CN116466313A - Radar expansion target detection method based on model-driven deep neural network - Google Patents

Radar expansion target detection method based on model-driven deep neural network Download PDF

Info

Publication number
CN116466313A
CN116466313A CN202310303790.XA CN202310303790A CN116466313A CN 116466313 A CN116466313 A CN 116466313A CN 202310303790 A CN202310303790 A CN 202310303790A CN 116466313 A CN116466313 A CN 116466313A
Authority
CN
China
Prior art keywords
target
network
detection
layer
radar
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
Application number
CN202310303790.XA
Other languages
Chinese (zh)
Inventor
赵文静
崔国龙
汪翔
汪育苗
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.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
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 University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202310303790.XA priority Critical patent/CN116466313A/en
Publication of CN116466313A publication Critical patent/CN116466313A/en
Pending legal-status Critical Current

Links

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
    • G01S7/41—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
    • G01S7/417—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 involving the use of neural networks
    • 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
    • G01S13/04—Systems determining presence of a target
    • 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
    • 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
    • G01S7/35—Details of non-pulse systems
    • G01S7/352—Receivers
    • G01S7/354—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
    • G01S7/41—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
    • G01S7/414—Discriminating targets with respect to background clutter
    • 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
    • G06N3/048—Activation functions
    • 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
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention provides a radar expansion target detection method based on a model-driven deep neural network, and belongs to the technical field of radar target detection. Firstly, modeling a radar expansion target detection problem as an optimization problem of a binary variable under a likelihood ratio detection criterion; then, constructing a deep expansion network driven by the combination of the model information and the data information, performing offline training, and setting a decision threshold according to the false alarm probability requirement; and finally, in the online detection stage, decision judgment is carried out in real time, so that target detection is realized. The method utilizes the model information and the data information simultaneously, and has stronger generalization capability. The simulation experiment shows that the method has higher detection performance than the existing typical detection method by taking the detection probability as an evaluation performance index, can ensure the constant false alarm characteristic, and is suitable for practical application.

Description

Radar expansion target detection method based on model-driven deep neural network
Technical Field
The invention belongs to the technical field of radar target detection, and particularly relates to a radar extended target detection method in a non-Gaussian clutter environment.
Background
Radar detection performance is limited primarily by clutter, noise and other disturbances. In particular, for high resolution radar systems, the statistical properties of the clutter deviate from the gaussian distribution and exhibit significant non-gaussian properties. Meanwhile, when the range resolution unit of the radar is smaller than the target size, the target echo signal occupies a plurality of range units, which is called a distributed target or an extended target. For distance extended targets, conventional point target detection methods suffer from shadowing effects of neighboring distance units and cannot effectively accumulate multiple distance unit energies, resulting in degraded detection performance, even complete failure. Therefore, it is important to improve radar extended target detection performance under non-gaussian clutter.
Many research institutions at home and abroad develop a radar expansion target detection method under non-Gaussian clutter. Common detector design criteria include the Generalized Likelihood Ratio (GLRT) criteria, wald and Rao detection criteria. For example, for composite Gaussian distribution clutter, the United states naval research laboratory designed the SDD-GLRT algorithm using density information of target scattering points (K. Gerlach, "Spatially distributed target detection in non-Gaussian clutch," IEEE Trans. Aerosp. Electron. Syst., vol.35, no.3, pp.926-934, july 1999.); under the GLRT detection criterion, the naval aviation astronomy university utilizes ordered statistics to design an OS-GLRT detection algorithm. (Y.He, T.Jian, F.Su, et al, "Novel range-spread target detectors in non-Gaussian clutch," IEEE Trans. Aerosp. Electron. Syst., vol.46, no.3, pp.1312-1328,2010.)
Most of the above detection methods are designed under the GLRT rule, and although GLRT has a theoretical basis, the GLRT does not have an optimal structure due to the absence of a uniform maximum potential. In addition, most of the detection algorithms existing at present are designed based on statistical theory, and the performance of the detection algorithms is seriously dependent on the statistical characteristics of targets and clutter. In an actual scene, when the preset target/clutter statistical characteristic is not matched with the actual situation, the detection performance of the method is seriously degraded. Therefore, the detection algorithm with better performance and strong robustness based on other strategies has important value in the radar target detection field.
The deep learning technology has strong data processing capability, can learn effective information contained in data, and is gradually applied to the radar target detection field and achieves a certain effect. However, most existing deep neural networks are "black box" processes, and lack interpretability. In addition, conventional deep networks contain a large number of parameters to learn, require a large amount of training data, and the performance is severely dependent on the data. For the field of radar signal processing, and in particular for non-cooperative targets, annotation data containing the target is limited. Therefore, how to combine the statistical signal processing theory with the deep learning technology to design a new radar expansion target detection algorithm under the condition of small sample data is worth deeply researching.
Disclosure of Invention
In order to solve the problems, the invention provides a radar expansion target detection method applicable to a non-Gaussian clutter environment and based on a model-driven deep neural network. Firstly, modeling a radar expansion target detection problem as an optimization problem of a binary variable under an LRT detection criterion; then, constructing a deep expansion network driven by the combination of the model information and the data information, performing offline training, and setting a decision threshold according to the false alarm probability requirement; and finally, in the online detection stage, decision judgment is carried out in real time to judge whether the target signal exists or does not exist, so that target detection is realized.
The technical scheme of the invention is as follows:
a radar expansion target detection method based on a model-driven deep neural network comprises the following steps:
step 1: modeling a radar expansion target detection problem as an optimization problem under a likelihood ratio detection criterion;
the application scene is as follows: the distributed MIMO radar comprises M transmitting array elements at a transmitting end, K receiving array elements at a receiving end, namely MK paths are shared from the transmitting end to the receiving end, each antenna transmits L pulses in a coherent processing interval, and each antenna at the transmitting end transmits waveforms which are mutually orthogonal;
let the target echo signal occupy H distance units, let y mk,h Representing the mk path, the received signal vector for the h range bin, and radar expansion target detection under non-Gaussian clutter is represented as:
wherein alpha is mk,h Representing the sum of the target scattering and channel propagation effects of the mk path, the h distance unit;
p mk for the corresponding target steering vector, denoted p mk ＝[1,exp(j2πf mk T r ),…,exp(j2π(L-1)f mk T r )] T ,f mk For Doppler shift of target, T r For pulse repetition time c mk,h For clutter vectors, a complex Gaussian model is generally used for modeling, and is expressed as a slow-varying component tau mk,h And a fast-varying component g mk,h The product of (a), i.eWherein g mk,h Is subject to complex gaussian distribution with mean value zero variance sigma, H 0 Indicating no target, H 1 Indicating that there is a target;
for the above detection problem, the likelihood ratio detection criteria are:
wherein eta is a decision threshold based on likelihood ratio detection criterion, alpha mk Representing the sum of the target scattering and channel propagation effects of the mk-th path, τ mk Representing the slowly varying component, y, of the mk-th path mk A received signal vector representing the mk th path;
the discrete binary variable ω∈ {0,1} is introduced, and the above detection criteria are expressed as:
wherein ω=0 indicates that the target is absent, and ω=1 indicates that the target is present;
step 2: building a model information and data jointly driven deep expansion network
Aiming at the optimization problem, a conventional projection gradient algorithm is developed into a deep neural network; each expansion layer in the network is of a fully-connected neural network structure and consists of an input layer, a hidden layer and an output layer;
the inputs to the layer t network are:
wherein,,is an estimate of ω for the upper layer network, +.>Is the dimension-increasing vector of the layer t network for increasing the input layer dimension,/for>Is an optimized objective function f b Gradient with respect to ω;
the hidden layer of the layer t network is expressed as:
h t ＝f ReLU (W 1 (t)x(t)+b 1 (t))
wherein f ReLU (. Cndot.) represents the nonlinear activation function of the hidden layer;
the output layer of the layer t network is expressed as:
wherein f sigmoid (. Cndot.) nonlinear activation function, { W 1 (t),b 1 (t),W 2 (t),b 2 (t),W 3 (t),b 3 (t) } is a layer t network parameter;
step 3: setting a decision threshold
Firstly preprocessing non-target data in a training set, inputting the non-target data into a trained network, ordering the output results of the network in a descending order, and finally determining a decision threshold according to the false alarm probability;
step 4: on-line detection judgment
Preprocessing the test set, inputting the test set into a trained network, and judging according to the following formula;
if the network output value is larger than the threshold, the target is judged to exist, otherwise, the target is judged to not exist.
Further, the decision threshold set in the step 3Wherein->After inputting the non-target data in the training set to the network, the +.>P as a result of fa Is the preset false alarm probability.
The beneficial effects of the invention are as follows:
the invention provides a radar expansion target detection method based on a model-driven deep neural network. The network has interpretability and can quickly converge with small sample data compared to the conventional deep neural network mentioned in the technical background. Compared with the traditional detection method based on the statistical model, the method utilizes model information and data information simultaneously, and has stronger generalization capability. The simulation experiment shows that the method has higher detection performance than the existing detection method in a non-Gaussian clutter environment by taking the detection probability as an evaluation performance index, can ensure the constant false alarm characteristic, and is suitable for practical application.
Drawings
FIG. 1 is a block diagram of a layer t network architecture of a deep deployment network.
FIG. 2 is a process flow diagram of a detection method based on a model-data joint driving deep expansion network.
Fig. 3 is a diagram of a radar system configuration used in a simulation experiment.
Fig. 4 is a graph showing the variation of the detection probability with the signal to noise ratio according to the present invention and the classical detection algorithm.
Fig. 5 is a graph showing the detection probability variation with false alarm probability according to the present invention and the classical detection algorithm.
Detailed Description
The invention will be described in further detail with reference to the accompanying drawings and the detailed description.
The radar expansion target detection method based on the model-driven deep neural network comprises the following steps:
1. modeling radar extended target detection problem as optimization problem
For the distributed MIMO radar provided with M transmitting array elements and K receiving array elements, the transmitting end and the receiving end are respectively provided with antennas which are far apart. It is assumed that each antenna transmits L pulses within one coherent processing interval, and each antenna at the transmitting end transmits waveforms orthogonal to each other.
Let y assume that the target echo signal occupies H distance units mk,h Represents the mk path, the received signal vector for the h range bin. The radar expansion target detection problem under non-Gaussian clutter is expressed as
Wherein alpha is mk,h Reflecting the effect of the diffusion of the targets and the propagation of the channels in the mk-th path, p mk For the corresponding steering vector, denoted p mk ＝[1,exp(j2πf mk T r ),…,exp(j2π(L-1)f mk T r )] T ,f mk For Doppler shift of target, T r Is the pulse repetition time. c mk,h For clutter vectors, a complex Gaussian model is generally used for modeling, and is expressed as a slow-varying component tau mk,h And a fast-varying component g mk,h The product of (a), i.eWherein g mk,h Is subject to a complex gaussian distribution with zero mean and zero variance Σ.
For the above detection problem, the likelihood ratio detection criteria are expressed as
Where η is a decision threshold based on likelihood ratio detection criteria.
The discrete binary variable omega epsilon {0,1} is introduced, and the above detection criteria are expressed as
Estimating the scattering coefficient of the target signal and the texture component of the clutter by using the maximum likelihood estimation method, wherein the problem is further equivalently expressed as
Where ω=0 indicates that the target is not present, and ω=1 indicates that the target is present.
2. Constructing deep-expansion network
Aiming at the optimization problem, a conventional projection gradient algorithm is developed into a deep neural network. As shown in fig. 1, each expansion layer in the network is a fully-connected neural network structure, and is composed of an input layer, a hidden layer and an output layer.
The inputs to the layer t network are noted as
Wherein the method comprises the steps ofIs an estimate of ω for the upper layer network, +.>Is the dimension vector of the layer t network for increasing the dimension of the input layer,/for the layer t network>Is an optimized objective function f b Gradient with respect to ω.
Implicit layer representation of a layer t network is
h t ＝f ReLU (W 1 (t)x(t)+b 1 (t))
Wherein f ReLU (. Cndot.) represents the nonlinear activation function of the hidden layer.
The output layer of the layer t network is expressed as
Wherein f sigmoid (. Cndot.) nonlinear activation function, { W 1 (t),b 1 (t),W 2 (t),b 2 (t),W 3 (t),b 3 (t) } is a layer t network parameter.
3. Setting a decision threshold
As shown in fig. 2, letFor non-target data in training set, for +.>Pretreatment to obtain->Inputting the data into a trained network, and recording an output result as ψ= { ω z Z=1, …, Z }. The network output results are ordered in descending order to obtain +.>According to false alarm probability P fa Get->The detection statistics are used as decision threshold, expressed as +.>
4. On-line detection judgment
As shown in fig. 2, in the online detection phase, the test set is preprocessed, then input into the trained network, and decision is made according to the following formula.
If the network output value is larger than the threshold in the step 3, the target is judged to exist, otherwise, the target is judged to not exist.
Specific examples:
parameter setting: as shown in fig. 3, a specific embodiment of the present invention is performed under a distributed MIMO radar system. The number of the transmitting antenna array elements and the number of the receiving antenna array elements are assumed to be 2, wherein the included angles between 2 transmitting antennas and the target are 0 degrees and 65 degrees, and the included angles between 2 receiving antennas and the target are-30 degrees and 40 degrees. Assuming that the number of transmit pulses is m=10 in one coherent processing interval, the transmit pulses are repeatedThe complex frequency is 500Hz, the carrier frequency is set to be 1GHz, the moving speed of a measured target is 108km/h, and the number of distance units occupied by a target echo signal is 6. The clutter amplitude statistical characteristic is modeled as K distribution, the texture component obeys the gamma distribution, the speckle component obeys the mean value to be zero, and the covariance matrix is sigma=sigma 0 +I M Is complex gaussian distribution of sigma 0 The modeling is in the form of an index,
wherein ρ is a first-order delay correlation coefficient of the clutter, and is set to ρ=0.9;for the noise ratio, set to 10dB; f (f) dc The Doppler frequency is normalized for clutter and set to 0.05.
Embodiment one:
let signal-to-noise ratio (SCR) be-21 dB to 0dB and false alarm probability be 0.001. It is assumed that the target echo signal energy is scattered uniformly in each range bin and that the target scattering amplitude follows a gaussian distribution. FIG. 4 shows the detection probability of the prior art and the present invention as a function of SCR. As shown in FIG. 4, the technology of the invention has higher detection probability, and the detection performance of the radar expansion target in the non-Gaussian clutter environment is obviously improved.
Embodiment two:
assume that the false alarm probability is 10 -3 By a linear variation in the range of 1, the signal-to-noise ratio SCR is set to-12 dB. Fig. 5 shows the detection probability of the prior art and the present invention as a function of the false alarm probability. The result shows that the detection probability of the invention is obviously improved under different false alarm probabilities.
Embodiment III:
table 1 below shows the actual false alarm probability and the variation of the preset false alarm probability under different clutter shape parameters, wherein the clutter shape parameters are set to 1 and 5, and the preset false alarm probability is 10 -5 ,10 -4 ,10 -3 ,10 -2 . As shown in the table, the actual false alarm probability of the present technology is almost the same as the set situation. In addition, the false alarm probability changes slightly under different clutter shape parameters, which indicates that the technology can ensure constant false alarm probability about the clutter shape parameters.
TABLE 1

Claims (2)

1. A radar expansion target detection method based on a model-driven deep neural network comprises the following steps:
step 1: modeling a radar expansion target detection problem as an optimization problem under a likelihood ratio detection criterion;
the application scene is as follows: the distributed MIMO radar comprises M transmitting array elements at a transmitting end, K receiving array elements at a receiving end, namely MK paths are shared from the transmitting end to the receiving end, each antenna transmits L pulses in a coherent processing interval, and each antenna at the transmitting end transmits waveforms which are mutually orthogonal;
let the target echo signal occupy H distance units, let y mk,h Representing the mk path, the received signal vector for the h range bin, and radar expansion target detection under non-Gaussian clutter is represented as:
wherein alpha is mk,h Representing the sum of the target scattering and channel propagation effects of the mk path, the h distance unit; p is p mk For the corresponding target steering vector, denoted p mk ＝[1,exp(j2πf mk T r ),,exp(j2π(L-1)f mk T r )] T ,f mk For Doppler shift of target, T r For pulse repetition time c mk,h For clutter vectors, a complex Gaussian model is generally used for modeling, and is expressed as a slow-varying component tau mk,h And a fast-varying component g mk,h The product of (a), i.eWherein g mk,h Is subject to complex gaussian distribution with mean value zero variance sigma, H 0 Indicating no target, H 1 Indicating that there is a target;
for the above detection problem, the likelihood ratio detection criteria are:
wherein eta is a decision threshold based on likelihood ratio detection criterion, alpha mk Representing the sum of the target scattering and channel propagation effects of the mk-th path, τ mk Representing the slowly varying component, y, of the mk-th path mk A received signal vector representing the mk th path;
the discrete binary variable ω∈ {0,1} is introduced, and the above detection criteria are expressed as:
wherein ω=0 indicates that the target is absent, and ω=1 indicates that the target is present;
step 2: building a model information and data jointly driven deep expansion network
Aiming at the optimization problem, a conventional projection gradient algorithm is developed into a deep neural network; each expansion layer in the network is of a fully-connected neural network structure and consists of an input layer, a hidden layer and an output layer;
the inputs to the layer t network are:
wherein,,is an estimate of ω for the upper layer network, +.>Is the dimension-increasing vector of the layer t network for increasing the input layer dimension,/for>Is an optimized objective function f b Gradient with respect to ω;
the hidden layer of the layer t network is expressed as:
h t ＝f ReLU (W 1 (t)x(t)+b 1 (t))
wherein f ReLU (. Cndot.) represents the nonlinear activation function of the hidden layer;
the output layer of the layer t network is expressed as:
wherein f sigmoid (. Cndot.) nonlinear activation function, { W 1 (t),b 1 (t),W 2 (t),b 2 (t),W 3 (t),b 3 (t) } is a layer t network parameter;
step 3: setting a decision threshold
Firstly preprocessing non-target data in a training set, inputting the non-target data into a trained network, ordering the output results of the network in a descending order, and finally determining a decision threshold according to the false alarm probability;
step 4: on-line detection judgment
Preprocessing the test set, inputting the test set into a trained network, and judging according to the following formula;
if the network output value is larger than the threshold, the target is judged to exist, otherwise, the target is judged to not exist.
2. The method for detecting radar expansion targets based on model-driven deep neural network as set forth in claim 1, wherein said decision threshold set in step 3Wherein->After inputting the non-target data in the training set to the network, the +.>P as a result of fa Is the preset false alarm probability.
CN202310303790.XA 2023-03-27 2023-03-27 Radar expansion target detection method based on model-driven deep neural network Pending CN116466313A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310303790.XA CN116466313A (en) 2023-03-27 2023-03-27 Radar expansion target detection method based on model-driven deep neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310303790.XA CN116466313A (en) 2023-03-27 2023-03-27 Radar expansion target detection method based on model-driven deep neural network

