CN110991386B - Robust nearest neighbor radar target one-dimensional range profile identification method and device - Google Patents

Robust nearest neighbor radar target one-dimensional range profile identification method and device Download PDF

Info

Publication number
CN110991386B
CN110991386B CN201911286732.0A CN201911286732A CN110991386B CN 110991386 B CN110991386 B CN 110991386B CN 201911286732 A CN201911286732 A CN 201911286732A CN 110991386 B CN110991386 B CN 110991386B
Authority
CN
China
Prior art keywords
range profile
dimensional range
samples
sample
tested
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911286732.0A
Other languages
Chinese (zh)
Other versions
CN110991386A (en
Inventor
郝志刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changsha Core Lianxin Intelligent System Co ltd
Original Assignee
Changsha Core Lianxin Intelligent System Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Changsha Core Lianxin Intelligent System Co ltd filed Critical Changsha Core Lianxin Intelligent System Co ltd
Priority to CN201911286732.0A priority Critical patent/CN110991386B/en
Publication of CN110991386A publication Critical patent/CN110991386A/en
Application granted granted Critical
Publication of CN110991386B publication Critical patent/CN110991386B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The application relates to a robust nearest neighbor radar target one-dimensional range profile identification method and system. The method comprises the following steps: the method comprises the steps of obtaining mixed Gaussian distribution of training samples of various labels in a one-dimensional range profile training sample set, setting soft attribute labels of the training samples in the one-dimensional range profile training sample set according to the mixed Gaussian distribution, obtaining a feature vector of a one-dimensional range profile sample to be tested, obtaining a plurality of training samples with the nearest distance according to the feature vector as neighbor samples, obtaining distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples, conducting discount processing on the soft attribute labels corresponding to the neighbor samples according to an evidence discount rule and the distance measurement, fusing the soft attribute labels corresponding to the neighbor samples after discount processing according to an evidence combination rule to obtain quantitative trust measurement of the one-dimensional range profile sample to be tested, and identifying the one-dimensional range profile to be tested according to the quantitative trust measurement. By adopting the method, the problem of robustness of one-dimensional range profile identification of the robust nearest neighbor radar target can be solved.

