CN107656250A - A kind of Intelligent radar sea target detection system and method based on artificial bee colony algorithm - Google Patents

Abstract

The invention discloses an intelligent radar sea target detection system and method based on an artificial bee colony algorithm. The system is composed of a radar, a database, and a host computer connected in sequence. The radar irradiates the detected sea area and stores the radar sea clutter data in the The database, the host computer includes a data preprocessing module, a robust forecast model modeling module, an intelligent optimization module, a target detection module, a model update module and a result display module: the present invention aims at the complex characteristics of maritime target detection , reconstruct the radar clutter data, detect the target data, and introduce the artificial bee colony algorithm, so as to provide a radar sea target detection system and method with online detection and high intelligence.

Description

An intelligent radar maritime target detection system and method based on artificial bee colony algorithm

technical field

The invention relates to the field of radar data processing, in particular to an intelligent radar sea target detection system and method based on an artificial bee colony algorithm.

Background technique

Sea clutter is radar backscatter echoes from the sea surface. In recent decades, with the in-depth understanding of sea clutter, Germany, Norway and other countries have successively tried to use radar to observe sea clutter to obtain radar wave images to invert wave information, so as to obtain real-time information about the state of the ocean, such as the Wave height, direction and period, etc., so as to further detect small targets at sea, which is of great significance to maritime activities.

Maritime target detection technology plays an important role, and providing accurate target judgment is one of the important tasks of marine radar. The radar automatic detection system makes a judgment under a given detection threshold according to the judgment criterion, and the strong sea clutter often becomes the main interference of the weak target signal. How to deal with sea clutter will directly affect the detection ability of radar in the marine environment: 1) Identify navigation buoys, small pieces of ice, and oil pollution floating on the sea surface, which may bring potential crises to navigation; 3) Monitor illegal fishing Fish is an important task for environmental monitoring.

In traditional target detection, sea clutter is considered to be a noise that interferes with navigation and is removed. However, when the radar observes the target on the sea, the weak moving target echo is often buried in the sea clutter, and the signal clutter ratio is low, so it is difficult for the radar to detect the target. The detection performance of the radar has a great influence. For various maritime warning and early warning radars, the main goal of the research is to improve the detection ability of targets in the background of sea clutter. Therefore, it not only has important theoretical and practical significance, but also is a difficult point and a hot spot in the detection of maritime targets at home and abroad.

Contents of the invention

In order to overcome the shortcomings of existing radar sea target detection methods that cannot realize online detection and poor intelligence, the present invention provides an intelligent radar sea target detection system and method based on artificial bee colony algorithm that realizes online detection and has strong intelligence.

The technical solution adopted by the present invention to solve its technical problems is:

An intelligent radar sea target detection system based on the artificial bee colony algorithm, including a radar, a database and a host computer, the radar, the database and the host computer are connected in sequence, the radar illuminates the detected sea area, and stores the radar sea clutter data To the database, the host computer includes:

The data preprocessing module is used to preprocess the radar sea clutter data, which is completed by the following process:

(1) The radar irradiates the detected sea area, and stores the radar sea clutter data into the database;

(2) Collect N radar sea clutter echo signal amplitudes x _i from the database as training samples, i=1,...,N;

(3) Normalize the training samples to obtain the normalized amplitude

Among them, minx represents the minimum value in the training sample, and maxx represents the maximum value in the training sample;

(4) Reconstruct the normalized training samples to obtain the input matrix X and the corresponding output matrix Y respectively:

Among them, D represents the reconstruction dimension, D is a natural number, and D<N, and the value range of D is 50-70;

The robust forecast model modeling module is used to establish a forecast model, which is completed by the following process:

Substitute the X and Y obtained by the data preprocessing module into the following linear equation:

in

The weight factor v _i is calculated by the following formula:

in is the estimate of the standard deviation of the error variable ξ _i , c ₁ and c ₂ are constants;

Solve the estimated function f(x):

Among them, M is the number of support vectors, 1 _v ＝[1,...,1] ^T , The superscript T denotes the transpose of the matrix, is the Lagrangian multiplier, b ^* is the bias, K=exp(-||x _i -x _j ||/θ ² ), where i=1,...,M, j=1,...,M , and exp(-||xx _i ||/θ ² ) are the kernel functions of the support vector machine, x _j is the amplitude of the jth radar sea clutter echo signal, θ is the kernel parameter, x is the input variable, γ is the penalty coefficient;

The intelligent optimization module uses the artificial bee colony algorithm to optimize the kernel parameters θ and penalty coefficient γ of the robust forecasting model, and completes it through the following process:

