Classification system for radar and sonar applications
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G01S-007/41
G01S-007/02
G01S-013/87
G01S-013/00
출원번호
UP-0401097
(2006-04-10)
등록번호
US-7567203
(2009-08-05)
발명자
/ 주소
Dizaji, Reza M.
Ghadaki, Hamid
출원인 / 주소
Raytheon Canada Limited
대리인 / 주소
Daly, Crowley, Mofford & Durkee, LLP
인용정보
피인용 횟수 :
9인용 특허 :
20
초록▼
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
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.
대표청구항▼
The invention claimed is: 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 com
The invention claimed is: 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 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; (12) calculating a mean of the total number of detection points in plots associated with the given radar track segment; (13) calculating a variance in the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line; (14) 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; (15) calculating a variance in the first difference of the slope of the given radar track segment; (16) calculating a mean of the displacement of the path of the given radar track segment from a polynomial least-squares best-fit line; (17) 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; or (18) calculating a mean of the first difference of the slope of the given radar track segment. 3. 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. 4. The method of claim 3, 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. 5. The method of claim 3, 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. 6. A classifier comprising: a) a pre-processing stage adapted to receive raw track data and to segment the raw track data to provide a plurality of track segment data for a given track; b) a feature extraction stage coupled to receive the track segment data from said pre-processing stage and to operate upon the track segment data to provide a plurality of feature values; and c) a classification stage coupled to operate upon the plurality of feature values provided by said feature extraction stage to provide a classification result. 7. The classifier of claim 6 adapted to classify a target obtained from a system having a sensor subsystem which provides primary data and optionally secondary data to the classifier. 8. A classifier comprising: a) a pre-processing stage adapted to receive raw track data and to segment the raw track data to provide a plurality of track segment data for a given track; b) a feature extraction stage coupled to receive the track segment data from said pre-processing stage and to operate upon the track segment data to provide a plurality of feature values; and c) a classification stage coupled to operate upon the plurality of feature values provided by said feature extraction stage to provide a classification result wherein the classifier is adapted to classify a target obtained from a system having a sensor subsystem which provides primary data and optionally secondary data to the classifier and wherein the sensor subsystem includes a transponder subsystem for providing the optional secondary data. 9. The classifier of claim 7 wherein: the preprocessing stage forms a track segment for the given track and generates principal data based upon the primary data or a combination of secondary and primary data, and extension data based upon the primary data; the feature extraction stage processes at least one of the primary and secondary data associated with the given track to provide a plurality of feature values; and the classification stage generates a principal classification result and an extension classification result for the track segment of the given track based upon at least a portion of the feature values or the classification stage generates a combined classification result based upon at least a portion of the feature values. 10. The classifier of claim 9 further comprising a combiner stage coupled to said classification stage wherein the combiner stage combines the extension and principal classification results to provide a classification result for the track segment when the classification stage provides the principal and extension classification results. 11. The classifier of claim 10 wherein the sensor subsystem comprises a radar system having a primary surveillance radar and a secondary surveillance radar and wherein the primary data corresponds to primary radar data provided by the primary surveillance radar and the secondary data corresponds to secondary radar data provided by the secondary surveillance radar and each track segment corresponds to a radar track segment. 12. The classifier of claim 11, 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. 13. The classifier of claim 12, wherein the classification stage includes: a) a principal feature classifier path coupled to the feature extraction stage, the principal feature classifier path generates the principal classification result; and, b) an extension feature classifier path coupled to the feature extraction stage, the extension feature classifier path generates the extension classification result. 14. The classifier of claim 13, wherein the extension feature classifier path includes: a) an extension feature processing stage coupled to the feature extraction stage, the extension feature processing stage receives the plurality of feature values based upon 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 to determine characteristics that should be provided to a classifier; and, b) an extension feature classifier stage coupled to the extension feature processing stage, the extension feature classifier stage classifies the post-processed extension feature vector to provide the extension classification result. 15. The classifier of claim 13, wherein the principal feature classifier path includes: a) a principal feature processing stage coupled to the feature extraction stage, the principal feature processing stage receives the plurality of feature values based upon 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 coupled to the principal feature processing stage, the principal feature classifier stage classifies post-processed principal feature vector to provide the principal classification result. 16. The classifier of claim 13, wherein at least one of the extension feature classifier path and the principal feature classifier path employ a machine learning technique for performing classification. 17. The classifier of claim 16, wherein the machine learning technique for performing classification includes at least one of: (a) a linear Support Vector machine; and (b) a non-linear Support Vector machine. 18. The classifier of claim 11, wherein the pre-processing stage is coupled 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. 19. The classifier of claim 6, wherein the pre-processing stage generates the track segment data for the given track segment to overlap at least one previously related track segment. 