CN112612006B - Deep learning-based non-uniform clutter suppression method for airborne radar - Google Patents
Deep learning-based non-uniform clutter suppression method for airborne radar Download PDFInfo
- Publication number
- CN112612006B CN112612006B CN202011340629.2A CN202011340629A CN112612006B CN 112612006 B CN112612006 B CN 112612006B CN 202011340629 A CN202011340629 A CN 202011340629A CN 112612006 B CN112612006 B CN 112612006B
- Authority
- CN
- China
- Prior art keywords
- clutter
- space
- time
- spectrum
- mvdr
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000001629 suppression Effects 0.000 title claims abstract description 33
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 238000001228 spectrum Methods 0.000 claims abstract description 48
- 238000013528 artificial neural network Methods 0.000 claims abstract description 35
- 239000011159 matrix material Substances 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 26
- 238000009825 accumulation Methods 0.000 claims abstract description 8
- 238000006243 chemical reaction Methods 0.000 claims abstract description 8
- 238000001514 detection method Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 24
- 230000003044 adaptive effect Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 2
- 238000011084 recovery Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Classifications
-
- 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
- G01S7/2923—Extracting wanted echo-signals based on data belonging to a number of consecutive radar periods
-
- 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
- G01S7/2923—Extracting wanted echo-signals based on data belonging to a number of consecutive radar periods
- G01S7/2927—Extracting wanted echo-signals based on data belonging to a number of consecutive radar periods by deriving and controlling a threshold value
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
Abstract
The invention provides an airborne radar non-uniform clutter suppression method based on deep learning, which comprises the following steps: constructing an airborne radar array element level clutter model, and generating a space-time data set and a clutter covariance matrix data set of each range gate; constructing a deep neural network, and utilizing the generated initial MVDR spectrum set as an input data set; the MVDR spectrum set is generated and is used as an output data set to train the deep neural network; n adjacent to the distance gate to be detected after analog-digital conversion 2 Estimating the range gate data to obtain a clutter MVDR space-time spectrum, and inputting the clutter MVDR space-time spectrum into a deep neural network to obtain a space-time two-dimensional clutter spectrum; reconstructing a clutter covariance matrix of the range gate data to be detected by combining a space-time steering dictionary, and obtaining a space-time self-adaptive weight by combining a target space-time steering vector to realize clutter suppression and target accumulation processing on the range gate data to be detected of the airborne radar; and then carrying out constant false alarm processing to finish the detection of the moving target. The invention can effectively inhibit clutter under the actual non-uniform condition, and has the characteristics of less sample requirement and simple engineering realization.
Description
Technical Field
The invention relates to the technical field of radar signal processing, in particular to an airborne radar non-uniform clutter suppression method based on deep learning, an airborne radar and a storage medium.
Background
At present, an aircraft flying in the air is used as a carrier, the visible distance to a low-altitude or ultra-low-altitude flying target is far than that of a ground radar, the early warning time and the monitoring range provided by the radar are greatly increased, and the radar can be flexibly and rapidly deployed in a required place, so that the airborne radar plays an irreplaceable role of other sensors in the fields of early warning, detection, warning, reconnaissance, accurate striking and the like, and becomes a key of winning local war wins in future high technical conditions in China.
As the radar works in a down-looking state, the airborne radar faces the problem of ground (sea) clutter which is more serious than that of the ground radar, and the airborne radar has wide clutter distribution range and high intensity, and the clutter intensity can reach 60-90dB especially in hilly and mountain areas; meanwhile, as the carrier moves, the ground (sea) clutter in different directions is different relative to the speed of the carrier, so that the clutter spectrum is greatly widened, the targets are often submerged in the clutter, and the target detection capability is seriously affected, so that the clutter suppression problem becomes a problem which the airborne radar needs to solve.
The space-time adaptive processing (STAP) technology is mainly adopted to restrain clutter through combining adaptive filtering in a space domain and a time domain, and has become the most effective and key technology for restraining clutter of the airborne radar. However, the clutter covariance matrix of the distance unit to be detected required for space-time adaptive weight calculation is unknown, so that clutter data (training samples) of the distance unit adjacent to the distance unit need to be obtained through maximum likelihood estimation. According to existing guidelines, a system degree of freedom is ensured in which the clutter suppression performance loss is less than twice the minimum number of uniform training samples required within 3 dB.