Description

Robust nearest neighbor radar target one-dimensional range profile identification method and device
Technical Field
The application relates to the technical field of radar target identification, in particular to a robust nearest neighbor radar target one-dimensional range profile identification method and device.
Background
With the continuous expansion of the broadband radar technology to the civil field, the identification requirement based on the radar target one-dimensional range profile is more and more strong. The nearest neighbor method is the most classical and least primitive method in the target recognition direction, and is widely applied to engineering. Like other target identification methods, the nearest neighbor method also needs support of a large number of training sample data sets to classify and identify the test samples, and in the specific implementation process, a plurality of nearest neighbor training samples of the test samples are searched in a feature space, and decision judgment is made according to attribute information of the nearest neighbor samples.
However, because the one-dimensional range profile of the broadband radar target has the characteristics of amplitude, translation, attitude sensitivity and the like, and the presence of intentional or unintentional interference is added, the quality of a large amount of collected training sample data is different, and if the quality difference of the training samples is ignored, the decision result of the nearest neighbor method is possibly induced, and the robustness of the one-dimensional range profile identification of the robust nearest neighbor radar target is influenced.
Disclosure of Invention
Therefore, in order to solve the above technical problems, it is necessary to provide a robust nearest neighbor radar target one-dimensional range profile identification method and apparatus capable of solving the robustness problem affecting the robust nearest neighbor radar target one-dimensional range profile identification.
A robust nearest neighbor radar target one-dimensional range profile identification method comprises the following steps:
acquiring mixed Gaussian distribution of training samples of various class labels in a one-dimensional range profile training sample set;
setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance according to the feature vector to serve as neighbor samples, and acquiring distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
according to an evidence discount rule and the distance measurement, carrying out discount processing on the soft attribute label corresponding to the adjacent sample;
fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
and identifying the one-dimensional distance image to be tested according to the quantitative trust measurement.
In one embodiment, the method further comprises the following steps: by adopting an expectation maximization method, clustering analysis is carried out on training samples of different types of labels, and the mixed Gaussian distribution of the training samples of all types of labels in a one-dimensional range profile training sample set is obtained as follows:
Figure GDA0002386902430000021
/>
wherein the content of the first and second substances,
Figure GDA0002386902430000022
Figure GDA0002386902430000023
indicates the fifth->
Figure GDA0002386902430000028
The s-th cluster center corresponding to the class training sample,
Figure GDA0002386902430000024
represents the covariance matrix corresponding to the cluster center, < > or >>
Figure GDA0002386902430000029
Represents the number of cluster centers within a cluster, N (. | m, P) represents a Gaussian distribution with mean m and covariance P.
In one embodiment, the method further comprises the following steps: according to generalized Bayes' theorem, setting the soft attribute label of each training sample in the one-dimensional range profile training sample set as
Figure GDA0002386902430000025
Wherein the content of the first and second substances,
Figure GDA0002386902430000026
D x represents the dimension of vector x, { c 1 ，c 2 ，…，c L Denotes a category label.
In one embodiment, the method further comprises the following steps: normalizing the soft attribute label to obtain the normalized expression of the soft attribute label as follows:
Figure GDA0002386902430000027
in one embodiment, the method further comprises the following steps: acquiring a characteristic vector x of a one-dimensional range profile sample to be tested, and acquiring K training samples with the nearest Euclidean distance according to the characteristic vector x to serve as neighbor samples x (k) ；
Obtaining the distance measurement x between the one-dimensional range profile sample x to be tested and the neighboring sample (k) The distance metric of (d) is:
Figure GDA0002386902430000031
wherein, the one-dimensional range profile sample x is tested to be separated from the neighboring sample x (k) Is d (x, x) (k) )≥0。
In one embodiment, the method further comprises the following steps: according to the evidence discount rule and the distance measurement, the discount processing process of the soft attribute label corresponding to the neighboring sample is as follows:
Figure GDA0002386902430000032
in one embodiment, the method further comprises the following steps: according to an evidence combination rule, fusing soft attribute labels corresponding to the discounted neighbor samples to obtain a one-dimensional distance image sample quantitative trust metric mx (·) to be tested;
wherein, the first and the second end of the pipe are connected with each other,
Figure GDA0002386902430000033
