CN112149061A - Multi-class average maximization true and false target feature extraction method - Google Patents
Multi-class average maximization true and false target feature extraction method Download PDFInfo
- Publication number
- CN112149061A CN112149061A CN202011026888.8A CN202011026888A CN112149061A CN 112149061 A CN112149061 A CN 112149061A CN 202011026888 A CN202011026888 A CN 202011026888A CN 112149061 A CN112149061 A CN 112149061A
- Authority
- CN
- China
- Prior art keywords
- true
- characteristic
- target
- parameter
- class
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
- G01S13/06—Systems determining position data of a target
- G01S13/08—Systems for measuring distance only
- G01S13/10—Systems for measuring distance only using transmission of interrupted, pulse modulated waves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/28—Details of pulse systems
- G01S7/285—Receivers
- G01S7/292—Extracting wanted echo-signals
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/36—Means for anti-jamming, e.g. ECCM, i.e. electronic counter-counter measures
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/418—Theoretical aspects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Operations Research (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of radar target identification, and particularly relates to a method for extracting characteristics of multiple types of average maximized true and false targets. The method adopts the multi-component Gaussian distribution to represent the likelihood function of the target data, can still accurately describe the distribution condition of the target data under the condition that target sample data is non-Gaussian distribution, and screens out the most effective classification and identification features from the characteristic elements of the one-dimensional distance image of the target. The defect that the conventional method is only suitable for Gaussian distribution of sample data is overcome, so that the target identification performance is improved, simulation experiments are carried out on the one-dimensional range profile data of four types of simulation targets, and the experimental result shows that the method is effective.
Description
Technical Field
The invention belongs to the technical field of radar target identification, and particularly relates to a method for extracting characteristics of multiple types of average maximized true and false targets.
Background
In radar target identification, feature extraction is a very critical step, so that on one hand, the dimension of an input vector can be reduced, the calculated amount can be reduced, and original classification information can be well maintained. The subspace is a commonly used feature extraction method, that is, a corresponding subspace is obtained under the condition that a certain criterion is satisfied. For example, the decision vector subspace method is a method in which a decision vector subspace is obtained using the fisher criterion, whereas the feature subspace method is a method in which a feature subspace is obtained under a condition in which a reconstruction error is minimum, and a good recognition effect can be obtained under a condition in which sample data is gaussian distributed.
However, the feature extraction method such as subspace is only suitable for the case where the sample data is gaussian distributed, and the distribution of the sample data may be non-gaussian in practice, and the recognition performance of the conventional feature extraction method is significantly reduced for the non-gaussian distribution case. There is room for further improvement in the recognition performance of the conventional feature extraction method.
Disclosure of Invention
Aiming at the problems, the invention provides a multi-class average maximization feature extraction method, which uses multi-component Gaussian distribution to represent likelihood functions of various target data, and can still accurately describe the distribution condition of the target data under the condition of non-Gaussian distribution, thereby screening out the most effective features for classification identification, overcoming the defects of the conventional feature extraction method, and effectively improving the classification performance of radar true and false targets.
The technical scheme of the invention is as follows:
a method for extracting features of multiple classes of average maximized true and false targets is characterized by comprising the following steps:
s1, setting n-dimension column vector xcjC is more than or equal to 1 and less than or equal to g, j is more than or equal to 1 and less than or equal to N for the jth training one-dimensional range profile of the c-th true and false targetc,Wherein g is the number of classes, NcThe number of training one-dimensional range profile samples of a class c true and false target is shown, N is the total number of the training one-dimensional range profile samples, and a likelihood function of the class c is represented by a plurality of Gaussian components:
whereinIs the weight coefficient of each gaussian component,and isMcThe number of gaussian components for class c,a parameter representing a gaussian component term, p (·) being a probability density function;
s2, assuming that each characteristic element in the one-dimensional image data is independent of each other, if the element is independent of the category, the element is not a valid characteristic, and the element obeys the distribution q (x)cjk/γck),xcjkDenotes xcjThe k element of (a), γckDefining a two-valued variable psi for the distribution coefficients of the kth characteristic element of the class cckIf the feature element is valid, thenck1, otherwise psick0; make the probability of the feature element being valid as betak=P(ψck1), modeling according to a likelihood function:
wherein r isckDistribution coefficients of kth characteristic elements of a type c;
s3, estimating the parameters of the model established in the step S2 by a mean value maximization method:
s31, initializing a mean vector, a covariance matrix and a weight of the Gaussian component;
s32, calculating the updated value of the model parameter:
wherein n isθAnd nγAre respectively a parameter setAndthe number of parameters in (1), E (-) and var (-) respectively represent a mean parameter and a variance parameter in a parameter set;
s33, calculating the error before and after each parameter updating, if the error is less than the specified threshold value or exceeds the specified iteration times, the updating process is terminated and the step S4 is carried out, otherwise, the step S32 is returned to continue the iterative updating;
s4, ifIf the value is larger than the given threshold value, reserving the kth characteristic element, and forming a new characteristic vector by the reserved characteristic elements to be used as a characteristic extraction result.
