Classification system for radar and sonar applications

• US 20070024494A1
• Filed: 04/10/2006
• Published: 02/01/2007
• Est. Priority Date: 04/11/2005
• Status: Active Grant

• Current Assignee: Raytheon Canada Limited (Rtx Corporation)
• Original Assignee: Raytheon Canada Limited (Rtx Corporation)
• Inventors: Dizaji, Reza, Ghadaki, Hamid
• Granted Patent: US 7,567,203 B2
• Time in Patent Office: Days
• US Class Current: 342/90
• CPC Class Codes: G01S 13/723 by using numerical data
• CPC Class Codes: G01S 13/726 Multiple target tracking
• CPC Class Codes: G01S 13/872 Combinations of primary radar and secondary radar
• CPC Class Codes: G01S 13/91 for traffic control G01S13/93 takes precedence
• CPC Class Codes: G01S 13/951 ground based
• CPC Class Codes: G01S 2013/916 Airport surface monitoring [ASDE]
• CPC Class Codes: G01S 7/412 based on a comparison between measured values and known or stored values
• CPC Class Codes: G01S 7/415 Identification of targets based on measurements of movement associated with the target
• CPC Class Codes: Y02A 90/10 Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

A system and method for target classification for an aircraft surveillance radar is provided. In one implementation, the track classifier provides tracks with an updated probability value based on its likelihood to conform to aircraft and non-aircraft target behavior. The track classifier identifies false tracks that may arise from weather and biological targets, and can detect aircrafts lacking Secondary Surveillance Radar (SSR) data. Various features and combinations of features are evaluated using a proposed clustering performance index (CPI) and used to discriminate between aircrafts and false tracks.

Claims

1. A classifier for classifying a given radar track segment obtained from a radar system, the radar system having a primary surveillance radar for providing primary radar data and a secondary surveillance radar for providing secondary radar data, wherein the classifier comprises:

a) a pre-processing stage, the preprocessing stage forms the given radar track segment and generates principal data based on the primary radar data or a combination of secondary and primary radar data, and extension data based on the primary radar data;

b) a feature extraction stage connected to the pre-processing stage, the feature extraction stage processes-at least one of the primary and secondary radar data associated with the given radar track segment to provide a plurality of feature values;

c) a classification stage connected to the feature extraction stage, the classification stage generates a principal classification result and an extension classification result for the given radar track segment based on at least a portion of the feature values;

or the classification stage generates a combined classification result for combined principal and extension feature values; and

,

d) a combiner stage connected to the classification stage, the combiner stage combines the extension and principal classification results to provide a classification result for the given radar track segment when the classification stage provides the principal and extension classification results.

2. The classifier of claim 1, wherein the combiner stage further combines the combined classification result or the extension and principal classification results with the classification result of at least one previous radar track segment associated with the given radar track segment to provide a classification result for the given radar track segment.

3. The classifier of claim 1, wherein the classification stage includes:

a) a principal feature classifier path connected to the feature extraction stage, the principal feature classifier path generates the principal classification result; and

,

b) an extension feature classifier path connected to the feature extraction stage, the extension feature classifier path generates the extension classification result.

4. The classifier of claim 3, wherein the extension feature classifier path includes:

a) an extension feature processing stage connected to the feature extraction stage, the extension feature processing stage receives the plurality of feature values based on the extension data for the given radar track segment to generate an extension feature vector wherein each entry in the extension feature vector is calculated from either the given radar track segment or the given radar track segment and associated radar track segments, and post-processes the extension feature vector; and

,

b) an extension feature classifier stage connected to the extension feature processing stage, the extension feature classifier stage classifies the post-processed extension feature vector to provide the extension classification result.