Publications (1)

Publication Number Publication Date
CN116466313A true CN116466313A (en) 2023-07-21

Family

ID=87174334

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310303790.XA Pending CN116466313A (en) 2023-03-27 2023-03-27 Radar expansion target detection method based on model-driven deep neural network

Country Status (1)

Country Link
CN (1) CN116466313A (en)

Similar Documents

Publication Publication Date Title
US5373456A (en) Expert system for assessing accuracy of models of physical phenomena and for selecting alternate models in the presence of noise
CN108398678A (en) A kind of quick self-adapted method for tracking target of centralization MIMO radar
CN114363949B (en) Abnormal data detection method for UWB positioning system
CN111460597A (en) Radar station distribution method based on improved multi-target particle swarm optimization algorithm
Ristic et al. Bernoulli track‐before‐detect filter for maritime radar
CN112036239A (en) Radar signal working mode identification method and system based on deep learning network
Lo et al. Radial basis function neural network for direction-of-arrivals estimation
Sun et al. Labelled multi‐Bernoulli filter with amplitude information for tracking marine weak targets
Davey Probabilistic multihypothesis trackerwith an evolving poisson prior
CN113608193A (en) Radar multi-target distance and speed estimation method based on UNet
CN108196238B (en) Clutter map detection method based on adaptive matched filtering under Gaussian background
Sun et al. Deep learning based preamble detection and TOA estimation
CN111198366B (en) Method for quickly selecting finite array elements under distributed MIMO radar multitasking
CN116466313A (en) Radar expansion target detection method based on model-driven deep neural network
CN116299289A (en) MIMO radar expansion target self-adaptive detection method based on information geometry
Ristic et al. Gaussian mixture multitarget–multisensor Bernoulli tracker for multistatic sonobuoy fields
CN110441761A (en) Multi-sources Information Fusion Method based on the detection of distributed buoy
CN112986975B (en) Distance weighting-based passive radar network centralized detection method
Lu et al. Distributed radar multi-frame detection with least squares quantization
CN108387879A (en) Value detection method in clutter map unit based on adaptive normalization matched filtering
Wan et al. A range-Doppler-angle estimation method for passive bistatic radar
CN115015831B (en) Large-scale array target azimuth estimation method under combined influence of impulse noise and non-uniform noise
Silantyev et al. Analysis of Efficiency of Long-Term Multi-Scanning Radar Observation Methods
Yang et al. Waveform Design For Track-Before-Detect-Based Cognitive Radars
Deng et al. A multi-layer neural network classifier for radar clutter

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