After the feature vectors are obtained, the classification and identification of the targets can be finished by utilizing a classifier
The invention has the beneficial effects that: under the condition of non-Gaussian distribution, the distribution condition of target data can still be accurately described, so that the most effective characteristics for classification and identification are screened, the defects of the conventional characteristic extraction method are overcome, and the classification performance of radar true and false targets is effectively improved.
Detailed Description
The technical scheme of the invention is described in detail by combining simulation experiments as follows:
the method of the invention mainly comprises the following steps:
let xcjThe (N-dimensional column vector) is the jth training one-dimensional distance image of the c-th true and false target, c is more than or equal to 1 and less than or equal to g, j is more than or equal to 1 and less than or equal to Nc,Wherein N iscThe number of training one-dimensional range image samples of class c true and false targets, and N is the number of training one-dimensional range image samplesAnd (4) total number. Class c likelihood function representation using polynomial gaussian components
WhereinIs the weight coefficient of each gaussian component,and isMcThe number of gaussian components for class c,the parameter representing the gaussian component term, p (·) is a probability density function. Assuming that the feature elements in the one-dimensional image data are independent of each other, equation (1) may be changed to
If an element is independent of a class, then the element is not a valid feature, let it obey the distribution q (x)cjk/γck),γckFor the distribution coefficient of the kth characteristic element of class c, equation (2) is transformed into
Wherein psickIs a two-valued variable, if the feature element is valid, phick1, otherwise psick0. Make the probability of the feature element being valid as
βk=P(ψck=1) (4)
The formula (3) can be represented as
The parameters in equation (5) are estimated by a mean maximization method, which comprises the following steps:
step 1: initializing a mean vector, a covariance matrix and a weight of the Gaussian component;
step 2: calculating updated values of model parameters
Wherein n isθAnd nγAre respectively a parameter setAnde (-) and var (-) represent the mean parameter and variance parameter, respectively, in the parameter set. Since each mixture component is Gaussian distributed, nθ=2,nγ=2。
And step 3: and (3) calculating errors before and after the updating of each parameter, if the errors are smaller than a specified threshold value or exceed a specified iteration number, terminating the updating process, and otherwise, turning to the step 2 to continue the iterative updating.
If it isAnd if the value is larger than the given threshold, reserving the kth characteristic element, grouping the reserved characteristic elements into a new characteristic vector, and then finishing the classification and identification of the target by utilizing a classifier.
To verify the effectiveness of the proposed method, the following simulation experiments were performed.
Four point targets were designed: true objects, debris, light baits, and heavy baits. The bandwidth of radar emission pulse is 1000MHZ (the range resolution is 0.15m, the radar radial sampling interval is 0.075m), the target is set as a uniform scattering point target, the scattering point of a true target is 7, and the number of the scattering points of the other three targets is 11. In the one-dimensional distance images of every 1 degree within the range of the target attitude angle of 0-80 degrees, the one-dimensional distance images of the target attitude angle of 0 degree, 2 degrees, 4 degrees, 6 degrees, and 90 degrees are taken for training, and the one-dimensional distance images of the rest attitude angles are taken as test data, so that each category of targets has 45 test samples.
For four targets (true target, fragment, light bait and heavy bait), in the range of 0-90 degrees of attitude angle, the multi-class average maximization feature extraction method and the method based on judgment are utilizedThe identification experiment is carried out by the method for extracting the other vector subspace characteristics, and the result is shown in the table I. The experimental parameters were: mcThe iteration stop threshold value is taken to be 0.0001, and the threshold for selecting the characteristic element is taken to be 0.001. In addition, the mean and covariance matrices of the various types of samples are used as initial values of the mean and variance matrices of the gaussian components.