5. The classifier of claim 3, wherein the principal feature classifier path includes:

a) a principal feature processing stage connected to the feature extraction stage, the principal feature processing stage receives the plurality of feature values based on the principal data for the given radar track segment to generate a principal feature vector wherein each entry in the principal feature vector is calculated from either the given radar track segment or the given radar track segment and associated radar track segments, and post-processes the principal feature vector; and

b) a principal feature classifier stage connected to the principal feature processing stage, the principal feature classifier stage classifies post-processed principal feature vector to provide the principal classification result.

6. The classifier of claim 5, wherein at least one of the extension classifier stage and the principal classifier stage employ a machine learning technique performing classification.

7. The classifier of claim 6, wherein the machine leaning technique includes one of:

a linear Support Vector machine;

or a non-linear Support Vector machine.

8. The classifier of claim 1, wherein the pre-processing stage is connected to the combiner stage for providing an indication of whether secondary radar data is associated with the given radar track segment, wherein the indication is used to forego the feature extraction and classification stages and classify the given track segment as being indicative of an aircraft.

9. The classifier of claim 1, wherein the pre-processing stage generates the given radar track segment to overlap at least one of the previous related track segments.

10. The classifier of claim 5, wherein at least one of the feature processing stages generates one of the feature vectors based upon at least one of:

(a) the given radar track segment and overlapping associated radar track signals;

or (b) repeating a portion of the feature values for the given radar track segment.

11. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a variance in the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line.

12. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a variance in the first difference of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line.

13. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a variance in the first difference of the slope of the given radar track segment.

14. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a mean of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line.

15. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a mean of the first difference of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line.

16. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes calculating a mean of the first difference of the slope of the given radar track segment.

17. The classifier of claim 1, wherein one of the features calculated by the feature extraction stage includes at least one of:

(1) calculating a variance in the second difference of the speed (jerk) of the target associated with the given radar track segment;

(2) calculating a mean of the second difference of the speed (jerk) of the target associated with the given radar track segment;

(3) calculating a variance in the first difference of the speed (acceleration) of the target associated with the given radar track segment;

(4) calculating the total distance covered by the given radar track segment;

(5) calculating a mean of the first difference of the speed (acceleration) of the target associated with the given radar track segment;

(6) calculating the mean speed of the target associated with the given radar track segment;

(7) calculating the mean number of radar scans between successive plots used to generate the given radar track segment;

(8) calculating the mean range of the target associated with the given radar track segment;

(9) calculating a mean of the range-compensated mean amplitude of the data points in plots associated with the given radar track segment;

(10) calculating a mean of the range-compensated peak amplitude of the data points in plots associated with the given radar track segment;

(11) calculating a mean difference of the range-compensated peak and mean amplitudes of the data points in plots associated with the given radar track segment;

or

(12) calculating a mean of the total number of detection points in plots associated with the given radar track segment.

18. The classifier of claim 1, wherein the classifier is connected to a track generator of the radar system to receive a given radar track and the pre-processing stage segments the given radar track to provide the given radar track segment and associated radar track segments.

19. The classifier of claim 1, wherein the classifier is connected to a plot extractor of the radar system to receive a plurality of detections from a series of plots, the plurality of detections being associated with a given target and the pre-processing stage forms the given radar track segment and associated radar track segments from the plurality of detections.

20. The classifier of claim 1, wherein the primary radar data is also used for training and testing the feature extraction stage, the classification stage and the combiner stage.

21. The classifier of claim 1, wherein the secondary radar data is also used for training and testing the classification stage and combiner stage.

22. A method for classifying a given radar track segment obtained from a radar system, the radar system having a primary surveillance radar for providing primary radar data and a secondary surveillance radar for providing secondary radar data, wherein the method comprises:

a) forming the given radar track segment and generating principal data based on the primary and secondary radar data, and extension data based on the primary radar data;

b) processing at least one of the primary and secondary radar data associated with the given radar track segment and a portion of a previous associated radar track segment to provide a plurality of feature values;

c) generating either a principal classification result and an extension classification result or a combined classification result for combined principal and extension feature values for the given radar track segment based on at least a portion of the feature values; and

,

d) combining the extension and principal classification results to provide a classification result for the given radar track segment when the principal and extension classification results are generated.