indicating the size of the collision between different neighboring samples.
In one embodiment, the method further comprises the following steps: and carrying out hard decision on the quantitative trust metric according to a betting probability decision rule in a trust function theory as follows:
Figure GDA0002386902430000034
where | represents the number of collection elements.
A robust nearest-neighbor radar target one-dimensional range profile recognition apparatus, the apparatus comprising:
the label setting module is used for acquiring the Gaussian mixture distribution of the training samples of all classes of labels in the one-dimensional range profile training sample set; setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
the distance acquisition module is used for acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance as neighbor samples according to the feature vector, and acquiring the distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
the trust measurement module is used for carrying out discount processing on the soft attribute label corresponding to the neighbor sample according to an evidence discount rule and the distance measurement; fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
and the identification module is used for identifying the one-dimensional distance image to be tested according to the quantitative trust measurement.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
acquiring mixed Gaussian distribution of training samples of various class labels in a one-dimensional range profile training sample set;
setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance as neighbor samples according to the feature vector, and acquiring distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
according to an evidence discount rule and the distance measurement, carrying out discount treatment on the soft attribute label corresponding to the adjacent sample;
according to an evidence combination rule, fusing soft attribute labels corresponding to the neighbor samples after discount processing to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
and identifying the one-dimensional distance image to be tested according to the quantitative trust measurement.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
acquiring mixed Gaussian distribution of training samples of various class labels in a one-dimensional range profile training sample set;
setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance according to the feature vector to serve as neighbor samples, and acquiring distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
according to an evidence discount rule and the distance measurement, carrying out discount processing on the soft attribute label corresponding to the adjacent sample;
fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
and identifying the one-dimensional distance image to be tested according to the quantitative trust measurement.
According to the robust nearest neighbor radar target one-dimensional range profile identification method, the robust nearest neighbor radar target one-dimensional range profile identification device, the computer equipment and the storage medium, the advantages of the trust function theory in the aspects of uncertain information representation and reasoning are fully utilized, the uncertain attributes of training samples with different qualities are effectively represented, and the robust reasoning of the uncertain information is completed in the subsequent nearest neighbor classification identification. In addition, different methods can be adopted for Gaussian mixture clustering, distance measurement, evidence discount processing, evidence combination rules, decision rules and the like related to each step, and the method is independent of a specific implementation approach and has wide applicability.
Drawings
FIG. 1 is a schematic flow chart of a robust nearest neighbor radar target one-dimensional range profile identification method in an embodiment;
FIG. 2 is a block diagram of a robust nearest-neighbor radar target one-dimensional range profile recognition apparatus in an embodiment;
FIG. 3 is a diagram of the internal structure of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
In one embodiment, as shown in fig. 1, a robust one-dimensional range profile identification method for a nearest neighbor radar target is provided, which can be applied in a terminal, and includes the following steps:
and 102, acquiring the mixed Gaussian distribution of the training samples of each class label in the one-dimensional range profile training sample set.
The one-dimensional range image is the vector sum of the projection of target scattering point sub-echoes on radar rays acquired by broadband radar signals, and not only provides the geometrical shape and structural characteristics of the target, but also contains more relevant information required by target identification. The one-dimensional range profile is actually a scatter intensity profile for each range bin on the target.
The training sample set comprises a plurality of training samples, each training sample is classified under a class label, and the class labels are the belonged classes of different radar targets.
The mixed gaussian distribution can be obtained by using a gaussian mixed clustering algorithm of methods such as expectation maximization, probability distribution and the like.
And 104, setting soft attribute labels of all training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution.
