive pulse returns in active systems and samples in passive systems are converted into vectors based upon the fourth order cumulant information for each successive return or sample. The vectors for each such pulse are compared to corresponding class information based upon the corresponding fourth order cumulant information. The comparison uses loglikelihood ratios of the different pairs of classes. Decision making is based upon the value of each loglikelihood ratio in comparison with a class threshold.

Claims

1. Apparatus for cataloging physical objects into one of a plurality of predetermined classes based upon energy received from the object, said apparatus comprising:

means for receiving the energy from the object;

means connected to said receiving means for generating an object vector based upon a fourth order cumulant function for the received energy;

means for storing class information based upon the fourth order cumulant functions for each of the plurality of predetermined classes; and

means for classifying the object into one of the predetermined classes based upon the class information.

2. Apparatus as recited in claim 1 wherein said generating means includes means for generating a fourth order cumulant object vector for the received energy and wherein said comparing means compares the fourth order cumulant object vector with the class information based upon fourth order cumulant function information for each class.

4. Apparatus as recited in claim 1 wherein said generating means includes means for generating a fourth order cumulant frequency spectrum object vector and wherein said comparing means compares the fourth order cumulant frequency spectrum object vector with class information based upon the fourth order cumulant frequency spectrum for each class.

5. Apparatus as recited in claim 4 wherein said means for generating the fourth order cumulant frequency spectrum object vector includes:

means for generating a fourth order cumulant object vector according to;

##EQU14## where s* represents a complex conjugate, and j₁ and j₂, and j₃ are time delays; and

means for generating a fourth order cumulant frequency spectrum object vector X(1) according to;

##EQU15##

6. Apparatus as recited in claim 1 wherein the energy from the object is received as a succession of energy pulses, wherein said generating means produces an object vector for each energy pulse and wherein said classifying means analyzes the object vectors in sequence.

7. Apparatus as recited in claim 6 wherein said classifying means includes decision means for classifying each energy pulse by means of a loglikelihood ratio using the object vector and corresponding information from said storing means for each of two of the plurality of predetermined classes.

8. Apparatus as recited in claim 6 wherein said classifying means includes decision means for classifying each energy pulse by means of a loglikelihood ratio using the object vector and, in sequence, corresponding information from said storing means for each possible pair of the plurality of predetermined classes.

9. Apparatus as recited in claim 8 wherein said decision means includes means for establishing, for each pair of classes, a pair of thresholds that determine whether each energy pulse is classified into one of the pair of classes or is unclassified.

3. Apparatus as recited in claim 3 wherein said means for generating the fourth order cumulant object vector operates according to:

##EQU13## where s* represents a complex conjugate, and j₁ and j₂, and j₃ are time delays.

10. Apparatus for classifying a physical object located in a medium into one of a predetermined number of classes, said apparatus comprising:

means for transmitting at least one energy pulse toward the object;

means for receiving an echo from the object in response to a transmitted energy pulse;

means connected to said receiving means for processing each received echo to obtain a fourth order cumulant frequency spectrum vector corresponding to the received echo;

means for storing information for each of the predetermined number of classes based upon the fourth order cumulant frequency spectrum for each predetermined class;

means connected to said processing means and said storing means for classifying each energy pulse into one of three states; and

means connected to said pulse classifying means for producing the classification for each object based upon echoes from the object.

11. Apparatus as recited in claim 10 wherein said processing means includes a fourth order cumulant generator, a fourth order cumulant frequency spectrum generator and means for converting the fourth order cumulant frequency spectrum to as fourth order cumulant frequency spectrum vector for each echo.

12. Apparatus as recited in claim 10 wherein said processing means includes:

a fourth order cumulant generator for producing a fourth order cumulant signal, Λ

_pq (X(n)), for each pulse "n" according to;

##EQU16## a fourth order cumulant frequency spectrum generator for producing a fourth order cumulant frequency spectrum signal for each pulse according to;

##EQU17## a vector generator for converting the fourth order cumulant frequency spectrum signal into a vector having the form;

space="preserve" listing-type="equation">X(1)=[X.sub.1,1 (1), X.sub.1,2 (1), . . . , X.sub.64,64 (1)].