In an actual clutter environment, spatial variation of a geographic environment (such as a sea-land boundary, a plain-mountain boundary, an urban-rural boundary and vegetation), artificial isolated strong points (such as bridges, telegraph poles, city centers, iron towers, corner reflectors) and the like can cause the clutter power to obviously change in distance. In such a complex landform environment, the clutter power generally varies along the distance, so that the number of uniform training samples for estimating the clutter covariance matrix is seriously insufficient, and a larger error exists between the estimated clutter covariance matrix and the real clutter covariance matrix, thereby obviously reducing the clutter suppression performance of the airborne radar.
Meanwhile, three major ways of solving the above problems are as follows:
the first type is a power selective training method (PST), i.e., a non-adaptive or adaptive method is used to select a distance unit with higher power as a training sample, and a clutter covariance matrix is estimated, so as to form an ideal clutter suppression notch to suppress clutter. However, since the correlation of clutter over distance decreases with increasing distance, the strong clutter training region selected by this method, if farther from the distance cell to be detected, greatly impairs clutter suppression capability due to its poor correlation.
The second type is a direct data domain method (DDD), that is, a clutter covariance matrix is estimated by forming a plurality of samples after smoothing only using the range bin data to be detected. The method does not need to use other distance unit data except the distance unit to be detected as a sample, so that a series of problems caused by clutter non-uniformity are avoided. However, smoothing results in a large aperture loss. In addition, since the target signal of the distance unit to be detected is filtered in advance, the problem of target cancellation caused by incomplete filtering of the target signal can occur under the conditions of array element error, channel error and the like.
A third class of methods is to improve clutter suppression performance under small sample conditions by reducing training sample requirements. Representative methods are a dimension-reduction class STAP method and a sparse recovery class STAP method. Representative dimension reduction STAP methods are local joint processing (JDL) and multi-channel joint adaptive processing (mDT). The JDL method transforms the data processing domain into the wave beam-Doppler domain, and combines adjacent wave beams with adjacent Doppler channels to carry out combined self-adaptive processing, thereby effectively reducing the degree of freedom of the system and further reducing the requirement on uniform training samples. The method has low robustness to errors due to small degree of freedom in a space domain, and is not commonly adopted in actual engineering. The mDT method utilizes the full airspace channel to combine the adjacent m Doppler channels for self-adaptive processing, thereby not only reducing the degree of freedom of the system, but also having robustness to errors, and therefore being the STAP technology adopted in the current practical engineering. However, the method still has more system degrees of freedom, and the requirement on the number of uniform training samples is relatively large, for example, under the condition that an airspace is 20 channels and m is taken to be 3, the number of the required uniform training samples is 120, so that the clutter covariance matrix can not be effectively estimated under the complex landform environment with severe clutter power change, and the loss of clutter suppression performance of the airborne radar is brought. The sparse recovery method mainly comprises a sparse recovery method (IAA) based on multi-sample iterative self-adaption, a sparse recovery method (SA-MUSIC) based on subspace expansion, a sparse Bayesian learning method (SBL) and the like. According to the method, the sparse characteristic of the airborne radar clutter in the space-time plane is utilized, and the accurate clutter covariance matrix is obtained by reconstructing a small amount of observation data of the airborne radar by combining an existing sparse recovery algorithm, so that the accurate estimation of the clutter characteristic is realized under the condition of a small sample. However, three major problems remain with the sparse recovery class STAP approach at present: firstly, the model dependence problem is that once the sparse mathematical model is not matched with actual data, the performance of the algorithm is greatly reduced; secondly, the problem of huge operand is solved, and the operand required by the existing sparse recovery type algorithm is far greater than that of the traditional STAP method, so that the method cannot be practically used under the current hardware condition; third, the problem of airspace error sensitivity is solved, the space-time guiding dictionary constructed by the existing sparse recovery method is based on an ideal array manifold, and in practice, the airborne radar array inevitably has airspace error, namely, the preset dictionary is mismatched with the real array manifold, so that the algorithm performance is drastically reduced.