The soft attribute labels are used for representing uncertainty of the class attributes of the training samples, so that the training samples with different qualities are represented, and the soft attribute labels of the training samples can be determined by specifically adopting generalized Bayes' theorem in a belief function theory to analyze.
And 106, acquiring a feature vector of the one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance according to the feature vector to serve as neighbor samples, and acquiring the distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples.
The distance measurement may be implemented by euclidean distance or mahalanobis distance, which is not limited herein.
And 108, according to the evidence discount rule and the distance measurement, carrying out discount processing on the soft attribute label corresponding to the adjacent sample.
The evidence discount rule is based on evidence theory and is a correction means before evidence combination. In this step, the soft attribute labels of the neighboring samples are corrected.
In this step, the evidence discount rule may also be evidence discount processing based on classical, context, etc.
And step 110, fusing the soft attribute labels corresponding to the discounted neighboring samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested.
And fusing the soft attribute labels to obtain the quantitative trust measurement of the one-dimensional distance image sample to be tested. It is worth mentioning that the Evidence combination rules and Evidence discount rules both come from the document Shafer G.A physical Theory of Evidence [ M ], princeton: princeton University Press,1976.
The evidence combination rule can also be based on local and global conflict proportion distribution and other evidence combination rules.
And 112, identifying the one-dimensional range profile to be tested according to the quantitative trust measurement.
In making the decision, a decision rule based on trust, betting probability and the like can also be adopted.
In the robust nearest neighbor radar target one-dimensional range profile identification method, the advantages of the trust function theory in the aspects of uncertain information representation and reasoning are fully utilized, the uncertain attributes of training samples with different qualities are effectively represented, and the robust reasoning of the uncertain information is completed in the subsequent nearest neighbor classification identification. In addition, different methods can be adopted for Gaussian mixture clustering, distance measurement, evidence discount processing, evidence combination rules, decision rules and the like related to each step, and the method is independent of a specific implementation approach and has wide applicability.
In one embodiment, the specific step of calculating the mixed gaussian distribution includes: by adopting an expectation maximization method, clustering analysis is carried out on training samples of different types of labels, and the mixed Gaussian distribution of the training samples of all types of labels in the one-dimensional range profile training sample set is obtained as follows:
Figure GDA0002386902430000071
wherein the content of the first and second substances,
Figure GDA0002386902430000072
Figure GDA0002386902430000073
indicates the fifth->
Figure GDA0002386902430000079
The s-th cluster center corresponding to the class training sample,
Figure GDA0002386902430000074
represents the covariance matrix corresponding to the cluster center, < > or >>
Figure GDA00023869024300000710
Represents the number of cluster centers within a cluster, N (. | m, P) represents a Gaussian distribution with mean m and covariance P.
In another embodiment, the step of setting the soft attribute tag comprises: according to the generalized Bayes theorem, the soft attribute labels of all the training samples in the one-dimensional range profile training sample set are set to
Figure GDA0002386902430000075
Wherein, the first and the second end of the pipe are connected with each other,
Figure GDA0002386902430000076
D x represents the dimension of vector x, { c 1 ，c 2 ，…，c L Indicates a category label.
In addition, soft attribute labels are required to be normalized to obtain
Figure GDA0002386902430000077
In one embodiment, the acquiring of the neighbor samples and the distance measurement are required, which specifically includes: acquiring a characteristic vector x of a one-dimensional range profile sample to be tested, and acquiring K training samples with the nearest Euclidean distance according to the characteristic vector x as neighbor samples x (k) (ii) a Obtaining the distance measurement x between the one-dimensional range profile sample x to be tested and the neighboring sample (k) The distance metric of (d) is:
Figure GDA0002386902430000078
wherein, the one-dimensional range profile sample x is tested to be away from the neighbor sample x (k) Is d (x, x) (k) ) Is more than or equal to 0. In particular, xAnd x (k) Are all vector representations.
In one embodiment, the step of performing a discount process includes: according to the evidence discount rule and the distance measurement, the discount processing process of the soft attribute label corresponding to the adjacent sample is as follows:
Figure GDA0002386902430000081
in one embodiment, the step of performing label fusion comprises: according to an evidence combination rule, fusing soft attribute labels corresponding to the discounted neighbor samples to obtain a one-dimensional distance image sample quantitative trust metric mx (·) to be tested;
wherein the content of the first and second substances,
Figure GDA0002386902430000082
indicating the size of the collision between different neighboring samples.
