CN111582026A - Sparse drive ALE-based underwater target detection method and system of support vector machine - Google Patents
Sparse drive ALE-based underwater target detection method and system of support vector machine Download PDFInfo
- Publication number
- CN111582026A CN111582026A CN202010242406.6A CN202010242406A CN111582026A CN 111582026 A CN111582026 A CN 111582026A CN 202010242406 A CN202010242406 A CN 202010242406A CN 111582026 A CN111582026 A CN 111582026A
- Authority
- CN
- China
- Prior art keywords
- entropy
- ale
- support vector
- vector machine
- sparse
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 45
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 44
- 238000001228 spectrum Methods 0.000 claims abstract description 67
- 238000012545 processing Methods 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 21
- 230000003044 adaptive effect Effects 0.000 claims description 35
- 238000012549 training Methods 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 23
- 238000012360 testing method Methods 0.000 claims description 14
- 238000010276 construction Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 8
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 5
- 230000003111 delayed effect Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims 1
- 230000007613 environmental effect Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 7
- 230000006872 improvement Effects 0.000 description 6
- 238000001914 filtration Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 108700041286 delta Proteins 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003623 enhancer Substances 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention provides an underwater target detection method and system based on a sparse drive ALE support vector machine, wherein the method comprises the following steps: carrying out beam forming on the array receiving data of the underwater unmanned platform; forming a beam signal; carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals; the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics; and inputting the two-dimensional entropy characteristics into a pre-trained support vector machine, and outputting a judgment result of target detection. The underwater target detection method reduces steady-state errors of the ALE, improves the line spectrum output signal-to-noise ratio of the sparsely driven ALE, improves the detection probability of line spectrum weak targets, and improves the environmental adaptability of the underwater unmanned platform for target detection under the conditions of non-uniform noise background and wide-band strong interference.
Description
Technical Field
The invention relates to the field of underwater target detection, in particular to an underwater target detection method and system based on a sparse drive ALE support vector machine.
Background
With the development of vibration reduction and noise reduction technologies, the radiation noise level of ships is greatly reduced, and great challenges are brought to the passive detection of low-noise and quiet underwater targets by an underwater unmanned platform. The line spectrum of the underwater target generally has better phase stability and higher intensity, but the line spectrum of the weak target is easily submerged in the continuous spectrum component under the influence of the variance of the continuous spectrum of the noise, and the broadband interference and the background noise in the signal need to be filtered.
A conventional Adaptive Line spectrum Enhancer (ALE) can enhance a Line spectrum signal and suppress broadband noise, but the performance of the ALE is limited due to a steady-state error problem of Adaptive filtering; the energy detection method based on the automatic threshold is seriously influenced by non-uniform noise background, strong interference and the like, and the weak target detection capability is difficult to meet the environment self-adaptability required by the underwater unmanned platform.
Disclosure of Invention
The invention aims to overcome the technical defects and provide an underwater target detection method based on a sparse drive ALE support vector machine, so as to overcome the defect of weak target detection capability of an underwater unmanned platform in complex environments such as non-uniform noise background, broadband strong interference and the like and improve the detection probability of line spectrum weak targets.
In order to achieve the above object, embodiment 1 of the present invention provides an underwater target detection method based on a support vector machine for sparsely driving ALE, the method including:
carrying out beam forming on the array receiving data of the underwater unmanned platform; forming a beam signal;
carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals;
the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics;
and inputting the two-dimensional entropy characteristics into a pre-trained support vector machine, and outputting a judgment result of target detection.
As an improvement of the foregoing method, the performing sparse drive ALE processing on the beam signal to obtain an enhanced beam domain signal specifically includes:
step S1) initial frequency domain weight w of adaptive filterF(0) Expressed as:
wF(0)=[wF,0(0),wF,1(0),...,wF,L-1(0)]T
let n represent the number of iterations, with an initial value of 0;
step S2), the time domain output y (n) of the nth iteration adaptive filter is expressed as:
wherein ,wF(n)=[wF,0(n),wF,1(n),...,wF,L-1(n)]TThe frequency domain weight of the nth iteration adaptive filter is obtained; x (n- Δ) is the time delayed beam signal:
x(n-Δ)=[x(n-Δ),x(n-Δ-1),...,x(n-Δ-L+1)]T
wherein, x (n) is a reference signal, Δ is a time delay, and L is a filter order;
step S3) calculates an output estimation error e (n) of the nth iterative adaptive filter:
e(n)=x(n)-y(n)
step S4) calculating the frequency domain weight w of the (n +1) th adaptive filterF(n+1):
According to the principle of gradient descent, by1/2Norm regularized sparsely driven ALE processing, frequency domain weight w of the (n +1) th iterative adaptive filterF(n +1) is:
wherein ,representing bitwise multiplication of vectors, | · non-woven phosphor-3/2A power of-3/2, μ is the step size, ρ ═ μ κ is a constant, and κ is a constant, representing bitwise modulo of the vector; x is the number ofF(n- Δ) is a discrete Fourier transform of x (n- Δ);
step S5) if | | wF(n+1)-wF(n) | | is less than the threshold, then step S6 is entered, otherwise, after n adds 1, step S2 is entered;
step S6), calculating a final enhanced beam domain signal y (n +1) after the sparse drive ALE processing:
wherein x (n +1- Δ) ═ x (n +1- Δ), x (n- Δ),.., x (n- Δ -L +2)]T,xF(n +1- Δ) is the discrete Fourier transform of x (n +1- Δ).
As an improvement of the above method, the converting the enhanced beam domain signal into a beam sound spectrum through discrete fourier transform, calculating an entropy characteristic of the beam sound spectrum, and drawing an entropy characteristic curve specifically includes:
converting the enhanced beam domain signal y (n +1) into a beam acoustic spectrum through discrete Fourier transform;
setting the wave beam sound spectrum of the mth time window to contain N frequency points with the amplitudes of c1,c2,…cNThe weight p of the nth frequency pointnThe ratio of the amplitude to the sum of the amplitudes of all the frequency points in the time window is as follows:
according to the amplitude weight, the Shannon Entropy value Encopy of the mth time windowmComprises the following steps:
Entropymis the entropy property of the beam sound spectrum;
control with time window m as abscissamAn entropy characteristic curve is plotted for the ordinate.
As an improvement of the above method, the calculating a mean value and a standard deviation of the entropy characteristic curve to construct the two-dimensional entropy feature specifically includes:
calculating the mean E of the entropy curvesaveAnd standard deviation Estd:
Wherein M is the number of entropy values in the entropy characteristic curve;
average value E of entropy curveaveAnd standard deviation EstdAnd forming a two-dimensional feature vector as a two-dimensional entropy feature.
As an improvement of the above method, the method further comprises: the step of training the support vector machine specifically comprises:
establishing a training set, comprising: receiving data and a tag thereof by an array of the underwater unmanned platform;
carrying out beam forming on array receiving data of the underwater unmanned platform of the training set; forming a beam signal; carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals; the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve to form a two-dimensional entropy characteristic;
selecting a Gaussian kernel function as a binary classifier of a support vector machine; when its output is 0, it represents hypothesis test H0: no target is present; its output is 1When, represents hypothesis test H1: a presence target;
inputting the two-dimensional entropy characteristics of the training set into a support vector machine for training, and optimizing the parameters of the support vector machine according to the characteristic distribution of the training data.
Embodiment 2 of the present invention provides an underwater target detection system of a support vector machine based on sparse drive ALE, characterized in that the system comprises: the system comprises an underwater unmanned platform, a trained support vector machine for target detection, a beam forming module, a sparse drive ALE processing module, a two-dimensional entropy characteristic construction module and a target detection module;
the beam forming module is used for carrying out beam forming on the data received by the array of the underwater unmanned platform; forming a beam signal;
the sparse drive ALE processing module is used for carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals;
the two-dimensional entropy characteristic construction module is used for converting the enhanced beam domain signals into beam sound spectrums through discrete Fourier transform, calculating entropy characteristics of the beam sound spectrums and drawing entropy characteristic curves; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics;
and the target detection module is used for inputting the two-dimensional entropy characteristics into a pre-trained support vector machine and outputting a judgment result of target detection.
As an improvement of the above system, a specific implementation process of the sparse drive ALE processing module is as follows:
step S1) initial frequency domain weight w of adaptive filterF(0) Expressed as:
wF(0)=[wF,0(0),wF,1(0),...,wF,L-1(0)]T
let n represent the number of iterations, with an initial value of 0;
step S2), the time domain output y (n) of the nth iteration adaptive filter is expressed as:
wherein ,wF(n)=[wF,0(n),wF,1(n),...,wF,L-1(n)]TThe frequency domain weight of the nth iteration adaptive filter is obtained; x (n- Δ) is the time delayed beam signal:
x(n-Δ)=[x(n-Δ),x(n-Δ-1),...,x(n-Δ-L+1)]T
wherein, x (n) is a reference signal, Δ is a time delay, and L is a filter order;
step S3) calculates an output estimation error e (n) of the nth iterative adaptive filter:
e(n)=x(n)-y(n)
step S4) calculating the frequency domain weight w of the (n +1) th adaptive filterF(n+1):
According to the principle of gradient descent, by1/2Norm regularized sparsely driven ALE processing, frequency domain weight w of the (n +1) th iterative adaptive filterF(n +1) is:
wherein ,representing bitwise multiplication of vectors, | · non-woven phosphor-3/2A power of-3/2, μ is the step size, ρ ═ μ κ is a constant, and κ is a constant, representing bitwise modulo of the vector; x is the number ofF(n- Δ) is a discrete Fourier transform of x (n- Δ);
step S5) if | | wF(n+1)-wF(n) | | is less than the threshold, then step S6 is entered, otherwise, after n adds 1, step S2 is entered;
step S6), calculating a final enhanced beam domain signal y (n +1) after the sparse drive ALE processing:
wherein x (n +1- Δ) ═ x (n +1- Δ), x (n- Δ),.., x (n- Δ -L +2)]T,xF(n +1- Δ) is the discrete Fourier transform of x (n +1- Δ).
As an improvement of the above system, the specific implementation process of the two-dimensional entropy feature construction module is as follows:
converting the enhanced beam domain signal y (n +1) into a beam acoustic spectrum through discrete Fourier transform;
setting the wave beam sound spectrum of the mth time window to contain N frequency points with the amplitudes of c1,c2,…cNThe weight p of the nth frequency pointnThe ratio of the amplitude to the sum of the amplitudes of all the frequency points in the time window is as follows:
according to the amplitude weight, the Shannon Entropy value Encopy of the mth time windowmComprises the following steps:
Entropymis the entropy property of the beam sound spectrum;
control with time window m as abscissamDrawing an entropy characteristic curve for a vertical coordinate;
calculating the mean E of the entropy curvesaveAnd standard deviation Estd:
Wherein M is the number of entropy values in the entropy characteristic curve;
average value E of entropy curveaveAnd standard deviation EstdForming two-dimensional feature vectors as two-dimensional entropyAnd (5) characterizing.
The invention has the advantages that:
by the underwater target detection method provided by the invention, the steady-state error of the ALE is reduced, the line spectrum output signal-to-noise ratio of the sparsely driven ALE is improved, the detection probability of a line spectrum weak target is improved, and the environmental adaptability of the underwater unmanned platform for target detection is improved under the non-uniform noise background and the strong broadband interference.
Drawings
FIG. 1 is a flow chart of the method of the present invention for underwater target detection based on a sparsely driven ALE support vector machine;
FIG. 2 is a flow diagram of the sparse drive ALE process of the present invention;
FIG. 3 is a steady state error comparison of a simulated conventional ALE versus a sparse drive ALE;
FIG. 4 is a comparison of the amplitude-frequency characteristics of a simulated conventional ALE weighting vector and an optimal wiener weight;
FIG. 5 is a comparison of amplitude-frequency characteristics of a simulated sparse drive ALE weighted vector and an optimal wiener weight;
FIG. 6 is a graph of simulated conventional ALE output LOFAR;
FIG. 7 is a graph of simulated sparse drive ALE output LOFAR;
FIG. 8 is a schematic diagram of the training and testing results of the marine test data of the present invention.
Detailed Description
The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, embodiment 1 of the present invention provides an underwater target detection method based on sparse drive ALE supervised learning, including the following steps:
step 1) carrying out beam forming on array receiving data of an underwater unmanned platform to form beam signals;
the array of the underwater unmanned platform comprises M array elements, and the received data is α0(n),…αm(n)…αM-1(n); the beamformed signals are:
x(n)=s(n)+v(n)
where s (n) is a line spectrum signal and v (n) is broadband noise.
Step 2), sparse drive ALE processing is carried out on the beam signals to obtain processed beam signals y (n);
sparse drive ALE is to consider the sparsity of line spectral targets in the frequency domain, since l1/2Sparsity of norm is better than that of common l1Norm, hence sparseness l1/2The norm regularization model is combined with the conventional ALE, frequency domain weighting vectors are thinned through iteration of frequency domain adaptive filtering, and a more sparse line spectrum can be obtained on a weighted output signal in a frequency domain, so that the steady-state error in the conventional ALE is reduced, and the line spectrum output signal-to-noise ratio is improved.
Conventional ALE is implemented based on Least Mean Square (LMS) adaptive filtering, given an output y (n) expressed as:
y(n)=wT(n)x(n-Δ)
wherein x (n- Δ) ═ x (n- Δ), x (n- Δ -1),.., x (n- Δ -L +1)]TIs the input sequence of the filter, and the output of the adaptive filter estimates the error as e (n) ═ x (n) -y (n), where x (n) is the reference signal, then the cost function of the conventional ALE according to the LMS criterion is:
J(n)=e2(n)
sparsely driven ALE is a combination of a sparse regularization model with a conventional ALE, since 0<p<1 time,/pNorm regularization ratio l1Norm regularization is more sparse, so sparse driving ALE introduces l of frequency domain weighting vector in cost function of adaptive line spectrum enhancement1/2Norm, then the cost function is expressed as
J(n)=e2(n)+κ||wF(n)||1/2
Wherein k is constant, gradient descent method is adopted, and cost function is used for weighting vectorGradient finding
thus according to the gradient descent principle, by1/2The norm regularized sparse drive ALE weight update formula is as follows:
where μ is the step size and ρ ═ μ κ is a constant.
As shown in fig. 2, the steps specifically include:
step 2-1) initial frequency domain weight w of adaptive filterF(0) Expressed as:
wF(0)=[wF,0(0),wF,1(0),...,wF,L-1(0)]T
let n represent the number of iterations, with an initial value of 0;
step 2-2), the time domain output y (n) of the nth iteration adaptive filter is expressed as:
wherein ,wF(n)=[wF,0(n),wF,1(n),...,wF,L-1(n)]TThe frequency domain weight of the nth iteration adaptive filter is obtained; x (n- Δ) is the time delayed beam signal:
x(n-Δ)=[x(n-Δ),x(n-Δ-1),...,x(n-Δ-L+1)]T
wherein, x (n) is a reference signal, Δ is a time delay, and L is a filter order;
step 2-3) calculating the output estimation error e (n) of the nth iteration adaptive filter:
e(n)=x(n)-y(n)
step 2-4) calculating the frequency domain weight w of the (n +1) th self-adaptive filterF(n+1):
According to the principle of gradient descent, by1/2Norm regularized sparsely driven ALE processing, frequency domain weight w of the (n +1) th iterative adaptive filterF(n +1) is:
wherein ,representing bitwise multiplication of vectors, | · non-woven phosphor-3/2A power of-3/2, μ is the step size, ρ ═ μ κ is a constant, and κ is a constant, representing bitwise modulo of the vector; x is the number ofF(n- Δ) is a discrete Fourier transform of x (n- Δ);
step 2-5) if | | | wF(n+1)-wF(n) if the absolute value is less than the threshold value, entering the step 2-6), otherwise, after adding 1 to n, entering the step 2-2);
step 2-6), calculating a final enhanced beam domain signal y (n +1) after sparse drive ALE processing:
wherein x (n +1- Δ) ═ x (n +1- Δ), x (n- Δ),.., x (n- Δ -L +2)]T,xF(n +1- Δ) is the discrete Fourier transform of x (n +1- Δ).
After the wave beams are formed, sparse drive ALE processing is carried out on each wave beam signal, so that the power distribution of the target wave beam acoustic spectrum along the frequency axis direction has more obvious aggregation characteristic, and the power distribution of the non-target wave beam acoustic spectrum along the frequency axis direction is more balanced.
Step 3) the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, and entropy characteristic curves of the beam sound spectrums, namely curves of Shannon entropy of the beam sound spectrums changing along with time, are calculated;
the degree of aggregation of the power distributions can be represented by entropy, which is referred to as shannon entropy, and thus the entropy of the beam output is calculated. Suppose that the sound spectrum of each time window in the beam domain signal contains N frequency points with the amplitude values of c1,c2,…cNWeight p of nth frequency pointnDefined as the ratio of its amplitude to the sum of the amplitudes of all the frequency points in the time window:
expressing Encopy according to the magnitude weight of the m-th time window by the Shannon Entropy valuemComprises the following steps:
step 4), extracting entropy characteristics of each beam acoustic spectrum, and forming a two-dimensional characteristic vector by the mean value and the standard deviation of the entropy characteristic curve;
the difference of the sound spectrum entropy characteristics of the wired spectrum target beam and the wireless spectrum target beam is obvious. Because the entropy value in the entropy characteristic curve of the line spectrum target beam sound spectrum is low and the variation range is large, the mean value E of the entropy curve is adopted in the inventionaveAnd standard deviation EstdAnd (3) forming a two-dimensional feature vector:
step 5) inputting entropy characteristics to a support vector machine for training by utilizing the small sample learning capacity of the support vector machine, selecting a Gaussian kernel function, and optimizing parameters of the support vector machine according to the characteristic distribution of training data; the method specifically comprises the following steps:
establishing a training set, comprising: receiving data and a tag thereof by an array of the underwater unmanned platform;
carrying out beam forming on array receiving data of the underwater unmanned platform of the training set; forming a beam signal; carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals; the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve to form a two-dimensional entropy characteristic;
selecting a Gaussian kernel function as a binary classifier of a support vector machine; inputting the two-dimensional entropy characteristics of the training set into a support vector machine for training, and optimizing the parameters of the support vector machine according to the characteristic distribution of the training data.
And 6) inputting the entropy characteristics of the step 4) into a trained support vector machine, and outputting a judgment result serving as target detection.
When the output of the support vector machine is 0, the representative hypothesis test H0: no target is present; when the output is 1, the representative hypothesis test H1: there is a target.
In the invention, a support vector machine is adopted to judge whether a line spectrum target exists. Based on (E) extracted as aboveave,Estd) And (4) selecting a Gaussian kernel to carry out support vector machine classifier training, and seeking a superior boundary through parameter optimization to detect the weak target containing the line spectrum.
Example 2
An embodiment 2 of the present invention provides an underwater target detection system of a support vector machine based on sparse drive ALE, the system including: the system comprises an underwater unmanned platform, a trained support vector machine for target detection, a beam forming module, a sparse drive ALE processing module, a two-dimensional entropy characteristic construction module and a target detection module;
the beam forming module is used for carrying out beam forming on the data received by the array of the underwater unmanned platform; forming a beam signal;
the sparse drive ALE processing module is used for carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals;
the two-dimensional entropy characteristic construction module is used for converting the enhanced beam domain signals into beam sound spectrums through discrete Fourier transform, calculating entropy characteristics of the beam sound spectrums and drawing entropy characteristic curves; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics;
and the target detection module is used for inputting the two-dimensional entropy characteristics into a pre-trained support vector machine and outputting a judgment result of target detection.
Following sparse drive ALE (l)1/2ALE) in comparison with conventional ALE (C-ALE). The simulation signal is a superimposed signal of four sinusoidal signals of 76Hz, 88Hz, 138Hz and 194Hz, the input signal-to-noise ratio is-20 dB, the order of the filter is 2000, and step length parameters of the two algorithms are both mu-10-5Parameter ρ 9 × 10 of sparse drive ALE-10. Fig. 3 is a comparison of the steady state error of the conventional ALE and the sparse drive ALE, and it can be seen that the steady state error of the sparse drive ALE is significantly lower than the conventional ALE. Fig. 4 shows the amplitude-frequency characteristics of the conventional ALE weighting vector and the Optimal wiener weight (Optimal) after reaching the steady state, and fig. 5 shows the amplitude-frequency characteristics of the sparse driving ALE weighting vector and the Optimal wiener weight (Optimal) after reaching the steady state. Fig. 6 is a graph of output LOFAR of a conventional ALE, and fig. 7 is a graph of output LOFAR of a sparse drive ALE, comparing to see that the output signal-to-noise ratio of the sparse drive ALE is higher than that of the conventional ALE.
The detection performance of the invention is verified by using the offshore test data of the underwater unmanned platform. The depth of a test cooperative target sound source is fixed to be 10m, a broadband noise signal with the emission frequency of 1-4 kHz and a line spectrum is emitted, the average spectrum level is 115dB, the training data duration is 40s, and the test data duration is 120 s. The beam scanning interval is 1 degree, the scanning range is [ -139 degrees, 150 degrees ], the sparse drive ALE line spectrum enhancement is carried out on the beam output, the entropy characteristics are extracted from the enhanced beam acoustic spectrum, the training set comprises 5800 samples, and the testing set comprises 17400 samples. The training result of the support vector machine and the entropy characteristic distribution of the training set and the test set are shown in fig. 8, the comparison condition of the detection performance of the cooperative target of the invention and the automatic threshold method is listed in table 1, the automatic threshold method is influenced by strong interference of a broadband in test data, the detection probability of the cooperative target in a long distance is lower, and the invention effectively overcomes the missing detection phenomenon of the cooperative target, has higher detection probability and is less influenced by environmental interference.
TABLE 1
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (8)
1. A method for underwater target detection based on a support vector machine for sparse drive ALE, the method comprising:
carrying out beam forming on the array receiving data of the underwater unmanned platform; forming a beam signal;
carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals;
the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics;
and inputting the two-dimensional entropy characteristics into a pre-trained support vector machine, and outputting a judgment result of target detection.
2. The method for detecting the underwater target of the support vector machine based on the sparse drive ALE as claimed in claim 1, wherein the sparse drive ALE processing is performed on the beam signals to obtain enhanced beam domain signals, specifically comprising:
step S1) initial frequency domain weight w of adaptive filterF(0) Expressed as:
wF(0)=[wF,0(0),wF,1(0),...,wF,L-1(0)]T
let n represent the number of iterations, with an initial value of 0;
step S2), the time domain output y (n) of the nth iteration adaptive filter is expressed as:
wherein ,wF(n)=[wF,0(n),wF,1(n),...,wF,L-1(n)]TThe frequency domain weight of the nth iteration adaptive filter is obtained; x (n- Δ) is the time delayed beam signal:
x(n-Δ)=[x(n-Δ),x(n-Δ-1),...,x(n-Δ-L+1)]T
wherein, x (n) is a reference signal, Δ is a time delay, and L is a filter order;
step S3) calculates an output estimation error e (n) of the nth iterative adaptive filter:
e(n)=x(n)-y(n)
step S4) calculating the frequency domain weight w of the (n +1) th adaptive filterF(n+1):
According to the principle of gradient descent, by1/2Norm regularized sparsely driven ALE processing, frequency domain weight w of the (n +1) th iterative adaptive filterF(n +1) is:
||wF(n)||1/2=(|wF,0(n)|1/2+|wF,1(n)|1/2+...+|wF,L-1(n)|1/2)2
wherein ,representing bitwise multiplication of vectors, | · non-woven phosphor-3/2A power of-3/2, μ is the step size, ρ ═ μ κ is a constant, and κ is a constant, representing bitwise modulo of the vector; x is the number ofF(n- Δ) is a discrete Fourier transform of x (n- Δ);
step S5) if | | wF(n+1)-wF(n) | | is less than the threshold, then step S6 is entered, otherwise, after n adds 1, step S2 is entered;
step S6), calculating a final enhanced beam domain signal y (n +1) after the sparse drive ALE processing:
wherein x (n +1- Δ) ═ x (n +1- Δ), x (n- Δ),.., x (n- Δ -L +2)]T,xF(n +1- Δ) is the discrete Fourier transform of x (n +1- Δ).
3. The method for detecting the underwater target based on the sparse drive ALE support vector machine according to claim 2, wherein the step of converting the enhanced beam domain signal into a beam sound spectrum through discrete Fourier transform, calculating the entropy characteristic of the beam sound spectrum and drawing an entropy characteristic curve specifically comprises the steps of:
converting the enhanced beam domain signal y (n +1) into a beam acoustic spectrum through discrete Fourier transform;
setting the wave beam sound spectrum of the mth time window to contain N frequency points with the amplitudes of c1,c2,…cNThe weight p of the nth frequency pointnThe ratio of the amplitude to the sum of the amplitudes of all the frequency points in the time window is as follows:
according to the amplitude weight, the Shannon Entropy value Encopy of the mth time windowmComprises the following steps:
Entropymis the entropy property of the beam sound spectrum;
control with time window m as abscissamAn entropy characteristic curve is plotted for the ordinate.
4. The method for detecting the underwater target of the support vector machine based on the sparse drive ALE according to claim 3, wherein the calculating of the mean and the standard deviation of the entropy characteristic curve and the constructing of the two-dimensional entropy feature specifically comprises:
calculating the mean E of the entropy curvesaveAnd standard deviation Estd:
Wherein M is the number of entropy values in the entropy characteristic curve;
average value E of entropy curveaveAnd standard deviation EstdAnd forming a two-dimensional feature vector as a two-dimensional entropy feature.
5. The method of claim 4, further comprising: the step of training the support vector machine specifically comprises:
establishing a training set, comprising: receiving data and a tag thereof by an array of the underwater unmanned platform;
carrying out beam forming on array receiving data of the underwater unmanned platform of the training set; forming a beam signal; carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals; the enhanced beam domain signals are converted into beam sound spectrums through discrete Fourier transform, the entropy characteristics of the beam sound spectrums are calculated, and entropy characteristic curves are drawn; calculating the mean value and the standard deviation of the entropy characteristic curve to form a two-dimensional entropy characteristic;
selecting a Gaussian kernel function as a binary classifier of a support vector machine; when its output is 0, it represents hypothesis test H0: no target is present; when its output is 1, it represents hypothesis test H1: a presence target;
inputting the two-dimensional entropy characteristics of the training set into a support vector machine for training, and optimizing the parameters of the support vector machine according to the characteristic distribution of the training data.
6. An underwater target detection system based on a sparsely driven ALE support vector machine, the system comprising: the system comprises an underwater unmanned platform, a trained support vector machine for target detection, a beam forming module, a sparse drive ALE processing module, a two-dimensional entropy characteristic construction module and a target detection module;
the beam forming module is used for carrying out beam forming on the data received by the array of the underwater unmanned platform; forming a beam signal;
the sparse drive ALE processing module is used for carrying out sparse drive ALE processing on the beam signals to obtain enhanced beam domain signals;
the two-dimensional entropy characteristic construction module is used for converting the enhanced beam domain signals into beam sound spectrums through discrete Fourier transform, calculating entropy characteristics of the beam sound spectrums and drawing entropy characteristic curves; calculating the mean value and the standard deviation of the entropy characteristic curve, and constructing two-dimensional entropy characteristics;
and the target detection module is used for inputting the two-dimensional entropy characteristics into a pre-trained support vector machine and outputting a judgment result of target detection.
7. The system for detecting the underwater target of the sparse-drive ALE-based support vector machine according to claim 6, wherein the sparse-drive ALE processing module is implemented by the following steps:
step S1) initial frequency domain weight w of adaptive filterF(0) Expressed as:
wF(0)=[wF,0(0),wF,1(0),...,wF,L-1(0)]T
let n represent the number of iterations, with an initial value of 0;
step S2), the time domain output y (n) of the nth iteration adaptive filter is expressed as:
wherein ,wF(n)=[wF,0(n),wF,1(n),...,wF,L-1(n)]TThe frequency domain weight of the nth iteration adaptive filter is obtained; x (n- Δ) is the time delayed beam signal:
x(n-Δ)=[x(n-Δ),x(n-Δ-1),...,x(n-Δ-L+1)]T
wherein, x (n) is a reference signal, Δ is a time delay, and L is a filter order;
step S3) calculates an output estimation error e (n) of the nth iterative adaptive filter:
e(n)=x(n)-y(n)
step S4) calculating the frequency domain weight w of the (n +1) th adaptive filterF(n+1):
According to the principle of gradient descent, by1/2Norm regularized sparsely driven ALE processing, frequency domain weight w of the (n +1) th iterative adaptive filterF(n +1) is:
||wF(n)||1/2=(|wF,0(n)|1/2+|wF,1(n)|1/2+...+|wF,L-1(n)|1/2)2
wherein ,representing bitwise multiplication of vectors, | · non-woven phosphor-3/2A power of-3/2, μ is the step size, ρ ═ μ κ is a constant, and κ is a constant, representing bitwise modulo of the vector; x is the number ofF(n-Delta) is a discrete Fourier of x (n-Delta)Transforming;
step S5) if | | wF(n+1)-wF(n) | | is less than the threshold, then step S6 is entered, otherwise, after n adds 1, step S2 is entered;
step S6), calculating a final enhanced beam domain signal y (n +1) after the sparse drive ALE processing:
wherein x (n +1- Δ) ═ x (n +1- Δ), x (n- Δ),.., x (n- Δ -L +2)]T,xF(n +1- Δ) is the discrete Fourier transform of x (n +1- Δ).
8. The system for detecting the underwater target based on the sparse drive ALE support vector machine according to claim 7, wherein the two-dimensional entropy feature construction module is implemented by the following specific processes:
converting the enhanced beam domain signal y (n +1) into a beam acoustic spectrum through discrete Fourier transform;
setting the wave beam sound spectrum of the mth time window to contain N frequency points with the amplitudes of c1,c2,…cNThe weight p of the nth frequency pointnThe ratio of the amplitude to the sum of the amplitudes of all the frequency points in the time window is as follows:
according to the amplitude weight, the Shannon Entropy value Encopy of the mth time windowmComprises the following steps:
Entropymis the entropy property of the beam sound spectrum;
control with time window m as abscissamDrawing an entropy characteristic curve for a vertical coordinate;
calculating the mean E of the entropy curvesaveAnd standard deviation Estd:
Wherein M is the number of entropy values in the entropy characteristic curve;
average value E of entropy curveaveAnd standard deviation EstdAnd forming a two-dimensional feature vector as a two-dimensional entropy feature.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010242406.6A CN111582026B (en) | 2020-03-31 | 2020-03-31 | Underwater target detection method and system based on sparse drive ALE (ALE) support vector machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010242406.6A CN111582026B (en) | 2020-03-31 | 2020-03-31 | Underwater target detection method and system based on sparse drive ALE (ALE) support vector machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111582026A true CN111582026A (en) | 2020-08-25 |
CN111582026B CN111582026B (en) | 2023-05-23 |
Family
ID=72111458
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010242406.6A Active CN111582026B (en) | 2020-03-31 | 2020-03-31 | Underwater target detection method and system based on sparse drive ALE (ALE) support vector machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111582026B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112462352A (en) * | 2020-10-30 | 2021-03-09 | 哈尔滨工程大学 | Line spectrum enhancement method suitable for low signal-to-noise ratio condition |