Therefore, to achieve effective suppression of clutter under practically non-uniform conditions, an STAP method with small sample requirements, robust error, and low computational complexity is needed for airborne radar.
Disclosure of Invention
The invention provides a deep learning-based non-uniform clutter suppression method for the airborne radar, which aims to solve the problem that the clutter suppression performance of the airborne radar is low under the actual non-uniform condition, can effectively suppress the clutter under the actual non-uniform condition, and has the characteristics of less sample requirements and simple engineering realization.
In order to solve the technical problems, the technical scheme of the invention is as follows: an airborne radar non-uniform clutter suppression method based on deep learning comprises the following steps:
s1: constructing an airborne radar array element level clutter model, and generating a space-time data set and a clutter covariance matrix data set of each range gate by simulating by combining various non-ideal factors;
s2: using n adjacent to the distance gate to be detected 1 The distance gate data is used as a training sample to generate an initial minimum variance undistorted response MVDR spectrum set; meanwhile, clutter covariance matrix corresponding to the range gate data to be detected is obtained to generate a minimum variance distortion-free response MVDR spectrum set;
s3: constructing a deep neural network, taking the initial MVDR spectrum set obtained in the step S2 as a neural network input data set, taking the MVDR spectrum set generated by the clutter covariance matrix as a neural network output data set, and training the deep neural network until convergence;
s4: the airborne radar carries out analog-digital conversion on the received actual observation data set, and then n adjacent to the range gate to be detected 2 Estimating the range gate data to obtain a clutter MVDR space-time spectrum, and inputting a trained deep neural network for processing to obtain a space-time two-dimensional clutter spectrum;
s5: reconstructing a clutter covariance matrix of the range gate data to be detected by using a space-time two-dimensional clutter spectrum and a space-time steering dictionary, and calculating by combining a known target space-time steering vector to obtain a space-time self-adaptive weight value so as to realize clutter suppression and target accumulation treatment on the range gate data to be detected of the airborne radar;
s6: and performing constant false alarm processing on the data processed by the clutter suppression and the accumulated targets to finish the detection of the moving targets.
Preferably, the non-idealities include, but are not limited to, spatial error, clutter relief, and yaw angle.
Further, the airspace error considers that the array element amplitude-phase error meets complex Gaussian distribution, the variance of the complex Gaussian distribution is selected from 0 to 0.05, and the interval is 0.002; the time domain decorrelation caused by clutter fluctuation obeys Gaussian distribution, the standard deviation of wind speed is 0 to 0.1, and the interval is 0.01; the yaw angle range of the carrier is set to 0-5 degrees, and the interval is 0.2 degrees.
Still further, the MVDR space-time spectrum has a calculation formula as follows:
wherein, R is clutter covariance matrix, phi is space-time guiding dictionary, specifically defined as:
wherein v is a space-time two-dimensional guide vector, f s Represents normalized spatial frequency, f d Expressed as normalized Doppler frequency, N s Representing the number of space frequency divisions, N d Indicating the number of doppler frequency divisions.
Further, when the MVDR space-time spectrum is obtained, the space frequency is divided into 4-5 times of the space domain array element number, and the Doppler frequency is divided into 4-5 times of the time domain pulse number.
Still further, the deep neural network employs a convolutional neural network.
Still further, in step S4, n 2 And 3-6, namely 3-6 adjacent distance gate data of the distance gate to be detected are taken as training samples to estimate and obtain clutter MVDR space-time spectrum.
Still further, the calculation formula of the space-time adaptive weight is as follows:
W＝μR -1 S
wherein ,for normalization constants, S represents a target space-time steering vector; s is S H Target space-time steering vector conjugate transposition;
and (3) applying the space-time self-adaptive weight to the range gate data to be detected to perform clutter suppression and target accumulation.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention utilizes the characteristic that the current deep learning technology can realize the high resolution reconstruction of images, trains the constructed deep neural network through the rough MVDR space-time spectrum set and the high resolution MVDR space-time spectrum set obtained by the airborne radar array element level clutter modeling simulation, and can realize the function of generating initial spectrum for small samples and generating high resolution space-time spectrum; and then inputting the space-time spectrum estimated by a small amount of samples in the actual observation data of the airborne radar into a deep neural network for processing, further obtaining a high-resolution space-time spectrum corresponding to the actual observation data, and finally constructing an accurate clutter covariance matrix based on the high-resolution spectrum and calculating to obtain a space-time self-adaptive weight. Therefore, clutter can be effectively inhibited under the actual non-uniform condition, and the method has the characteristics of less sample requirement and simple engineering realization.