13. Apparatus as recited in claim 10 wherein said pulse classifying means includes:

loglikelihood ratio generating means connected to said vector generating means and said storing means for generating a loglikelihood ratio for each vector with respect to a pair of classes;

threshold means for determining whether each energy pulse is to be classified into one or none of the classes in the pair; and

means for recording the classification of a pulse into one of a pair of classes.

14. Apparatus as recited in claim 13 wherein said storing means stores information for "class(p)" and "class(q)" classes in one pair of classes including A_p and A_q mean vectors for a given echo pulse and covariance matrices M_p and M_q that depend upon a given vector for an echo and a corresponding means vector and wherein said loglikelihood ratio generating means produces a loglikelihood ratio signal Λ

_pq (X(1)) for a first pulse according to;

##EQU18##

15. Apparatus as recited in claim 14 wherein said loglikelihood ratio generating means produces a loglikelihood ratio, Λ

_pq (X(1)), for a successive pulse "n" according to;

##EQU19## where N_p, N_q, C_p, C_q, B_p and B_q are cross covariances in said storing means based upon the covariance matrices.

16. Apparatus as recited in claim 13 wherein said threshold means establishes, for a given pair of classes, first and second thresholds and enables said recording means to record a classification when the loglikelihood ratio exceeds one of the two thresholds, said loglikelihood ratio generating means and said threshold means providing an analysis of each echo for each possible pair of classes.

17. Apparatus as recited in claim 16 wherein said loglikelihood ratio generating means generates a Λ

_pq signal for a given pair, "p" and "q", of classes and said threshold means operates according to;

space="preserve" listing-type="equation">IF Λ

.sub.pq (X(η

))≧

a THEN set class(p)=class(p)+1

and

space="preserve" listing-type="equation">IF Λ

.sub.pq (X(η

))≦

b THEN set class(q)=class(q)+1

where "a" and "b" are the thresholds for classes "p" and "q" respectively and said recording means includes a class(p) recording means and a class(q) recording means for being incremented when Λ

_pq exceeds "a" and "b" respectively.

18. A method for classifying an object located in a medium into one of a plurality of classes comprising the steps of receiving one of a plurality of energy samples from the object through the medium and for each sample, in succession:

generating a fourth order cumulant vector corresponding to an energy sample;

comparing the fourth order cumulant vector with class information derived from the fourth order cumulant information for each of the classes taken two at a time;

testing the result of each comparison for each non-redundant pair of classes to determine pulse classification; and

accumulating the testing results for each of the energy samples to classify the object.

19. A method as recited in claim 18 wherein said step for generating a fourth order cumulant vector includes:

generating a fourth order cumulant object vector according to;

##EQU20## where s* represents a complex conjugate, and j₁ and j₂, and j₃ are time delays;

generating a fourth order cumulant frequency spectrum object vector according to;

##EQU21## and converting the fourth order cumulant frequency spectrum signal into a vector X(1) having the form;

space="preserve" listing-type="equation">X(1)=[X.sub.1,1 (1), X.sub.1,2 (1), . . . , X.sub.64,64 (1)].

20. A method as recited in claim 18 wherein the class information derived from the fourth order cumulant information includes "class(p)" and "class(q)" information for each class in a pair of classes including A_p and A_q mean vectors and covariance matrices M_p and M_q that depend upon information for each predetermined class of objects and wherein the step of comparing the fourth order cumulant vector with the class information includes the step of obtaining a loglikelihood ratio Λ

_pq (X(1)) for a first pulse according to;

##EQU22##

21. A method as recited in claim 18 wherein the step of obtaining a loglikelihood ratio includes the step of obtaining a loglikelihood ratio, Λ

_pq (X(n)), for a successive pulse "n" according to;

##EQU23## where N_p, N_q, C_p, C_q, B_p and B_q are stored cross covariances based upon the covariance matrices.

22. A method as recited in claim 18 wherein said testing step includes:

comparing each Λ

_pq loglikelihood signal for a given pair, "p" and "q", of classes with thresholds "a" and "b" for classes "p" and "q" respectively according to;

space="preserve" listing-type="equation">IF Λ

.sub.pq (X(η

))≧

a THEN set class(p)=class(p)+1

and

space="preserve" listing-type="equation">IF Λ

.sub.pq (X(η

))≦

b THEN set class(q)=class(q)+1

and

incrementing class(p) and class(q) recording means each time Λ

_pq exceeds "a" and "b" respectively.