CN113343914A (en) * | 2021-06-30 | 2021-09-03 | 东南大学 | Radiation noise line spectrum frequency domain self-adaptive enhancement method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101846594A (en) * | 2010-06-22 | 2010-09-29 | 上海交通大学 | Fault detection device based on beam forming acoustic-image mode recognition and detection method thereof |
US20170004403A1 (en) * | 2015-05-21 | 2017-01-05 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems and methods for predicting personal attributes based on public interaction data |
CN106355151A (en) * | 2016-08-30 | 2017-01-25 | 电子科技大学 | Recognition method, based on deep belief network, of three-dimensional SAR images |
CN107766818A (en) * | 2017-10-18 | 2018-03-06 | 哈尔滨工程大学 | A kind of didactic submerged structure environment line feature extraction method |
CN109977724A (en) * | 2017-12-27 | 2019-07-05 | 中国科学院声学研究所 | A kind of Underwater Target Classification method |
-
2020
- 2020-03-31 CN CN202010242406.6A patent/CN111582026B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101846594A (en) * | 2010-06-22 | 2010-09-29 | 上海交通大学 | Fault detection device based on beam forming acoustic-image mode recognition and detection method thereof |
US20170004403A1 (en) * | 2015-05-21 | 2017-01-05 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems and methods for predicting personal attributes based on public interaction data |
CN106355151A (en) * | 2016-08-30 | 2017-01-25 | 电子科技大学 | Recognition method, based on deep belief network, of three-dimensional SAR images |
CN107766818A (en) * | 2017-10-18 | 2018-03-06 | 哈尔滨工程大学 | A kind of didactic submerged structure environment line feature extraction method |
CN109977724A (en) * | 2017-12-27 | 2019-07-05 | 中国科学院声学研究所 | A kind of Underwater Target Classification method |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112462352A (en) * | 2020-10-30 | 2021-03-09 | 哈尔滨工程大学 | Line spectrum enhancement method suitable for low signal-to-noise ratio condition |
CN113343914A (en) * | 2021-06-30 | 2021-09-03 | 东南大学 | Radiation noise line spectrum frequency domain self-adaptive enhancement method |
Also Published As
Publication number | Publication date |
---|---|
CN111582026B (en) | 2023-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9264799B2 (en) | Method and apparatus for acoustic area monitoring by exploiting ultra large scale arrays of microphones | |
CN109444820B (en) | Method for detecting target after interference suppression of multi-channel radar when clutter and interference coexist | |
CN111965632B (en) | Radar target detection method based on Riemann manifold dimensionality reduction | |
CN111582026B (en) | Underwater target detection method and system based on sparse drive ALE (ALE) support vector machine | |
CN114595732B (en) | Radar radiation source sorting method based on depth clustering | |
CN110888105A (en) | DOA estimation method based on convolutional neural network and received signal strength | |
CN107479050B (en) | Target detection method and device based on symmetric spectral characteristics and sub-symmetric characteristics | |
CN111650574B (en) | Underwater space-time self-adaptive processing method and system based on sparse recovery | |
CN107783096A (en) | A kind of two-dimensional background equalization methods shown for bearing history figure | |
CN111446997B (en) | Self-adaptive digital beam synthesis method based on deep learning | |
CN110361709B (en) | Vehicle-mounted millimeter wave radar target identification method based on dynamic false alarm probability | |
CN114089307B (en) | Radar detection and classification method and system under target and interference conditions | |
CN116400335A (en) | Horizontal array active sonar target echo detection method and system | |
CN111931570B (en) | Through-wall imaging radar human body target detection method based on full convolution network | |
CN111414580B (en) | Reverberation suppression method under low signal-to-mixing ratio condition | |
Xiangwei et al. | Adaptive clutter map detector in nonhomogeneous environment | |
Melkanovich | Features of implementation of adaptive signals processing for a cylindrical antenna array with a horizontal generatrix | |
CN116679278B (en) | Target radar detection method under strong ground clutter interference | |
CN115144830B (en) | Small target radar detection method under strong ground clutter interference | |
CN115062658B (en) | Overlapping radar signal modulation type identification method based on self-adaptive threshold network | |
CN114299356A (en) | Passive underwater acoustic target identification method based on multi-scale convolutional neural network | |
Hamid et al. | Improvements in broadband processing of passive arrays for target detection in low SNR conditions | |
CN116755097A (en) | Multichannel frequency domain line spectrum enhancement method and system for simulated ray craft | |
CN117970299A (en) | Target echo time delay estimation method and system under strong direct background interference | |
Remadevi et al. | Adaptive Tow Ship Noise Cancellation Using Deep Regression Neural Network |
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 |