20. The classifier of claim 15, wherein at least one of the feature processing stages generates one of the feature vectors based upon at least one of: (a) the track segment for the given track and overlapping associated radar track segments; and (b) a portion of the feature values for the track segment for the given radar track. 21. The classifier of claim 11, wherein the classifier is coupled 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. 22. The classifier of claim 11, wherein: the classifier is coupled 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. 23. The classifier of claim 7, wherein at least one of the primary radar data and the secondary radar data are used for training and testing at least one of: (1) the feature extraction stage; (2) the classification stage; and (3) the combiner stage. 24. A method for classifying a given observation obtained from a sensor system, the sensor system having a sensor subsystem for providing primary data and optionally a transponder system for providing secondary data, wherein the method comprises: a) receiving raw track data; b) segmenting the raw track data to provide a plurality of track segment data for a given track; c) operating on the track segment data to provide a plurality of feature values; and d) operating on the plurality of feature values to provide a classification result for the given track segment. 25. A method for classifying a given observation obtained from a sensor subsystem, the sensor system having a sensor subsystem for providing primary data and optionally a transponder system for providing secondary data, wherein the method comprises: a) receiving new track data; b) segmenting the raw track data to provide a plurality of tract segment data for a given track; c) operating on the track segment data to provide a plurality of feature values; and d) operating on the plurality of feature values to provide a classification result for the given track segment wherein segmenting the raw track data to provide a plurality of track segment data for a given track includes: (1) generating principal data based upon the primary and secondary data; and (2) generating extension data based upon the primary data. 26. The method of claim 25 wherein operating on the plurality of feature values to provide a classification result for the given track segment includes: generating a principal classification result; generating an extension classification result; combining the extension classification result and the principal classification result to provide a classification result for the given track when the principal and extension classification results are generated and the method further comprises: combining one of the combined classification result or the extension and principal classification results with the classification result of at least one previous track segment associated with the given track segment to provide a classification result for the given track segment. 27. The method of claim 26, wherein (d) includes: 1) receiving a plurality of feature values based upon the extension data for the given track segment; 2) generating an extension feature vector using at least a plurality of feature values based upon the extension data wherein each entry in the extension feature vector is calculated from either the given track segment or the given track segment and associated track segments; 3) post-processing the extension feature vector to determine characteristics that should be provided to a classifier; and 4) classifying the post-processed extension feature vector to provide the extension classification result. 28. The method of claim 27 wherein step (d) further includes: 5) receiving a plurality of feature values based upon the principal data for the given track segment; 6) generating a principal feature vector using at least the plurality of feature values based upon the extension data wherein each entry in the principal feature vector is calculated from either the given track segment or the given track segment and associated track segments; 7) post-processing the principal feature vector to determine characteristics that should be provided to a classifier; and 8) classifying the post-processed principal feature vector to provide the principal classification result. 29. The method of claim 28, wherein the sensor system is a radar system having a primary surveillance radar and a secondary surveillance radar and wherein the primary data corresponds to primary radar data and the secondary radar corresponds to secondary radar data and wherein (a) includes providing an indication of whether secondary radar data is associated with the given radar track segment, wherein the indication is used to forego processing and classification performed by (b)-(c) and classifying the given radar track segment as being indicative of an aircraft in (d). 30. The method of claim 29, wherein (b) includes forming the given track segment such that the given track segment overlaps at least one previous related track segment. 31. The method of claim 30, wherein at least one of (d)(2) and (d)(6) include at least one of: a) generating the feature vector based upon the given track segment and overlapping associated radar track signals; and b) generating the feature vector based upon repeating a portion of the feature values for the given radar track segment. 32. The method of claim 24, wherein the method includes assessing the features according to: a) calculating a plurality of feature values for several of the features based upon 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 upon the number of mis-classified radar track segments for assessing either one of the features or a combination of the features.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (20)
Tsang Shiu Ming ; Salvatore ; Jr. Gerald P., Apparatus and method for mitigating multipath.
Schutte Peter C. (12 Gail Ct. Huntington NY 11743) Koutsoudis Theodore (5 Vin Ct. Farmingdale NY 11735) Kadar Ivan (75-11 195th St. Flushing NY 11366), Apparatus for combining data produced from multiple radars illuminating a common target area.
Haley Paul H. (Monroeville PA) Oravec James J. (Plum Boro PA) Petlevich ; Jr. Walter J. (Uniontown PA) Schweizer Philipp F. (Monroeville PA), Method and apparatus for classifying objects in sonar images.
Blackman Samuel S. (Los Angeles CA) Dempster Robert J. (Los Angeles CA), Method and apparatus for continuous time representation of multiple hypothesis tracking data.
Abrahamson Staffan,SEX ; Axelsson Dan,SEX ; Brusmark Bertil,SEX ; Gustafsson Anders,SEX ; Strifors Hans,SEX, Method of detecting and classifying objects by means of radar.
Wang, Jian; Brookner, Eli; Drake, Peter R.; Fournier, Bradley; Ponsford, Anthony M.; Chang, Yueh-Chi, Methods and apparatus for 3D radar data from 2D primary surveillance radar and passive adjunct radar.
Young, Brett J.; Johnson, Jason A., Target-tracking radar classifier with glint detection and method for target classification using measured target epsilon and target glint information.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.