23. The method of claim 22, wherein step (d) includes further combining the combined classification result or the extension and principal classification results with the classification result of at least one previous radar track segment associated with the given radar track segment to provide a classification result for the given radar track segment.

24. The method of claim 22, wherein step (c) includes

e) receiving the plurality of feature values based on the extension data for the given radar track segment;

f) generating an extension feature vector wherein each entry in the extension feature vector is calculated from either the given radar track segment or the given radar track segment and associated radar track segments;

g) post-processing the extension feature vector; and

,

h) classifying the post-processed extension feature vector to provide the extension classification result.

25. The method of claim 24, wherein step (c) further includes:

i) receiving the plurality of feature values based on the principal data for the given radar track segment;

j) generating a principal feature vector wherein each entry in the principal feature vector is calculated from either the given radar track segment or the given radar track segment and associated radar track segments;

k) post-processing the principal feature vector; and

,

l) classifying the post-processed principal feature vector to provide the principal classification result.

26. The method of claim 25, wherein the method further includes using machine learning techniques for performing classification in at least one of steps (h) and (I).

27. The method of claim 26, wherein the method includes using at least one of:

(1) a linear Support Vector Machine for the machine learning technique;

or (2) a non-linear Support Vector Machine for the machine learning technique.

28. The method of claim 22, wherein step (b) includes forming the given radar track segment such that the given track segment overlaps at least one of the previous related track segments.

29. The method of claim 24, wherein at least one of steps (f) and (j) include generating one of the feature vectors based on the given radar track segment and overlapping associated radar track signals.

30. The method of claim 24, wherein at least one of steps (f) and (j) include generating one of the feature vectors based on repeating a portion of the feature values for the given radar track segment.

31. The method of claim 22, wherein step (b) includes one of:

calculating one of the features based on a variance in the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line;

calculating one of the features based on a variance in the first difference of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line;

calculating one of the features based on a variance in the first difference of the slope of the given radar track segment;

calculating one of the features based on a mean of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line;

calculating one of the features based on a mean of the first difference of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line;

calculating one of the features based on a mean of the first difference of the slope of the given radar track segment;

calculating one of the features based on a variance in the second difference of the speed (jerk) of the target associated with the given radar track segment;

calculating one of the features based on a variance in the first difference of the speed (acceleration) of the target associated with the given radar track segment;

calculating one of the features based on the total distance covered by the given radar track segment;

calculating one of the features based on a mean of the second difference of the speed (jerk) of the target associated with the given radar track segment;

calculating one of the features based on a mean of the first difference of the speed (acceleration) of the target associated with the given radar track segment;

calculating one of the features based on a mean speed of the target associated with the given radar track segment;

calculating one of the features based on the mean number of radar scans between successive plots used to generate the given radar track segment;

calculating one of the features based on the mean range of the target associated with the given radar track segment;

calculating one of the features based on a mean of the range-compensated mean amplitude of the data points in plots associated with the given radar track segment;

calculating one of the features based on a mean of the range-compensated peak amplitude of the data points in plots associated with the given radar track segment;

calculating one of the features based on a mean difference of the range-compensated peak and mean amplitudes of the data points in plots associated with the given radar track segment;

or

calculating one of the features based on a mean of the total number of detection points in plots associated with the given radar track segment.

32. The method of claim 22, wherein the method includes receiving a given radar track from a track generator of the radar system and step (a) includes segmenting the given radar track to provide the given radar track segment and associated radar track segments.

33. The method of claim 22, wherein the method includes assessing the features according to:

a) calculating a plurality of feature values for several of the features based on a plurality of training radar track segments;

b) partitioning the plurality of feature values calculated for at least one feature into a plurality of classes;

c) randomly picking classified points for each class and calculating the number of mis-classified radar track segments; and

,

d) computing a performance index based on the number of mis-classified radar track segments for assessing either one of the features or a combination of the features.