Step 1: Initialize the parameters of the artificial bee colony algorithm, set the number of honey sources P, the maximum iteration number iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the honey source represents the feasible solution of the problem, since the model has two Each parameter needs to be optimized, so the dimension of the position p _i is 2 dimensions, the position p _i of the nectar source is randomly generated according to the following formula = (p _i1 , p _i2 ), and the initial iteration number iter = 0;

p _ij =L _d +rand()*(U _d -L _d )(i=1,2,...,P,j=1,2)

Step 2: assign a leading bee to the nectar source p _i , search according to the formula, and generate a new nectar source V _i ;

Step 3: Calculate the fitness value of V _i , and determine the retained honey source according to the method of greedy selection;

Step 4: Calculate the probability that the nectar source that leads the bees to find is replaced;

Step 5: follow the bees to search in the same way as the lead bees, and determine the nectar source to keep according to the method of greedy selection;

Step 6: Determine whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding role of the leading bee becomes a scout bee, otherwise go directly to step 8;

Step 7: Scout bees randomly generate new honey sources;

Step 8: iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, otherwise go to step 2.

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

The target detection module is used for target detection, which is completed by the following process:

1) Collect D sea clutter echo signal amplitudes at sampling time t to obtain TX=[x _t-D+1 ,...,x _t ], where x _t-D+1 represents the Sea clutter echo signal amplitude, x _t represents the sea clutter echo signal amplitude at the tth sampling moment;

2) Carry out normalization processing;

3) Substituting the estimated function f(x) obtained by the modeling module of the robust prediction model to calculate the sea clutter forecast value at the sampling time (t+1).

4) Calculate the difference e between the predicted value of sea clutter and the measured value of radar echo, and calculate the control limit Q _α :

Among them, α is the confidence level, θ ₁ , θ ₂ , θ ₃ , and h ₀ are intermediate variables, λ _j ⁱ represents the i-th power of the jth eigenvalue of the covariance matrix, k is the sample dimension, and C _α is the positive Statistical distribution reliability is α;

5) Detection and judgment: when the e ² difference is greater than the control limit Q _α , there is a target at this point, otherwise there is no target.

The model update module is used to collect data according to the set sampling time interval, compare the obtained measured data with the model forecast value, and if the relative error is greater than 10%, add new data to the training sample data and update the forecast model.

The result display module is used to display the detection result of the target detection module on the host computer.

A kind of radar sea target detection method used by the intelligent radar sea target detection system based on artificial bee colony algorithm, described method comprises the following steps:

(1) The radar irradiates the detected sea area, and stores the radar sea clutter data into the database;

(2) Collect N radar sea clutter echo signal amplitudes x _i from the database as training samples, i=1,...,N;

(3) Normalize the training samples to obtain the normalized amplitude

Among them, minx represents the minimum value in the training sample, and maxx represents the maximum value in the training sample;

(4) Reconstruct the normalized training samples to obtain the input matrix X and the corresponding output matrix Y respectively:

Among them, D represents the reconstruction dimension, D is a natural number, and D<N, and the value range of D is 50-70;

(5) Substitute the obtained X and Y into the following linear equation:

in

The weight factor v _i is calculated by the following formula:

in is the estimate of the standard deviation of the error variable ξ _i , c ₁ and c ₂ are constants;

Solve the estimated function f(x):

Among them, M is the number of support vectors, 1 _v ＝[1,...,1] ^T , The superscript T denotes the transpose of the matrix, is the Lagrangian multiplier, b ^* is the bias, K=exp(-||x _i -x _j ||/θ ² ), where i=1,...,M, j=1,...,M , and exp(-||xx _i ||/θ ² ) are the kernel functions of the support vector machine, x _j is the amplitude of the jth radar sea clutter echo signal, θ is the kernel parameter, x is the input variable, γ is the penalty coefficient;

(6) Use the artificial bee colony algorithm to optimize the kernel parameter θ and the penalty coefficient γ of step (5), and use the following process to complete:

(6.1) Initialize the parameters of the artificial bee colony algorithm, set the number of honey sources P, the maximum iteration number iter _max , the minimum and maximum values L _d and U _d of the initial search space; the position of the honey source represents the feasible solution of the problem, since the model has two Each parameter needs to be optimized, so the dimension of the position p _i is 2 dimensions, the position p _i of the nectar source is randomly generated according to the following formula = (p _i1 , p _i2 ), and the initial iteration number iter = 0;

p _ij =L _d +rand()*(U _d -L _d )(i=1,2,...,P,j=1,2)

(6.2) Assign a leading bee to the honey source p _i , search according to the formula, and generate a new honey source V _i ;

(6.3) Calculate the fitness value of _Vi , and determine the nectar source to keep according to the method of greedy selection;

(6.4) Calculate the probability that the nectar source that leads the bee to find is more followed;

(6.5) Follower bees search in the same way as lead bees, and determine the nectar source to keep according to the method of greedy selection;

(6.6) Determine whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding role of the leading bee becomes a scout bee, otherwise directly go to step (6.8);

(6.7) Scout bees randomly generate new honey sources;

(6.8) iter=iter+1, judging whether the maximum number of iterations has been reached, if satisfied, then output the optimal parameters, otherwise go to step (6.2).

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

(7) Collect D sea clutter echo signal amplitudes at sampling time t to obtain TX=[x _t-D+1 ,...,x _t ], where x _t-D+1 represents the t-D+1th sampling time The amplitude of the sea clutter echo signal, x _t represents the amplitude of the sea clutter echo signal at the tth sampling moment;

(8) Carry out normalization processing;

(9) Substituting the estimated function f(x) obtained in step (5) to calculate the predicted value of sea clutter at the sampling time (t+1).

(10) Calculate the difference e between the predicted value of sea clutter and the measured value of radar echo, and calculate the control limit Q _α :

Among them, α is the confidence level, θ ₁ , θ ₂ , θ ₃ , and h ₀ are intermediate variables, λ _j ⁱ represents the i-th power of the jth eigenvalue of the covariance matrix, k is the sample dimension, and C _α is the positive Statistical distribution reliability is α;

(11) Detection and judgment: when the e ² difference is greater than the control limit Q _α , there is a target at this point, otherwise there is no target.

(12) Collect data according to the set sampling time interval, compare the obtained measured data with the model forecast value, if the relative error is greater than 10%, add new data to the training sample data, and update the forecast model.

The technical idea of the present invention is: the present invention aims at the chaotic characteristics of the radar sea clutter, reconstructs the radar sea clutter data, performs nonlinear fitting on the reconstructed data, and establishes a forecast model of the radar sea clutter, Calculate the difference between the predicted value and the measured value of the radar sea clutter. The error will be significantly greater when there is a target than when there is no target. The artificial bee colony algorithm is introduced to achieve strong intelligent target detection under the background of sea clutter.

The beneficial effects of the present invention are mainly manifested in: 1. Sea targets can be detected online; 2. The detection method used only needs fewer samples; 3. It has strong intelligence and is less affected by human factors.

Description of drawings

Fig. 1 is the hardware structural diagram of the system proposed by the present invention;

Fig. 2 is a functional block diagram of the host computer proposed by the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. The embodiments of the present invention are used to explain the present invention, rather than to limit the present invention. Within the spirit of the present invention and the protection scope of the claims, any modification and change made to the present invention will fall into the protection scope of the present invention.

Example 1

With reference to Fig. 1, Fig. 2, a kind of intelligent radar sea target detection system based on artificial bee colony algorithm, comprises radar 1, database 2, and host computer 3, and radar 1, database 2 and host computer 3 are connected successively, and described radar 1 The detected sea area is irradiated, and the radar sea clutter data is stored in the database 2, and the upper computer 3 includes:

The data preprocessing module 4 is used for radar sea clutter data preprocessing, which is completed by the following process:

1) Collect N radar sea clutter echo signal amplitudes x _i from the database as training samples, i=1,...,N;

2) Normalize the training samples to obtain the normalized amplitude

Among them, minx represents the minimum value in the training sample, and maxx represents the maximum value in the training sample;

3) Reconstruct the normalized training samples to obtain the input matrix X and the corresponding output matrix Y respectively:

Among them, D represents the reconstruction dimension, D is a natural number, and D<N, and the value range of D is 50-70;

Robust forecast model modeling module 5 is used to establish a forecast model, which is completed by the following process:

Substitute the obtained X and Y into the following linear equation:

in

The weight factor v _i is calculated by the following formula:

in is the estimate of the standard deviation of the error variable ξ _i , c ₁ , c ₂ are constants

Solve the estimated function f(x):

Among them, M is the number of support vectors, 1 _v ＝[1,...,1] ^T , The superscript T denotes the transpose of the matrix, is the Lagrangian multiplier, b ^* is the bias, K=exp(-||x _i -x _j ||/θ ² ), where i=1,...,M, j=1,...,M , and exp(-||xx _i ||/θ ² ) are the kernel functions of the support vector machine, x _j is the amplitude of the jth radar sea clutter echo signal, θ is the kernel parameter, x is the input variable, γ is the penalty coefficient;

The intelligent optimization module 6 is used to optimize the kernel parameter θ and the penalty coefficient γ of the robust prediction model by using the artificial bee colony algorithm, and the following process is used to complete:

Step 1: Initialize the parameters of the artificial bee colony algorithm, set the number of honey sources P, the maximum iteration number iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the honey source represents the feasible solution of the problem, since the model has two Each parameter needs to be optimized, so the dimension of the position p _i is 2 dimensions, the position p _i of the nectar source is randomly generated according to the following formula = (p _i1 , p _i2 ), and the initial iteration number iter = 0;

p _ij =L _d +rand()*(U _d -L _d )(i=1,2,...,P,j=1,2)

Step 2: assign a leading bee to the nectar source p _i , search according to the formula, and generate a new nectar source V _i ;

Step 3: Calculate the fitness value of V _i , and determine the retained honey source according to the method of greedy selection;

Step 4: Calculate the probability that the nectar source that leads the bees to find is replaced;

Step 5: follow the bees to search in the same way as the lead bees, and determine the nectar source to keep according to the method of greedy selection;

Step 6: Determine whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding role of the leading bee becomes a scout bee, otherwise go directly to step 8;

Step 7: Scout bees randomly generate new honey sources;

Step 8: iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, otherwise go to step 2.

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

The target detection module 7 is used for target detection, which is completed by the following process:

1) Collect D sea clutter echo signal amplitudes at sampling time t to obtain TX=[x _t-D+1 ,...,x _t ], where x _t-D+1 represents the Sea clutter echo signal amplitude, x _t represents the sea clutter echo signal amplitude at the tth sampling moment;

2) Carry out normalization processing;

3) Substituting the function f(x) obtained by the robust prediction model modeling module to obtain the sea clutter forecast value at the sampling time (t+1);

4) Calculate the difference e between the predicted value of sea clutter and the measured value of radar echo, and calculate the control limit Q _α :

Among them, α is the confidence level, θ ₁ , θ ₂ , θ ₃ , and h ₀ are intermediate variables, λ _j ⁱ represents the i-th power of the jth eigenvalue of the covariance matrix, k is the sample dimension, and C _α is the positive Statistical distribution reliability is α;

5) Detection and judgment: when the e ² difference is greater than the control limit Q _α , there is a target at this point, otherwise there is no target.

The model update module 8 is used to collect data according to the set sampling time interval, compare the obtained measured data with the model forecast value, and if the relative error is greater than 10%, add new data to the training sample data and update the forecast model.

The result display module 9 is used to display the detection result of the target detection module on the host computer.

The hardware part of described upper computer 3 comprises: I/O element, is used for the acquisition of data and the transmission of information; Data memory, stores the required data sample of operation and operating parameter etc.; Program memory, stores the software program that realizes function module ; Calculator, to execute the program, to realize the specified function; display module, to display the set parameters and test results.

Example 2

With reference to Fig. 1, Fig. 2, a kind of intelligent radar sea target detection method based on artificial bee colony algorithm, described method comprises the following steps:

(1) The radar irradiates the detected sea area, and stores the radar sea clutter data into the database;

(2) Collect N radar sea clutter echo signal amplitudes x _i from the database as training samples, i=1,...,N;

(3) Normalize the training samples to obtain the normalized amplitude

Among them, minx represents the minimum value in the training sample, and maxx represents the maximum value in the training sample;

(4) Reconstruct the normalized training samples to obtain the input matrix X and the corresponding output matrix Y respectively:

Among them, D represents the reconstruction dimension, D is a natural number, and D<N, and the value range of D is 50-70;

(5) Substitute the obtained X and Y into the following linear equation:

in

The weight factor v _i is calculated by the following formula:

in is the estimate of the standard deviation of the error variable ξ _i , c ₁ and c ₂ are constants;

Solve the estimated function f(x):

Among them, M is the number of support vectors, 1 _v ＝[1,...,1] ^T , The superscript T denotes the transpose of the matrix, is the Lagrangian multiplier, b ^* is the bias, K=exp(-||x _i -x _j ||/θ ² ), where i=1,...,M, j=1,...,M , and exp(-||xx _i ||/θ ² ) are the kernel functions of the support vector machine, x _j is the amplitude of the jth radar sea clutter echo signal, θ is the kernel parameter, x is the input variable, γ is the penalty coefficient;

(6) Use the artificial bee colony algorithm to optimize the kernel parameter θ and the penalty coefficient γ of step (5), and use the following process to complete:

(6.1) Initialize the parameters of the artificial bee colony algorithm, set the number of honey sources P, the maximum iteration number iter _max , the minimum and maximum values L _d and U _d of the initial search space; the position of the honey source represents the feasible solution of the problem, because the model has two Each parameter needs to be optimized, so the dimension of the position p _i is 2 dimensions, the position p _i of the nectar source is randomly generated according to the following formula = (p _i1 , p _i2 ), and the initial iteration number iter = 0;

p _ij =L _d +rand()*(U _d -L _d )(i=1,2,...,P,j=1,2)

(6.2) Assign a leading bee to the honey source p _i , search according to the formula, and generate a new honey source V _i ;

(6.3) Calculate the fitness value of _Vi , and determine the nectar source to keep according to the method of greedy selection;

(6.4) Calculate the probability that the nectar source that leads the bee to find is more followed;

(6.5) Follower bees search in the same way as lead bees, and determine the nectar source to keep according to the method of greedy selection;

(6.6) Determine whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding role of the leading bee becomes a scout bee, otherwise directly go to step (6.8);

(6.7) Scout bees randomly generate new honey sources;

(6.8) iter=iter+1, judging whether the maximum number of iterations has been reached, if satisfied, then output the optimal parameters, otherwise go to step (6.2).

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

(7) Collect D sea clutter echo signal amplitudes at sampling time t to obtain TX=[x _t-D+1 ,...,x _t ], where x _t-D+1 represents the t-D+1th sampling time The amplitude of the sea clutter echo signal, x _t represents the amplitude of the sea clutter echo signal at the tth sampling moment;

(8) Carry out normalization processing;

(9) Substituting the estimated function f(x) obtained in step (5) to calculate the predicted value of sea clutter at the sampling time (t+1).

(10) Calculate the difference e between the predicted value of sea clutter and the measured value of radar echo, and calculate the control limit Q _α :

Among them, α is the confidence level, θ ₁ , θ ₂ , θ ₃ , and h ₀ are intermediate variables, λ _j ⁱ represents the i-th power of the jth eigenvalue of the covariance matrix, k is the sample dimension, and C _α is the positive Statistical distribution reliability is α;

(11) Detection and judgment: when the e ² difference is greater than the control limit Q _α , there is a target at this point, otherwise there is no target.

(12) Collect data according to the set sampling time interval, compare the obtained measured data with the model forecast value, if the relative error is greater than 10%, add new data to the training sample data, and update the forecast model.

As can be seen from the above embodiments, the present invention has established an intelligent radar sea target detection system and method, which can detect radar targets online; and the detection method used only needs fewer samples; in addition, the influence of human factors is reduced, and the intelligence is high , strong robustness.

Claims (2)

1. a kind of Intelligent radar sea target detection system based on artificial bee colony algorithm, including radar, database and upper Machine, radar, database and host computer are sequentially connected, it is characterised in that：The radar is irradiated to detected marine site, and by thunder Up to sea clutter data storage to described database, described host computer includes data preprocessing module, robust forecasting model is built Mould module, intelligent optimizing module, module of target detection, model modification module and result display module：

The data preprocessing module, to carry out radar sea clutter data prediction, completed using following process：

(1) radar is irradiated to detected marine site, and by radar sea clutter data storage to described database；

(2) N number of radar sea clutter echo-signal amplitude x is gathered from database_iAs training sample, i=1 ..., N；

(3) training sample is normalized, obtains normalizing amplitude

<mrow> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mi>i</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> <mrow> <mi>max</mi> <mi> </mi> <mi>x</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> </mfrac> </mrow>

Wherein, minx represents the minimum value in training sample, and maxx represents the maximum in training sample；

(4) training sample after normalization is reconstructed, respectively obtains input matrix X and corresponding output matrix Y：

<mrow> <mi>Y</mi> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>D</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>D</mi> <mo>+</mo> <mn>2</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <mtable> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> </mtable> </mtd> </mtr> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mi>N</mi> </msub> </mtd> </mtr> </mtable> </mfenced> </mrow>

Wherein, D represents reconstruct dimension, and D is natural number, and D ＜ N, D span are 50-70；

The robust forecasting model modeling module is completed to establish forecasting model using following process：

X, Y that data preprocessing module is obtained substitute into following linear equation：

<mrow> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <msubsup> <mn>1</mn> <mi>v</mi> <mi>T</mi> </msubsup> </mtd> </mtr> <mtr> <mtd> <msub> <mn>1</mn> <mi>v</mi> </msub> </mtd> <mtd> <mrow> <mi>K</mi> <mo>+</mo> <msub> <mi>V</mi> <mi>&amp;gamma;</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msup> <mi>b</mi> <mo>*</mo> </msup> </mtd> </mtr> <mtr> <mtd> <msup> <mi>&amp;alpha;</mi> <mo>*</mo> </msup> </mtd> </mtr> </mtable> </mfenced> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mi>Y</mi> </mtd> </mtr> </mtable> </mfenced> </mrow>

Wherein

Weight factor v_iCalculated by following formula：

WhereinIt is error variance ξ_iThe estimation of standard deviation, c₁,c₂For constant；

Solve to obtain function f (x) to be estimated：

<mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <msup> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>*</mo> </msup> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mo>|</mo> <mo>|</mo> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>|</mo> <mo>|</mo> <mo>/</mo> <msup> <mi>&amp;theta;</mi> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mo>+</mo> <msup> <mi>b</mi> <mo>*</mo> </msup> </mrow>

Wherein, M is the number of supporting vector, 1_v=[1 ..., 1]^T,The transposition of subscript T representing matrixs,It is Lagrange multiplier, b^*It is amount of bias, K=exp (- | | x_i-x_j||/θ²), wherein i=1 ..., M, j=1 ..., M,With exp (- | | x-x_i||/θ²) be SVMs kernel function, x_jFor j-th of radar sea clutter Echo-signal amplitude, θ are nuclear parameters, and x represents input variable, and γ is penalty coefficient；

The intelligent optimizing module, to the nuclear parameter θ and penalty coefficient γ using artificial bee colony algorithm to robust forecasting model Optimize, completed using following process：

(A) parameter of artificial bee colony algorithm is initialized, if nectar source number P, greatest iteration number iter_max, the minimum in initial ranging space Value and maximum L_dAnd U_d；The feasible solution of the positional representation problem in nectar source, because model has two parameters to need to optimize, so position Put p_iDimension for 2 dimension, as the following formula at random generation nectar source position p_i=(p_i1,p_i2), put primary iteration number iter=0；

p_ij=L_d+rand()*(U_d-L_d) (i=1,2 ..., P, j=1,2)

(B) it is nectar source p_iDistribution one leads honeybee, scans for as the following formula, produces new nectar source V_i；

(C) V is calculated_iFitness value, according to greediness selection method determine retain nectar source；

(D) calculate lead the nectar source that honeybee is found by more with probability；

(E) honeybee is followed to use with leading honeybee identical mode to scan for, the nectar source for determining to retain according to the method for greediness selection；

(F) nectar source V is judged_iWhether the condition that satisfaction is abandoned, such as meet, it is corresponding to lead honeybee role to be changed into search bee, otherwise directly It is switched to (H)

(G) search bee randomly generates new nectar source；

(H) iter=iter+1, judge whether to have been maxed out iterations, export optimized parameter if meeting, otherwise turn To step (B).

Wherein, nectar source number is 100, the minimum value and maximum 0 and 100 in initial ranging space, maximum iteration 100.

The module of target detection, to carry out target detection, completed using following process：

(a) gather D sea clutter echo-signal amplitude in sampling instant t and obtain TX=[x_t-D+1,...,x_t], x_t-D+1Represent t- The sea clutter echo-signal amplitude of D+1 sampling instants, x_tThe sea clutter echo-signal amplitude of t sampling instants is represented, TX is represented Signal amplitude matrix of the sea clutter from t-D+1 sampling instants to t sampling instants；

(b) it is normalized；

<mrow> <mover> <mrow> <mi>T</mi> <mi>X</mi> </mrow> <mo>&amp;OverBar;</mo> </mover> <mo>=</mo> <mfrac> <mrow> <mi>T</mi> <mi>X</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> <mrow> <mi>max</mi> <mi> </mi> <mi>x</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> </mfrac> </mrow>

(c) sea that sampling instant (t+1) is calculated in the function f (x) to be estimated that robust forecasting model modeling module obtains is substituted into Clutter predicted value.

(d) difference e of sea clutter predicted value and radar return measured value is calculated, calculates control limit Q_α：

<mrow> <msub> <mi>Q</mi> <mi>&amp;alpha;</mi> </msub> <mo>=</mo> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> <msup> <mrow> <mo>&amp;lsqb;</mo> <mfrac> <mrow> <msub> <mi>C</mi> <mi>&amp;alpha;</mi> </msub> <msub> <mi>h</mi> <mn>0</mn> </msub> <msqrt> <mrow> <mn>2</mn> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> </mrow> </msqrt> </mrow> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> </mfrac> <mo>+</mo> <mn>1</mn> <mo>+</mo> <mfrac> <mrow> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> <msub> <mi>h</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <msub> <mi>h</mi> <mn>0</mn> </msub> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> </mfrac> <mo>&amp;rsqb;</mo> </mrow> <mfrac> <mn>1</mn> <msub> <mi>h</mi> <mn>0</mn> </msub> </mfrac> </msup> </mrow>

<mrow> <msub> <mi>&amp;theta;</mi> <mi>i</mi> </msub> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msubsup> <mi>&amp;lambda;</mi> <mi>j</mi> <mi>i</mi> </msubsup> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>3</mn> </mrow>

<mrow> <msub> <mi>h</mi> <mn>0</mn> </msub> <mo>=</mo> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mn>2</mn> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> <msub> <mi>&amp;theta;</mi> <mn>3</mn> </msub> </mrow> <mrow> <mn>3</mn> <msubsup> <mi>&amp;theta;</mi> <mn>2</mn> <mn>2</mn> </msubsup> </mrow> </mfrac> </mrow>

Wherein, α is confidence level, θ₁,θ₂,θ₃,h₀It is intermediate variable, λ_j ⁱThe i powers of j-th of characteristic value of covariance matrix are represented, K is sample dimension, C_αIt is the statistics that normal distribution confidence level is α；

(e) detection judgement is carried out：Work as e²Difference is more than control limit Q_αWhen, there is target in the point, otherwise without target.

The model modification module, to by the sampling time interval gathered data of setting, by obtained measured data and model Predicted value compares, if relative error is more than 10%, new data is added into training sample data, updates forecasting model.

The result display module, the testing result of module of target detection to be shown in host computer.

2. thunder used in the Intelligent radar sea target detection system based on artificial bee colony algorithm described in a kind of claim 1 Up to method for detecting targets at sea, it is characterised in that：Described method comprises the following steps：

(1) radar is irradiated to detected marine site, and by radar sea clutter data storage to described database；

(2) N number of radar sea clutter echo-signal amplitude x is gathered from database_iAs training sample, i=1 ..., N；

(3) training sample is normalized, obtains normalizing amplitude

<mrow> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mi>i</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> <mrow> <mi>max</mi> <mi> </mi> <mi>x</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> </mfrac> </mrow>

Wherein, minx represents the minimum value in training sample, and maxx represents the maximum in training sample；

(4) training sample after normalization is reconstructed, respectively obtains input matrix X and corresponding output matrix Y：

<mrow> <mi>Y</mi> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>D</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mrow> <mi>D</mi> <mo>+</mo> <mn>2</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <mtable> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> </mtable> </mtd> </mtr> <mtr> <mtd> <msub> <mover> <mi>x</mi> <mo>&amp;OverBar;</mo> </mover> <mi>N</mi> </msub> </mtd> </mtr> </mtable> </mfenced> </mrow>

Wherein, D represents reconstruct dimension, and D is natural number, and D ＜ N, D span are 50-70；

(5) obtained X, Y are substituted into following linear equation：

<mrow> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <msubsup> <mn>1</mn> <mi>v</mi> <mi>T</mi> </msubsup> </mtd> </mtr> <mtr> <mtd> <msub> <mn>1</mn> <mi>v</mi> </msub> </mtd> <mtd> <mrow> <mi>K</mi> <mo>+</mo> <msub> <mi>V</mi> <mi>&amp;gamma;</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msup> <mi>b</mi> <mo>*</mo> </msup> </mtd> </mtr> <mtr> <mtd> <msup> <mi>&amp;alpha;</mi> <mo>*</mo> </msup> </mtd> </mtr> </mtable> </mfenced> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mi>Y</mi> </mtd> </mtr> </mtable> </mfenced> </mrow>

Wherein

Weight factor v_iCalculated by following formula：

WhereinIt is error variance ξ_iThe estimation of standard deviation, c₁,c₂For constant；

Solve to obtain function f (x) to be estimated：

<mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <msup> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>*</mo> </msup> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mo>|</mo> <mo>|</mo> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>|</mo> <mo>|</mo> <mo>/</mo> <msup> <mi>&amp;theta;</mi> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mo>+</mo> <msup> <mi>b</mi> <mo>*</mo> </msup> </mrow>

Wherein, M is the number of supporting vector, 1_v=[1 ..., 1]^T,The transposition of subscript T representing matrixs,It is Lagrange multiplier, b^*It is amount of bias, K=exp (- | | x_i-x_j||/θ²), wherein i=1 ..., M, j=1 ..., M,With exp (- | | x-x_i||/θ²) be SVMs kernel function, x_jFor j-th of radar sea clutter Echo-signal amplitude, θ are nuclear parameters, and x represents input variable, and γ is penalty coefficient；

(6) manually ant colony algorithm optimizes to the nuclear parameter θ and penalty coefficient γ of step (5), is completed using following process：

(6.1) parameter of artificial bee colony algorithm is initialized, if nectar source number P, greatest iteration number iter_max, initial ranging space is most Small value and maximum L_dAnd U_d；The feasible solution of the positional representation problem in nectar source, because model has two parameters to need to optimize, so Position p_iDimension for 2 dimension, as the following formula at random generation nectar source position p_i=(p_i1,p_i2), put primary iteration number iter=0；

p_ij=L_d+rand()*(U_d-L_d) (i=1,2 ..., P, j=1,2)

(6.2) it is nectar source p_iDistribution one leads honeybee, scans for as the following formula, produces new nectar source V_i；

(6.3) V is calculated_iFitness value, according to greediness selection method determine retain nectar source；

(6.4) calculate lead the nectar source that honeybee is found by more with probability；

(6.5) honeybee is followed to use with leading honeybee identical mode to scan for, the honey for determining to retain according to the method for greediness selection Source；

(6.6) nectar source V is judged_iWhether the condition that satisfaction is abandoned, such as meet, it is corresponding to lead honeybee role to be changed into search bee, otherwise Pass directly to step (6.8)；

(6.7) search bee randomly generates new nectar source；

(6.8) iter=iter+1, judge whether to have been maxed out iterations, export optimized parameter if meeting, otherwise Go to step (6.2).

Wherein, nectar source number is 100, the minimum value and maximum 0 and 100 in initial ranging space, maximum iteration 100.

(7) gather D sea clutter echo-signal amplitude in sampling instant t and obtain TX=[x_t-D+1,...,x_t], x_t-D+1Represent t- The sea clutter echo-signal amplitude of D+1 sampling instants, x_tRepresent the sea clutter echo-signal amplitude of t sampling instants；

(8) it is normalized；

<mrow> <mover> <mrow> <mi>T</mi> <mi>X</mi> </mrow> <mo>&amp;OverBar;</mo> </mover> <mo>=</mo> <mfrac> <mrow> <mi>T</mi> <mi>X</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> <mrow> <mi>max</mi> <mi> </mi> <mi>x</mi> <mo>-</mo> <mi>min</mi> <mi> </mi> <mi>x</mi> </mrow> </mfrac> </mrow>

(9) the sea clutter predicted value that sampling instant (t+1) is calculated in the function f (x) to be estimated that step (5) obtains is substituted into.

(10) difference e of sea clutter predicted value and radar return measured value is calculated, calculates control limit Q_α：

<mrow> <msub> <mi>Q</mi> <mi>&amp;alpha;</mi> </msub> <mo>=</mo> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> <msup> <mrow> <mo>&amp;lsqb;</mo> <mfrac> <mrow> <msub> <mi>C</mi> <mi>&amp;alpha;</mi> </msub> <msub> <mi>h</mi> <mn>0</mn> </msub> <msqrt> <mrow> <mn>2</mn> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> </mrow> </msqrt> </mrow> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> </mfrac> <mo>+</mo> <mn>1</mn> <mo>+</mo> <mfrac> <mrow> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> <msub> <mi>h</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <msub> <mi>h</mi> <mn>0</mn> </msub> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> <msub> <mi>&amp;theta;</mi> <mn>2</mn> </msub> </mfrac> <mo>&amp;rsqb;</mo> </mrow> <mfrac> <mn>1</mn> <msub> <mi>h</mi> <mn>0</mn> </msub> </mfrac> </msup> </mrow>

<mrow> <msub> <mi>&amp;theta;</mi> <mi>i</mi> </msub> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msubsup> <mi>&amp;lambda;</mi> <mi>j</mi> <mi>i</mi> </msubsup> <mo>,</mo> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>3</mn> </mrow>

<mrow> <msub> <mi>h</mi> <mn>0</mn> </msub> <mo>=</mo> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mn>2</mn> <msub> <mi>&amp;theta;</mi> <mn>1</mn> </msub> <msub> <mi>&amp;theta;</mi> <mn>3</mn> </msub> </mrow> <mrow> <mn>3</mn> <msubsup> <mi>&amp;theta;</mi> <mn>2</mn> <mn>2</mn> </msubsup> </mrow> </mfrac> </mrow>

Wherein, α is confidence level, θ₁,θ₂,θ₃,h₀It is intermediate variable, λ_j ⁱThe i powers of j-th of characteristic value of covariance matrix are represented, K is sample dimension, C_αIt is the statistics that normal distribution confidence level is α；

(11) detection judgement is carried out：Work as e²Difference is more than control limit Q_αWhen, there is target in the point, otherwise without target.

(12) by the sampling time interval gathered data of setting, by obtained measured data compared with model prediction value, if phase 10% is more than to error, then new data is added into training sample data, updates forecasting model.