As can be seen from the results in table 1, for the true target, the recognition rate of the discriminant vector subspace feature extraction method is 83%, while the recognition rate of the multi-class average maximization feature extraction method of the present invention is 92%; for the fragments, the recognition rate of the discrimination vector subspace feature extraction method is 78%, while the recognition rate of the multi-class average maximization feature extraction method is 86%; for light baits, the recognition rate of the discrimination vector subspace feature extraction method is 80%, and the recognition rate of the multi-class average maximization feature extraction method is 85%; for heavy bait, the recognition rate of the discriminant vector subspace feature extraction method is 82%, while the recognition rate of the multi-class average maximization feature extraction method of the invention is 84%. On average, for four types of targets, the correct recognition rate of the multi-type average maximization feature extraction method is higher than that of a discrimination vector subspace feature extraction method, and the fact that the multi-type average maximization feature extraction method provided by the invention really improves the recognition performance of the multi-type targets is shown.
TABLE 1 identification results of the two methods
Claims (1)
1. A method for extracting features of multiple classes of average maximized true and false targets is characterized by comprising the following steps:
s1, setting n-dimension column vector xcjC is more than or equal to 1 and less than or equal to g, j is more than or equal to 1 and less than or equal to N for the jth training one-dimensional range profile of the c-th true and false targetc,Wherein g is the number of classes, NcTraining for class c true and false targetsThe number of the one-dimensional range profile samples is N, the total number of the training one-dimensional range profile samples is N, and a likelihood function of a class c is represented by a plurality of Gaussian components:
whereinIs the weight coefficient of each gaussian component,and isMcThe number of gaussian components for class c,a parameter representing a gaussian component term, p (·) being a probability density function;
s2, assuming that each characteristic element in the one-dimensional image data is independent of each other, if the element is independent of the category, the element is not a valid characteristic, and the element obeys the distribution q (x)cjk/γck),xcjkDenotes xcjThe k element of (a), γckDefining a two-valued variable psi for the distribution coefficient of the kth characteristic element of the class c, determined by the distribution typeckIf the feature element is valid, thenck1, otherwise psick0; make the probability of the feature element being valid as betak=P(ψck1), modeling according to a likelihood function:
wherein gamma isckDistribution coefficients of kth characteristic elements of a type c;
s3, estimating the parameters of the model established in the step S2 by a mean value maximization method:
s31, initializing a mean vector, a covariance matrix and a weight of the Gaussian component;
s32, calculating the updated value of the model parameter:
wherein n isθAnd nγAre respectively a parameter setAndthe number of parameters in (1), E (-) and var (-) respectively represent a mean parameter and a variance parameter in a parameter set;
s33, calculating the error before and after each parameter updating, if the error is less than the specified threshold value or exceeds the specified iteration times, the updating process is terminated and the step S4 is carried out, otherwise, the step S32 is returned to continue the iterative updating;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011026888.8A CN112149061A (en) | 2020-09-25 | 2020-09-25 | Multi-class average maximization true and false target feature extraction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011026888.8A CN112149061A (en) | 2020-09-25 | 2020-09-25 | Multi-class average maximization true and false target feature extraction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112149061A true CN112149061A (en) | 2020-12-29 |
Family
ID=73897416
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011026888.8A Pending CN112149061A (en) | 2020-09-25 | 2020-09-25 | Multi-class average maximization true and false target feature extraction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112149061A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113960553A (en) * | 2021-10-14 | 2022-01-21 | 电子科技大学 | Gaussian weight distribution feature extraction method in one-dimensional image recognition |
CN115712867A (en) * | 2022-11-03 | 2023-02-24 | 哈尔滨工程大学 | Multi-component radar signal modulation identification method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108828574A (en) * | 2018-04-26 | 2018-11-16 | 电子科技大学 | The separation enhancing true and false target's feature-extraction method in subspace between one type |
CN108845302A (en) * | 2018-08-23 | 2018-11-20 | 电子科技大学 | A kind of true and false target's feature-extraction method of k nearest neighbor transformation |
CN110007286A (en) * | 2019-04-22 | 2019-07-12 | 电子科技大学 | A kind of true and false target one-dimensional range profile feature extracting method of linear discriminant study |
CN110658507A (en) * | 2019-10-12 | 2020-01-07 | 电子科技大学 | Multi-class average maximization true and false target feature extraction method for radar target identification |
CN110687514A (en) * | 2019-10-16 | 2020-01-14 | 电子科技大学 | Nonlinear discrimination learning true and false target one-dimensional range profile feature extraction method |
-
2020
- 2020-09-25 CN CN202011026888.8A patent/CN112149061A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108828574A (en) * | 2018-04-26 | 2018-11-16 | 电子科技大学 | The separation enhancing true and false target's feature-extraction method in subspace between one type |
CN108845302A (en) * | 2018-08-23 | 2018-11-20 | 电子科技大学 | A kind of true and false target's feature-extraction method of k nearest neighbor transformation |
CN110007286A (en) * | 2019-04-22 | 2019-07-12 | 电子科技大学 | A kind of true and false target one-dimensional range profile feature extracting method of linear discriminant study |
CN110658507A (en) * | 2019-10-12 | 2020-01-07 | 电子科技大学 | Multi-class average maximization true and false target feature extraction method for radar target identification |
CN110687514A (en) * | 2019-10-16 | 2020-01-14 | 电子科技大学 | Nonlinear discrimination learning true and false target one-dimensional range profile feature extraction method |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113960553A (en) * | 2021-10-14 | 2022-01-21 | 电子科技大学 | Gaussian weight distribution feature extraction method in one-dimensional image recognition |
CN113960553B (en) * | 2021-10-14 | 2023-04-18 | 电子科技大学 | Gaussian weight distribution feature extraction method in one-dimensional image recognition |
CN115712867A (en) * | 2022-11-03 | 2023-02-24 | 哈尔滨工程大学 | Multi-component radar signal modulation identification method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110658507B (en) | Multi-class average maximization true and false target feature extraction method for radar target identification | |
CN110007286B (en) | Linear discriminant learning true and false target one-dimensional range profile feature extraction method | |
CN108345911B (en) | Steel plate surface defect detection method based on convolutional neural network multi-stage characteristics | |
CN112149061A (en) | Multi-class average maximization true and false target feature extraction method | |
CN108122008B (en) | SAR image recognition method based on sparse representation and multi-feature decision-level fusion | |
CN111401452B (en) | Image classification method of equal-variation convolution network model based on partial differential operator | |
CN107238822B (en) | Method for extracting orthogonal nonlinear subspace characteristics of true and false target one-dimensional range profile | |
CN107271965B (en) | Method for extracting true and false target one-dimensional range profile features in cluster subspace | |
CN108182440B (en) | Method for acquiring surrounding rock category based on slag slice image recognition | |
CN108845302B (en) | K-nearest neighbor transformation true and false target feature extraction method | |
CN108491864B (en) | Hyperspectral image classification based on automatic determination of convolution kernel size convolutional neural network | |
CN110443257B (en) | Significance detection method based on active learning | |
CN110687514B (en) | Nonlinear discrimination learning true and false target one-dimensional range profile feature extraction method | |
CN111652264B (en) | Negative migration sample screening method based on maximum mean value difference | |
CN110738695A (en) | image feature point mismatching and removing method based on local transformation model | |
CN114241233B (en) | Nonlinear class group sparse representation real and false target one-dimensional range profile identification method | |
CN112183617B (en) | RCS sequence feature extraction method for sample and class label maximum correlation subspace | |
CN117274702B (en) | Automatic classification method and system for cracks of mobile phone tempered glass film based on machine vision | |
CN111325158B (en) | CNN and RFC-based integrated learning polarized SAR image classification method | |
CN110068799B (en) | Sparse neighborhood center preserving RCS sequence feature extraction method | |
CN108182684B (en) | Image segmentation method and device based on weighted kernel function fuzzy clustering | |
CN114332452B (en) | Automatic detection method for image annotation result of target detection or target segmentation | |
CN110826599B (en) | Sparse representation sample distribution boundary retention feature extraction method | |
CN110780270B (en) | Target library attribute discrimination local regular learning subspace feature extraction method | |
CN111160464B (en) | Industrial high-order dynamic process soft measurement method based on multi-hidden-layer weighted dynamic model |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201229 |