In the method, the airborne radar array element level clutter modeling, the deep neural network construction and the neural network training links are all offline processing, and the subsequent actual measurement data neural network processing, the space-time self-adaptive weight construction and the constant false alarm processing part are online processing, so that the required operation amount is small, and the method is more suitable for practical engineering application.
Drawings
Fig. 1 is a flowchart of the method for non-uniform clutter suppression of the airborne radar according to the present embodiment.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some embodiments of the present invention, which are only for illustration and not to be construed as limitations of the present patent. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, the method for non-uniform clutter suppression of the airborne radar based on deep learning is characterized in that: the method comprises the following steps:
s1: constructing an airborne radar array element level clutter model, and generating a space-time data set and a clutter covariance matrix data set of each range gate by simulating by combining various non-ideal factors;
s2: using n adjacent to the distance gate to be detected 1 Taking the range gate data as a training sample to generate an initial rough MVDR spectrum set; meanwhile, a clutter covariance matrix corresponding to a distance gate to be detected is obtained to generate an MVDR spectrum set;
s3: constructing a deep neural network, taking the initial rough MVDR spectrum set obtained in the step S2 as a neural network input data set, taking the MVDR spectrum set generated by the clutter covariance matrix as a neural network output data set, and training the deep neural network until convergence;
s4: the airborne radar performs analog-digital conversion on the received actual observation data set, and then n adjacent to the distance gate to be detected after analog-digital conversion 2 Estimating clutter MVDR space-time spectrum according to the range gate data, and inputting a trained deep neural network for processing to obtain a space-time two-dimensional clutter spectrum;
s5: reconstructing a clutter covariance matrix of the range gate data to be detected by using a space-time two-dimensional clutter spectrum and a space-time steering dictionary, and calculating by combining a known target space-time steering vector to obtain a space-time self-adaptive weight value so as to realize clutter suppression and target accumulation treatment on the range gate data to be detected of the airborne radar;
s6: and performing constant false alarm processing on the data processed by the clutter suppression and the accumulated targets to finish the detection of the moving targets.
Assuming that the multichannel phased array airborne radar antenna is a uniform linear array placed on the positive side of n=8 array elements, the pulse number k=8 in the coherent processing time, the following describes the detailed steps of the whole embodiment with reference to the drawings and examples:
step S1, constructing an airborne radar array element level clutter model based on radar system parameters and a radar distance equation, and carrying out simulation modeling by considering various non-ideal factors, wherein the non-ideal factors comprise but are not limited to airspace errors, clutter fluctuation and yaw angles; the space domain error considers that the array element amplitude-phase error meets complex Gaussian distribution, and the variance is selected from 0 to 0.05, and the interval is 0.002; the time domain decorrelation caused by clutter fluctuation obeys Gaussian distribution, the standard deviation of wind speed is 0 to 0.1, and the interval is 0.01; the yaw angle range is set to 0 DEG to 5 DEG o Spaced 0.2 °; all the above three non-ideal factors are all considered in various level arrangement combinations when the clutter data are simulated.
Step S2, under various non-ideal conditions obtained through simulation, utilizing n adjacent to the distance gate to be detected 1 The range gate data is used as a training sample to generate initial rough MVThe DR spectrum set is used as the input data of the deep neural network, and the MVDR space-time spectrum obtained by the clutter covariance matrix corresponding to the range gate to be detected is used as the output data of the neural network; the MVDR space-time spectrum calculation formula is as follows:
wherein, R is clutter covariance matrix, phi is space-time guiding dictionary, specifically defined as:
wherein v is a space-time two-dimensional steering vector, f s Represents normalized spatial frequency, f d Expressed as normalized Doppler frequency, N s Representing the number of space frequency divisions, N d Indicating the number of doppler frequency divisions. When the MVDR space-time spectrum is obtained, the space frequency is divided into 4-5 times of the space domain array element number, and the Doppler frequency is divided into 4-5 times of the time domain pulse number. In this example, N is taken separately s＝4N and Nd =4k, n represents the number of spatial elements, K represents the number of time-domain pulses.
In a specific embodiment, the deep neural network uses a convolutional neural network with 5 convolutional layers, and the convolution window sizes are 24×205, 12×105, 6×45, 3×15 and 1×3, and the activation function selects a ReLU function, where the expression of the ReLU function is as follows:
ReLU(x)＝max{x,0}
wherein x represents any independent variable;
in this embodiment, training is performed on the deep neural network based on the neural network input data set and the neural network output data set, and the specific training is set to have a batch size of 20, and all samples are trained 200 times.
In a specific embodiment, each range gate in the airborne radar receives NK×1 space-time echo signals; and an analog-to-digital conversion unit in the airborne radar carries out analog-to-digital conversion on NK multiplied by 1 echo signals, so that the received data are digitized and stored.
In a specific embodiment, n 2 Taking 3-6, taking 4 adjacent distance gate data of the distance gate to be detected as training sample likelihood estimation to obtain a covariance matrix, then obtaining a clutter MVDR space-time spectrum through MVDR spectrum estimation, and inputting into a deep neural network for processing to obtain a space-time two-dimensional clutter spectrum.
Finally, calculating by combining the space-time two-dimensional clutter spectrum with a space-time dictionary to obtain more accurate range gate clutter covariance matrix data to be detected; and calculating a space-time self-adaptive weight by using the obtained clutter covariance matrix and a known target space-time steering vector, wherein the space-time self-adaptive weight is specifically as follows:
W＝μR -1 S
wherein ,for normalization constants, S represents a target space-time steering vector; s is S H Target space-time steering vector conjugate transpose.
And (3) applying the space-time self-adaptive weight to the range gate data to be detected to perform clutter suppression and target accumulation.
In the method described in this embodiment, the steps of modeling the clutter at the array element level of the airborne radar, constructing the deep neural network and training the deep neural network in steps S1, S2 and S3 are all offline processing, while the steps of processing the actual measurement data neural network, constructing the space-time adaptive processing weight and detecting the constant false alarm in steps S4, S5 and S6 are all online processing, so that the required operand is small, and the method is more suitable for practical engineering application.
It is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (7)
1. An airborne radar non-uniform clutter suppression method based on deep learning is characterized by comprising the following steps of: the method comprises the following steps:
s1: constructing an airborne radar array element level clutter model, and generating a space-time data set and a clutter covariance matrix data set of each range gate by simulating by combining various non-ideal factors;
s2: using n adjacent to the distance gate to be detected 1 The distance gate data is used as a training sample to generate an initial minimum variance undistorted response MVDR spectrum set; meanwhile, a clutter covariance matrix corresponding to a distance gate to be detected is obtained to generate a minimum variance distortion-free response MVDR spectrum set;
s3: constructing a deep neural network, taking the initial MVDR spectrum set obtained in the step S2 as a neural network input data set, taking the MVDR spectrum set generated by the clutter covariance matrix as a neural network output data set, and training the deep neural network until convergence;
s4: the airborne radar performs analog-to-digital conversion on the received actual observation data set, and then the distance gate to be detected after analog-to-digital conversion is adjacent to n 2 Estimating the range gate data to obtain a clutter MVDR space-time spectrum, and inputting a trained deep neural network for processing to obtain a space-time two-dimensional clutter spectrum;
s5: reconstructing a clutter covariance matrix of the range gate data to be detected by using a space-time two-dimensional clutter spectrum and a space-time steering dictionary, and calculating by combining a known target space-time steering vector to obtain a space-time self-adaptive weight value so as to realize clutter suppression and target accumulation treatment on the range gate data to be detected of the airborne radar;
s6: performing constant false alarm processing on the data processed by the clutter suppression and the accumulation targets to finish the detection of the moving targets;
the MVDR space-time spectrum has the following calculation formula:
wherein, R is clutter covariance matrix, phi is space-time guiding dictionary, specifically defined as:
wherein v is a space-time two-dimensional guide vector, f s Represents normalized spatial frequency, f d Expressed as normalized Doppler frequency, N s Representing the number of space frequency divisions, N d Indicating the number of doppler frequency divisions.
2. The deep learning-based airborne radar non-uniform clutter suppression method according to claim 1, wherein: such non-idealities include, but are not limited to, spatial error, clutter relief, and yaw angle.
3. The deep learning based airborne radar non-uniform clutter suppression method according to claim 2, wherein: the airspace error considers that the amplitude-phase error of the array element meets complex Gaussian distribution, and the variance is selected from 0 to 0.05, and the interval is 0.002; time domain decorrelation caused by clutter fluctuation obeys Gaussian distribution, and the standard deviation of wind speed is 0 to 0.1, and the interval is 0.01; the yaw angle range of the carrier is set to 0-5 degrees, and the interval is 0.2 degrees.
4. The deep learning-based airborne radar non-uniform clutter suppression method according to claim 1, wherein: when the MVDR space-time spectrum is obtained, the space frequency is divided into 4-5 times of the space domain array element number, and the Doppler frequency is divided into 4-5 times of the time domain pulse number.
5. The deep learning-based airborne radar non-uniform clutter suppression method of claim 4, wherein: the deep neural network adopts a convolutional neural network.
6. Deep learning-based airborne lightning as claimed in claim 5The method for suppressing the clutter is characterized by comprising the following steps of: in step S4, n 2 And 3-6, namely 3-6 adjacent distance gate data of the distance gate to be detected are taken as training samples to estimate and obtain clutter MVDR space-time spectrum.
7. The deep learning-based airborne radar non-uniform clutter suppression method of claim 6, wherein: the calculation formula of the space-time adaptive weight is as follows:
W＝μR -1 S
wherein ,for normalization constants, S represents a target space-time steering vector; s is S H Target space-time steering vector conjugate transposition;
and (3) applying the space-time self-adaptive weight to the range gate data to be detected to perform clutter suppression and target accumulation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011340629.2A CN112612006B (en) | 2020-11-25 | 2020-11-25 | Deep learning-based non-uniform clutter suppression method for airborne radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011340629.2A CN112612006B (en) | 2020-11-25 | 2020-11-25 | Deep learning-based non-uniform clutter suppression method for airborne radar |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112612006A CN112612006A (en) | 2021-04-06 |
| CN112612006B true CN112612006B (en) | 2023-08-22 |
Family
ID=75225180
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011340629.2A Active CN112612006B (en) | 2020-11-25 | 2020-11-25 | Deep learning-based non-uniform clutter suppression method for airborne radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112612006B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113219432B (en) * | 2021-05-14 | 2022-11-25 | 内蒙古工业大学 | Moving object detection method based on knowledge assistance and sparse Bayesian learning |
| CN113341391B (en) * | 2021-06-01 | 2022-05-10 | 电子科技大学 | Radar target multi-frame joint detection method in unknown environment based on deep learning |
| CN114492510B (en) * | 2021-12-30 | 2024-02-27 | 西北工业大学 | Intelligent sea clutter suppression method |
| CN114779198B (en) * | 2022-04-24 | 2022-09-23 | 中国人民解放军空军预警学院 | Conformal array airborne radar space-time clutter spectrum adaptive compensation and clutter suppression method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103364764A (en) * | 2013-06-25 | 2013-10-23 | 西安电子科技大学 | Airborne radar non-stationary clutter suppression method |
| CN105929371A (en) * | 2016-04-22 | 2016-09-07 | 西安电子科技大学 | Airborne radar clutter suppression method based on covariance matrix estimation |
| CN106772253A (en) * | 2016-11-25 | 2017-05-31 | 西安电子科技大学 | A kind of radar clutter suppression method under non-homogeneous clutter environment |
| WO2018045601A1 (en) * | 2016-09-09 | 2018-03-15 | 深圳大学 | Sparse recovery stap method for array error and system thereof |
| WO2018045567A1 (en) * | 2016-09-09 | 2018-03-15 | 深圳大学 | Robust stap method based on array manifold priori knowledge having measurement error |
| WO2018049595A1 (en) * | 2016-09-14 | 2018-03-22 | 深圳大学 | Admm-based robust sparse recovery stap method and system thereof |
| CN109116311A (en) * | 2018-09-19 | 2019-01-01 | 西安电子科技大学 | Knowledge based assists the clutter suppression method of sparse iteration covariance estimation |
-
2020
- 2020-11-25 CN CN202011340629.2A patent/CN112612006B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103364764A (en) * | 2013-06-25 | 2013-10-23 | 西安电子科技大学 | Airborne radar non-stationary clutter suppression method |
| CN105929371A (en) * | 2016-04-22 | 2016-09-07 | 西安电子科技大学 | Airborne radar clutter suppression method based on covariance matrix estimation |
| WO2018045601A1 (en) * | 2016-09-09 | 2018-03-15 | 深圳大学 | Sparse recovery stap method for array error and system thereof |
| WO2018045567A1 (en) * | 2016-09-09 | 2018-03-15 | 深圳大学 | Robust stap method based on array manifold priori knowledge having measurement error |
| WO2018049595A1 (en) * | 2016-09-14 | 2018-03-22 | 深圳大学 | Admm-based robust sparse recovery stap method and system thereof |
| CN106772253A (en) * | 2016-11-25 | 2017-05-31 | 西安电子科技大学 | A kind of radar clutter suppression method under non-homogeneous clutter environment |
| CN109116311A (en) * | 2018-09-19 | 2019-01-01 | 西安电子科技大学 | Knowledge based assists the clutter suppression method of sparse iteration covariance estimation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112612006A (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112612006B (en) | Deep learning-based non-uniform clutter suppression method for airborne radar | |
| CN109116311B (en) | Clutter suppression method based on knowledge-aided sparse iteration covariance estimation | |
| CN104035095B (en) | Based on the low level wind shear velocity estimation method of optimal processor during sky | |
| CN108761419B (en) | Low-altitude wind shear wind speed estimation method based on self-adaptive processing of combined space-time main channel | |
| CN109212500B (en) | High-precision KA-STAP (K-ary-based adaptive-noise) covariance matrix estimation method based on sparse reconstruction | |
| CN109324315B (en) | Space-time adaptive radar clutter suppression method based on double-layer block sparsity | |
| CN106772253B (en) | Radar clutter suppression method under non-uniform clutter environment | |
| CN111913157B (en) | Sea clutter suppression method based on radar signal space-time decorrelation model | |
| CN107167783B (en) | Sparse reconstruction method of conformal array clutter covariance matrix | |
| CN112612005B (en) | Radar main lobe interference resistance method based on deep learning | |
| CN105223560A (en) | Based on the airborne radar object detection method of the sparse recovery of clutter pitching azimuth spectrum | |
| CN103176168A (en) | Short-range cluster cancellation method for airborne non-side-looking array radar | |
| CN106443633A (en) | Shipborne high frequency ground wave radar sea clutter time domain suppression method | |
| CN110161489A (en) | A kind of strong and weak signals direction-finding method based on pseudo- frame | |
| CN115685096B (en) | Secondary radar side lobe suppression method based on logistic regression | |
| CN110554391A (en) | low-altitude wind shear wind speed estimation method based on DDD-GMB | |
| CN111220986A (en) | Echo power screening and DLCD (digital Living control computer) assisted low-altitude wind shear wind speed estimation method | |
| CN107748364A (en) | Low wind field speed estimation method based on contraction multistage wiener filter | |
| CN113376607B (en) | Airborne distributed radar small sample space-time self-adaptive processing method | |
| CN108761417B (en) | Airborne radar clutter suppression method based on knowledge-aided maximum likelihood | |
| CN111796288B (en) | Clutter frequency spectrum compensation technology-based three-coordinate radar moving target processing method | |
| CN111650574B (en) | Underwater space-time self-adaptive processing method and system based on sparse recovery | |
| CN103885042B (en) | Based on the array element error estimation of clutter subspace | |
| CN110376561B (en) | Time domain dimension reduction multi-fast-beat iterative array element amplitude-phase error estimation method | |
| CN104914421A (en) | Low-altitude wind shear wind speed estimating method based on sum-difference beam |
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 |