In one embodiment, the hard decision of the quantitative trust metric according to the betting probability decision rule in the trust function theory is as follows:
Figure GDA0002386902430000083
where | represents the number of collection elements.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least a portion of the steps in fig. 1 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 2, there is provided a robust one-dimensional range profile recognition apparatus for nearest-neighbor radar targets, comprising: a tag setting module 202, a distance acquisition module 204, a trust metric module 206, and an identification module 208, wherein:
the label setting module 202 is configured to obtain mixed gaussian distribution of training samples of each class label in the one-dimensional range profile training sample set; setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
a distance obtaining module 204, configured to obtain a feature vector of a one-dimensional range profile sample to be tested, obtain, according to the feature vector, a plurality of training samples closest to each other as neighbor samples, and obtain a distance measure between the one-dimensional range profile sample to be tested and the neighbor samples;
a trust measurement module 206, configured to perform discount processing on the soft attribute label corresponding to the neighboring sample according to an evidence discount rule and the distance measurement; fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
and the identifying module 208 is configured to identify the to-be-tested one-dimensional range profile according to the quantitative trust metric.
In one embodiment, the label setting module 202 is further configured to perform cluster analysis on training samples of different types of labels by using an expected maximization method, and obtain a gaussian mixture distribution of the training samples of each type of label in the one-dimensional range profile training sample set as follows:
Figure GDA0002386902430000091
wherein, the first and the second end of the pipe are connected with each other,
Figure GDA0002386902430000092
Figure GDA0002386902430000093
represents a fifth or fifth party>
Figure GDA0002386902430000096
The s-th cluster center corresponding to the class training sample,
Figure GDA0002386902430000094
represents the covariance matrix corresponding to the cluster center, < > or >>
Figure GDA0002386902430000097
Represents the number of cluster centers within a cluster, N (. | m, P) represents a Gaussian distribution with mean m and covariance P.
In one embodiment, the label setting module 202 is further configured to set the soft attribute labels of the training samples in the one-dimensional distance image training sample set to soft attribute labels according to a generalized bayes theorem by using the gaussian mixture distribution
Figure GDA0002386902430000095
Wherein, the first and the second end of the pipe are connected with each other,
Figure GDA0002386902430000101
D x represents the dimension of vector x, { c 1 ，c 2 ，…，c L Indicates a category label.
In one embodiment, the label setting module 202 is further configured to normalize the soft attribute label, and the normalized representation of the soft attribute label is obtained as:
Figure GDA0002386902430000102
in one embodiment, the distance obtaining module 204 is further configured to obtain a feature vector x of the one-dimensional range profile sample to be tested, and obtain K training samples with the shortest euclidean distance as neighboring samples x according to the feature vector x (k) (ii) a Obtaining the said treatmentTesting a distance measure x of a one-dimensional range profile sample x from the neighbor sample (k) The distance metric of (d) is:
Figure GDA0002386902430000103
wherein, the one-dimensional range profile sample x is tested to be away from the neighbor sample x (k) Is d (x, x) (k) )≥0。
In one embodiment, the trust metric module 206 is further configured to perform a discount process on the soft attribute labels corresponding to the neighbor samples according to the evidence discount rule and the distance metric as follows:
Figure GDA0002386902430000104
in one embodiment, the trust metric module 206 is further configured to fuse the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a one-dimensional distance image sample quantization trust metric mx (·) to be tested;
wherein the content of the first and second substances,
Figure GDA0002386902430000105
indicating the size of the collision between different neighboring samples.
In one embodiment, the identification module 208 is further configured to make a hard decision on the quantized trust metric according to a betting probability decision rule in the trust function theory as follows:
Figure GDA0002386902430000111
where | represents the number of collection elements.
For specific limitations of the robust nearest-neighbor radar target one-dimensional range profile identification device, reference may be made to the above limitations of the robust nearest-neighbor radar target one-dimensional range profile identification method, and details are not repeated here. The modules in the robust nearest-neighbor radar target one-dimensional range profile identification device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 3. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a robust nearest neighbor radar target one-dimensional range profile identification method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 3 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In an embodiment, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the above method embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (8)

1. A robust nearest neighbor radar target one-dimensional range profile identification method comprises the following steps:
acquiring mixed Gaussian distribution of training samples of various class labels in a one-dimensional range profile training sample set;
setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance as neighbor samples according to the feature vector, and acquiring distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
according to an evidence discount rule and the distance measurement, carrying out discount processing on the soft attribute label corresponding to the adjacent sample;
fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
identifying the one-dimensional range profile to be tested according to the quantitative trust measurement;
the obtaining of the mixed Gaussian distribution of the training samples of each class label in the one-dimensional range profile training sample set comprises:
by adopting an expectation maximization method, clustering analysis is carried out on training samples of different types of labels, and the mixed Gaussian distribution of the training samples of all types of labels in the one-dimensional range profile training sample set is obtained as follows:
Figure FDA0004112491820000011
wherein the content of the first and second substances,
Figure FDA0004112491820000012
Figure FDA0004112491820000013
represents the s cluster center corresponding to the i training sample>
Figure FDA0004112491820000014
A covariance matrix, M, representing the correspondence of the cluster centers l Representing the number of cluster centers within a cluster, N (· | m, P) represents a Gaussian distribution with mean m and covariance P;
according to the Gaussian mixture distribution, setting a soft attribute label of each training sample in the one-dimensional range profile training sample set, wherein the soft attribute label comprises the following steps:
according to the generalized Bayes theorem, the soft attribute labels of all the training samples in the one-dimensional range profile training sample set are set to
Figure FDA0004112491820000015
Wherein the content of the first and second substances,
Figure FDA0004112491820000021
D x represents the dimension of vector x, { c 1 ,c 2 ,…,c L Indicates a category label.
2. The method of claim 1, further comprising:
normalizing the soft attribute label to obtain the normalized expression of the soft attribute label as follows:
Figure FDA0004112491820000022
3. the method according to claim 1, wherein the obtaining a feature vector of a one-dimensional range profile sample to be tested, obtaining a plurality of the training samples closest to each other as neighbor samples according to the feature vector, and obtaining a distance measure between the one-dimensional range profile sample to be tested and the neighbor samples comprises:
acquiring a characteristic vector x of a one-dimensional range profile sample to be tested, and acquiring K training samples with the nearest Euclidean distance according to the characteristic vector x to serve as neighbor samples x (k) ；
Obtaining the distance measurement x between the one-dimensional range profile sample x to be tested and the neighboring sample (k) The distance metric of (d) is:
Figure FDA0004112491820000023
wherein, the one-dimensional range profile sample x is tested to be away from the neighbor sample x (k) Is d (x, x) (k) )≥0。
4. The method of claim 3, wherein discounting the soft attribute labels corresponding to the neighboring samples according to an evidence discount rule and the distance metric comprises:
according to the evidence discount rule and the distance measurement, the discount processing process of the soft attribute label corresponding to the neighboring sample is as follows:
Figure FDA0004112491820000024
5. the method according to claim 4, wherein the fusing the soft attribute labels corresponding to the discounted neighboring samples according to an evidence combination rule to obtain a quantitative confidence measure of the one-dimensional distance image sample to be tested, comprises:
according to an evidence combination rule, fusing soft attribute labels corresponding to the discounted neighbor samples to obtain a one-dimensional distance image sample quantitative trust metric mx (·) to be tested;
wherein the content of the first and second substances,
Figure FDA0004112491820000031
Figure FDA0004112491820000032
indicating the size of the collision between different neighboring samples.
6. The method of claim 5, wherein identifying the one-dimensional range profile to be tested according to the quantitative confidence metric comprises:
and carrying out hard decision on the quantitative trust metric according to a betting probability decision rule in a trust function theory as follows:
Figure FDA0004112491820000033
where | represents the number of collection elements.
7. A robust nearest-neighbor radar target one-dimensional range profile recognition apparatus, comprising:
the label setting module is used for acquiring the mixed Gaussian distribution of the training samples of each class label in the one-dimensional range profile training sample set; setting soft attribute labels of the training samples in the one-dimensional distance image training sample set according to the mixed Gaussian distribution;
the distance acquisition module is used for acquiring a feature vector of a one-dimensional range profile sample to be tested, acquiring a plurality of training samples with the nearest distance as neighbor samples according to the feature vector, and acquiring the distance measurement between the one-dimensional range profile sample to be tested and the neighbor samples;
the trust measurement module is used for carrying out discount processing on the soft attribute label corresponding to the neighbor sample according to an evidence discount rule and the distance measurement; fusing the soft attribute labels corresponding to the discounted neighbor samples according to an evidence combination rule to obtain a quantitative trust measurement of the one-dimensional distance image sample to be tested;
the identification module is used for identifying the one-dimensional distance image to be tested according to the quantitative trust measurement;
the distance acquisition module is further used for performing clustering analysis on the training samples of different types of labels by adopting an expected maximization method to obtain the following mixed Gaussian distribution of the training samples of all types of labels in the one-dimensional range profile training sample set:
Figure FDA0004112491820000041
wherein the content of the first and second substances,
Figure FDA0004112491820000042
Figure FDA0004112491820000043
represents the s cluster center corresponding to the i training sample>
Figure FDA0004112491820000044
A covariance matrix, M, representing the correspondence of the cluster centers l Representing the number of cluster centers within a cluster, N (· | m, P) represents a Gaussian distribution with mean m and covariance P;
the label setting module is also used for setting the soft attribute labels of all the training samples in the one-dimensional distance image training sample set into
Figure FDA0004112491820000045
Wherein the content of the first and second substances,
Figure FDA0004112491820000046
D x represents the dimension of vector x, { c 1 ,c 2 ,…,c L Indicates a category label.
8. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 6 when executing the computer program.
CN201911286732.0A 2019-12-14 2019-12-14 Robust nearest neighbor radar target one-dimensional range profile identification method and device Active CN110991386B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911286732.0A CN110991386B (en) 2019-12-14 2019-12-14 Robust nearest neighbor radar target one-dimensional range profile identification method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911286732.0A CN110991386B (en) 2019-12-14 2019-12-14 Robust nearest neighbor radar target one-dimensional range profile identification method and device

Publications (2)

Publication Number Publication Date
CN110991386A CN110991386A (en) 2020-04-10
CN110991386B true CN110991386B (en) 2023-04-18

Family

ID=70093598

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911286732.0A Active CN110991386B (en) 2019-12-14 2019-12-14 Robust nearest neighbor radar target one-dimensional range profile identification method and device

Country Status (1)

Country Link
CN (1) CN110991386B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106443632A (en) * 2016-12-01 2017-02-22 西安电子科技大学 Radar target identification method based on label maintaining multitask factor analyzing model
CN107463966A (en) * 2017-08-17 2017-12-12 电子科技大学 Radar range profile's target identification method based on dual-depth neutral net
WO2018090937A1 (en) * 2016-11-18 2018-05-24 深圳云天励飞技术有限公司 Image processing method, terminal and storage medium

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090092299A1 (en) * 2007-10-03 2009-04-09 Siemens Medical Solutions Usa, Inc. System and Method for Joint Classification Using Feature Space Cluster Labels
US8642872B2 (en) * 2008-03-03 2014-02-04 Microsoft Corporation Music steering with automatically detected musical attributes

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018090937A1 (en) * 2016-11-18 2018-05-24 深圳云天励飞技术有限公司 Image processing method, terminal and storage medium
CN106443632A (en) * 2016-12-01 2017-02-22 西安电子科技大学 Radar target identification method based on label maintaining multitask factor analyzing model
CN107463966A (en) * 2017-08-17 2017-12-12 电子科技大学 Radar range profile's target identification method based on dual-depth neutral net

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Andrew O. Finley 等.Efficient algorithms for Bayesian Nearest Neighbor Gaussian Processes.《Comput Graph Stat》.2019,全文. *
刘邱云 ; 付雪峰 ; 吴根秀 ; .基于可传递信度模型的自适应k-NN分类器.计算机研究与发展.2008,(S1), *
李睿 ; 王晓丹 ; 蕾蕾 ; 赵振冲 ; .一种基于RVM和DS的一维距离像融合识别方法.智能系统学报.(04), *
李锋 等.基于标签特征和相关性的多标签分类算法.《计算机工程与应用》.2019,全文. *
林志贵 ; 袁臣虎 ; 冯志红 ; .基于D-S理论的多源信息融合冲突问题处理方法.计算机工程与应用.2006,(35), *

Also Published As

Publication number Publication date
CN110991386A (en) 2020-04-10

Similar Documents

Publication Publication Date Title
Leegwater et al. Performance study of a score‐based likelihood ratio system for forensic fingermark comparison
US20210295162A1 (en) Neural network model training method and apparatus, computer device, and storage medium
Herman et al. Mutual information-based method for selecting informative feature sets
CN111860674A (en) Sample class identification method and device, computer equipment and storage medium
CN109285105B (en) Watermark detection method, watermark detection device, computer equipment and storage medium
CN113299346B (en) Classification model training and classifying method and device, computer equipment and storage medium
CN111126339A (en) Gesture recognition method and device, computer equipment and storage medium
CN116597384B (en) Space target identification method and device based on small sample training and computer equipment
Yuan et al. Evidential deep neural networks for uncertain data classification
Zhang et al. Leveraging uncertainty from deep learning for trustworthy material discovery workflows
CN110991538A (en) Sample classification method and device, storage medium and computer equipment
Warif et al. A comprehensive evaluation procedure for copy-move forgery detection methods: results from a systematic review
CN112464660B (en) Text classification model construction method and text data processing method
CN110991386B (en) Robust nearest neighbor radar target one-dimensional range profile identification method and device
CN113537020A (en) Complex SAR image target identification method based on improved neural network
Qu et al. Boundary detection using a Bayesian hierarchical model for multiscale spatial data
Pankaj et al. CGA: An image processing based software for surface strain analysis in sheet metal forming
CN114422450B (en) Network traffic analysis method and device based on multi-source network traffic data
CN110889432A (en) Feature point matching method and device, computer equipment and storage medium
CN112859034B (en) Natural environment radar echo amplitude model classification method and device
CN115424267A (en) Rotating target detection method and device based on Gaussian distribution
Alfaz et al. A deep convolutional neural network based approach to classify and detect crack in concrete surface using xception
CN114359232A (en) Image change detection method and device based on context covariance matrix
CN110852400A (en) Classification model evaluation method and device, computer equipment and storage medium
CN113095963A (en) Real estate cost data processing method, real estate cost data processing device, computer equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant