CN103177290A - Identification method for model of ship domain based on online self-organization neural network - Google Patents

Abstract

The invention provides an identification method for a model of ship domain based on an online self-organization fuzzy neural network. The method comprises the following steps of: selecting a ship safety zone model and determining a function, an input variable and a desired output value of the model of the ship safety zone model; establishing a dynamic fuzzy neural network comprising an input layer, a membership function layer, a T-norm layer and an output layer; training the dynamic fuzzy neural network by a training dataset comprising an input variable and an output value of the model till achieving accuracy requirements; and taking sailing parameters of two corresponding ships as input variables and inputting the input variables into the trained ship safety zone model so as to obtain ship safety areas of the two ships. By adopting the technical scheme, the safety model modified by using the method has better accuracy and higher safety compared with the conventional model of the ship domain.

Description

A kind of discrimination method of the ship domain model based on online self organizing neural network

Technical field

The present invention relates to a kind of ship domain correction of the model method, relate in particular to a kind of discrimination method of the ship domain model based on online self organizing neural network.

Background technology

Marine intelligent transportation traffic becomes as the important component part of China's science and technology development strategy the emerging crossing research focus that vessel traffic and effective and science merge gradually.And for the research of the individual boats and ships behavior of maritime traffic system, seem particularly important.20th century six the seventies, Japan adds concept that rattan [1] proposes the navigation safety field so far, in document [2] [3] [4] [5], as can be known, the researcher has proposed the navigation safety domain model of various difformities, size.Have a wide range of applications in the modern ships field.But, can't form all the time a unified model, cause the main cause of the problems referred to above to have: the factor of the navigation environment that (1) is different causes producing the model of difformity, size; (2) most model all forms according to the method for statistics or simulated experiment; (3) existing model easy to understand all, go but be difficult to be applied in reality.Document [3] [4] has proposed a kind of hexagon ship domain model of complexity, determine each limit size with ship's speed and boats and ships convolution parameter, this model makes the boats and ships in the collision prevention situation be convenient to adopt evolution algorithm that its flight path is optimized, but its complexity is higher, the physics meaning is more ambiguous, is not easy to understand and practical application.[5] provide the quantization method of boats and ships field boundary in several situations in conjunction with factors such as ship turning performances, the maneuvering performance of boats and ships is embodied in the method, but the funtcional relationship between moulded dimension and influence factor is artificial given a kind of rough estimation equation.It should be noted that " cross-sectional area " model that [2] propose is to be put together by former and later two semiellipses, by decisions such as Ship Controling parameter and headways, is one of classical model.

On the other hand, fuzzy system, neural network and fuzzy neural network are fast-developing, and extraordinaryly approach because it has, generalization ability, be applied in industrial circle rapidly.When design fuzzy system, neural network and fuzzy neural network, all must first determine regular number hidden nodes alive, the mode learning algorithm of application error backpropagation is trained simultaneously.As everyone knows, the method pace of learning is slow, easily is absorbed in local minimum point.Therefore, in the urgent need to finding one for the Fast Learning method of real-time application.For addressing the above problem, the researcher has proposed dynamic neural network.But there is following shortcoming in D-FNN:

It is to use the standard Gaussian function that dynamic fuzzy neural network (DFNN) is divided in the input space, in its rule, the width of all Gaussian functions of input variable is all identical, this point and reality are usually inconsistent, particularly when input variable has very different operation interval.

Dynamic fuzzy neural network (DFNN) is no matter how its subordinate function distributes, and its subordinate function is all identical with the quantity of fuzzy rule.This causes some subordinate function overlappings, the fuzzy rule indigestion that extracts.

In dynamic fuzzy neural network (DFNN), the Gaussian function width of article one fuzzy rule is what choose at random.

Have too many predefined parameter in dynamic fuzzy neural network (DFNN), and these parameters all lack physical significance, thus more difficult when selecting these special parameters.Must carry out the renormalization of normalization and output quantity for dynamic fuzzy neural network (DFNN) input quantity, will add intensive like this, the time that reaches Approximation effect preferably is long.

Therefore, the present invention proposes the model of a novel intelligent ship domain based on " cross-sectional area " model and Self-organized Fuzzy Neural Network.

Summary of the invention

The present invention is directed to the proposition of above problem, and the discrimination method of a kind of ship domain model based on online self organizing neural network of development has following steps:

-select the safety of ship regional model, determine function, input variable and the desired output of this model;

-foundation comprises the dynamic fuzzy neural network of input layer, subordinate function layer, T-norm layer and output layer;

-use the training dataset of the input variable and the output valve that comprise described model, described dynamic fuzzy neural network is trained until reach accuracy requirement;

-with the navigational parameter of two corresponding boats and ships, the safety of ship regional model after complete as input variable input training obtains the safety of ship zone of two boats and ships.

Described safety of ship regional model is the cross-sectional area model: this model is approximate to be put together by former and later two semiellipses, and the function of this model is shown below:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=L+(1+s)T_{90}U\\ R_{\mathrm{ba}}=L+T_{90}U\\ S_b=B+(1+t)T_{90}{D}_T\end{array}\right.$

Wherein, R _bf, R _baAnd S _bRepresent respectively radius and the transversal radius of the forward and backward ellipse in zone, T ₉₀Be ship turning 90 required time, the D of degree _TFor tactical diameter, s and t are environmental parameter;

Input variable is: P ^k=[L ^k, B ^k, U ₁ ^k, U ₂ ^k, α ^k], desired output is:

K=1,2.....n, U ₂ ^kRepresent the object ship ship's speed; α ^kRepresent two ship angles.

Described dynamic fuzzy neural network specifically comprises:

Input layer: have a plurality of nodes, each node represents the linguistic variable of an input;

The subordinate function layer: have a plurality of nodes, each node represents a subordinate function, and described subordinate function uses Gaussian function to be expressed as follows:

${\mathrm{\&mu;}}_{\mathrm{ij}}=(x_i,c_{\mathrm{ij}},{{\mathrm{\&sigma;}}_{\mathrm{ij}}}^L,{{\mathrm{\&sigma;}}_{\mathrm{ij}}}^R)=\left\{\begin{array}{c}\exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_{\mathrm{ij}}}^R\right)}^2}],x_i\&GreaterEqual;c_{\mathrm{ij}}\\ \exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_{\mathrm{ij}}}^L\right)}^2}],x_i<c_{\mathrm{ij}}\end{array}\right.$

Wherein, i=1,2......r, j=1,2.....u, wherein μ _ijBe x _iJ subordinate function, c _ijX _iThe center of j Gaussian function, σ _j ^L, σ _j ^RRepresent respectively x _iThe left and right width of j subordinate function, r is the input variable number, u is the total regular number of system, x represents the linguistic variable of an input, x=[x ₁, x ₂... ..x _i], x _iThe numerical value of expression x on the i dimension;

T-norm layer: have a plurality of nodes, each node represents the IF-part of a possible fuzzy rule, and j rule is output as:

Wherein:

${\mathrm{\&sigma;}}_j\left(x_i\right)=\left\{\begin{array}{c}{{\mathrm{\&sigma;}}_j}^R,x_i\&GreaterEqual;c_{\mathrm{ij}}\\ {{\mathrm{\&sigma;}}_j}^L,x_i<c_{\mathrm{ij}}\end{array}\right.$

Output layer: have at least a node, each node in this layer represents respectively an output variable, and this output is the stack of all input signals:

Y is output variable, ω _jThe THEN-part, for TSK model: ω _j=α _0j+ α _1jx ₁+ α _2jx ₂... .. ,+α _rjx _r, j=1,2 ... .., u.

Three variablees for the cross-sectional area model build three independently broad sense self organizing neural networks:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=f_{R_{\mathrm{bf}}}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\\ R_{\mathrm{ba}}=f_{R_{\mathrm{ba}}}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\\ S_b=f_{S_b}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\end{array}\right.$

Described dynamic fuzzy neural network training comprises following steps:

-with the degree of membership of asymmetric Gaussian function computational data to each center, expectation quality k _eWith can hold effective radius k _d, k _e=max{e _maxβ ^k-1, e _min, k _d=max{d _maxγ ^k-1, d _min;

-computing system error: || e ^k||=|| t ^k-y ^k||; Calculate mahalanobis distance:

${\mathrm{disk}}_{\mathrm{jk}}\left(X^k\right)=\sqrt{(X^k-C_j)T{\mathrm{\&Sigma;}}_j\left(X^k\right)(X^k-C_j)},$

Wherein

And find minimum mahalanobis distance d _min, find out from the nearest node of this sample;

When (1) || e _i||＞k _e, d _min＞k _dThe time, need to increase a fuzzy rule;

(2) || e _i||≤k _e, d _min≤ k _dThe time, the broad sense self organizing neural network can hold this data fully;

(3) || e _i||≤k _e, d _min＞k _dThe time, the broad sense self organizing neural network has preferably generalization ability only need adjust result parameter;

(4) || e _i||＞k _e, d _min≤ k _dThe time, adjust the RBF node that covers these data and upgrade result parameter;

-definition error rate of descent:

${\mathrm{err}}_i=\frac{{\left(p_i^{T}T\right)}^2}{p_i^T{p}_i{T}^{T}T},\mathrm{err}\&Element;R^{1*(r+1)v};$ Err is converted to: ERR=[ρ ₁, ρ ₂... ρ _u], ERR ∈ R ^{(r+1) u}

Define and calculate the importance of every rule:

${\mathrm{sig}}_j=\sqrt{\frac{{\mathrm{\&rho;}}_j^T{\mathrm{\&rho;}}_j}{r+1}},j=\mathrm{1,2}.....u;$

Judgement: if sig _j＞k _s, delete the j rule, wherein k _sBe the predefine parameter, be the parameter of judgement sensitivity;

-repetition training, until the network output error reaches requirement, finish training, when having barrier or other boats and ships to enter into the resulting safety of ship of the present invention field, the data that the deck officer can obtain according to the present invention judge the ship trajectory that this ship is new, avoid bumping.

After training obtains dynamic fuzzy neural network, the performance of service test data set check dynamic neural network: respectively to R _bf, R _baAnd S _bMake root-mean-square error.

Owing to having adopted technique scheme, the present invention has proposed OSFNN at the discrimination method that has proposed a kind of ship domain model based on online self organizing neural network, is equivalent to the online Self-organized Fuzzy Neural Network of TSK fuzzy system on function.In principle, compare with the D-FNN based on the Gaussian function of symmetry, acceptance domain based on the OSFNN of asymmetric Gaussian function provides more flexible, nonlinear transformation is approached any one nonlinear system widely, so more complicated on algorithm, has more generality; From using, the fuzzy rule that this algorithm extracts has good intelligibility.What be worth to propose is, although OSFNN is more complicated than D-FNN, and, its required predefined parameter is but lacked than D-FNN, so it implements than D-FNN and is more prone to.Relative and traditional ship domain model, the security model through correction of the present invention has better precision, and security is also higher.

Description of drawings

Technical scheme for clearer explanation embodiments of the invention or prior art, the below will do one to the accompanying drawing of required use in embodiment or description of the Prior Art and introduce simply, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain according to these accompanying drawings other accompanying drawing.

Fig. 1 is the topology diagram of broad sense self organizing neural network of the present invention

Fig. 2 is the algorithm flow chart of broad sense self organizing neural network of the present invention

Fig. 3 is the cross-sectional area boats and ships regional model that the present invention adopts

Fig. 4 is R of the present invention _bfThe graph of a relation of neuronal quantity and training set sample size

Fig. 5 is R of the present invention _bfRoot-mean-square error and training set sample size graph of a relation

Fig. 6 is R of the present invention _baThe graph of a relation of neuronal quantity and training set sample size

Fig. 7 is R of the present invention _baRoot-mean-square error and training set sample size graph of a relation

Fig. 8 is S of the present invention _bThe graph of a relation of neuronal quantity and training set sample size

Fig. 9 is S of the present invention _bRoot-mean-square error and training set sample size graph of a relation

Embodiment

For the purpose, technical scheme and the advantage that make embodiments of the invention is clearer, below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is known complete description:

As Fig. 1-shown in Figure 9: a kind of discrimination method of the ship domain model based on online self organizing neural network mainly comprises the steps:

At first, select the safety of ship regional model, determine function, input variable and the desired output of this model:

The boats and ships regional model that adopts is " cross-sectional area " model, wherein R _bf, R _baAnd S _bRepresent respectively front and back radius and the transversal radius in zone, model is determined by following formula:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=L+(1+s)T_{90}U\\ R_{\mathrm{ba}}=L+T_{90}U\\ S_b=B+(1+t)T_{90}{D}_T\end{array}\right.$

Wherein, L, B, U represent respectively length and width and the speed T of boats and ships ₉₀Be ship turning 90 required time, the D of degree _TFor tactical diameter, s and t represent environmental parameter, environment has: two ships travel (stem to stem) relatively, two ships travel in opposite directions (stern is to fore), two ships intersect travel (ship side is to fore).(the former represents this ship, and the latter represents object ship):

When two fores are correct,

$\left\{\begin{array}{c}s=2-\mathrm{\&Delta;U}/U_1\\ t=1\end{array}\right.$

During two ship right-angled intersections,

$\left\{\begin{array}{c}s=2-\mathrm{\&alpha;}/\mathrm{\&pi;}\\ t=\mathrm{\&alpha;}/\mathrm{\&pi;}\end{array}\right.$

When two bow tails join,

$\left\{\begin{array}{c}s=1\\ t=1\end{array}\right..$

Then, set up the dynamic fuzzy neural network that comprises input layer, subordinate function layer, T-norm layer and output layer:

If the OSFNN dynamic fuzzy neural network is divided into four layers (as Fig. 1), be respectively: input layer, subordinate function layer, T-norm layer, output layer,

Input layer: each node represents respectively the linguistic variable of an input.

The subordinate function layer: each node DGF represents respectively a subordinate function, and this subordinate function represents with following Gaussian function:

${\mathrm{\&mu;}}_{\mathrm{ij}}=(x_i,c_{\mathrm{ij}},{{\mathrm{\&sigma;}}_j}^L,{{\mathrm{\&sigma;}}_j}^R)=\left\{\begin{array}{c}\exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_j}^R\right)}^2}],x_i\&GreaterEqual;c_{\mathrm{ij}}\\ \exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_j}^L\right)}^2}],x_i<c_{\mathrm{ij}}\end{array}\right.$

Wherein, i=1,2......r, j=1,2.....u, wherein μ _ijBe x _iJ subordinate function, c _ijX _iThe center of j Gaussian function, σ _j ^L, σ _j ^RRepresent respectively x _iThe left and right width of j subordinate function, r is the input variable number, u is the total regular number of system.

T-norm layer: each node in this layer represents respectively an IF-part in possible fuzzy rule.

Therefore j rule is output as:

Wherein ${\mathrm{\&sigma;}}_j\left(x_i\right)=\left\{\begin{array}{c}{{\mathrm{\&sigma;}}_j}^R,x_i\&GreaterEqual;c_{\mathrm{ij}}\\ {{\mathrm{\&sigma;}}_j}^L,x_i<c_{\mathrm{ij}}\end{array}\right..$

Output layer: each node in this layer represents respectively an output variable, if only have an output node, is single output, if a plurality of output nodes are arranged, is many result outputs, and this output is the stack of all input signals:

Wherein, y is output variable, ω _jThe THEN-part, for TSK model: ω _j=α _0j+ α _1jx ₁+ α _2jx ₂... .. ,+α _rjx _r, j=1,2 ... .., u.

The present invention need build three independently OSFNN, is respectively:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=f_{R_{\mathrm{bf}}}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\\ R_{\mathrm{ba}}=f_{R_{\mathrm{ba}}}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\\ S_b=f_{S_b}(L,B,U_1,{\mathrm{<figure-callout\; id="2"\; label="U"\; filenames="HDA00003010589300042.png"\; state="\{\{state\}\}">U</figure-callout>}}_2,\mathrm{\&alpha;})\end{array}\right..$

Use comprises the training dataset of input variable and the output valve of described model, described dynamic fuzzy neural network is trained until reach accuracy requirement:

(1) with the degree of membership of asymmetric Gaussian function computational data to each Gauss center, the predefine parameter of initialization system.Expectation quality ke with can hold effective radius kd.

K wherein _e=max{e _maxβ ^k-1, e _min, k _d=max{d _maxγ ^k-1, d _min.

(2) computing system error: || e ^k||=|| t ^k-y ^k||; Calculate mahalanobis distance:

${\mathrm{disk}}_{\mathrm{jk}}\left(X^k\right)=\sqrt{(X^k-C_j)T{\mathrm{\&Sigma;}}_j\left(X^k\right)(X^k-C_j)},$

Wherein

And find minimum mahalanobis distance, find out from the nearest node of this sample.

(3) judge according to mahalanobis distance:

1.||e _i||＞k _e, d _min＞k _dThe time, need to increase a fuzzy rule.

2.||e _i||≤k _e, d _min≤ k _dThe time, OSFNN can hold this data fully.

3.||e _i||≤k _e, d _min＞k _dThe time, OSFNN has preferably generalization ability only need adjust result parameter.

4.||e _i||＞k _e, d _min≤ k _dThe time, adjust the RBF node that covers these data and upgrade result parameter.

(4) use equation of linear regression T=Ψ A+E and QR to decompose Ψ=PQ, wherein T: the desired output vector; A: weight vector; Ψ: regression matrix; The E error vector; P: orthogonal matrix, P ∈ R ^n*vQ: upper triangular matrix, Q ∈ R ^v*v

Definition error rate of descent:

${\mathrm{err}}_i=\frac{{\left(p_i^{T}T\right)}^2}{p_i^T{p}_i{T}^{T}T},\mathrm{err}\&Element;R^{1*(r+1)v}.$

Err is converted to: ERR=[ρ ₁, ρ ₂... ρ _u], ERR ∈ R ^{(r+1) u}

Define and calculate the importance of every rule:

${\mathrm{sig}}_j=\sqrt{\frac{{\mathrm{\&rho;}}_j^T{\mathrm{\&rho;}}_j}{r+1}},j=\mathrm{1,2}.....u.$

Judgement: if sig _j＞k _s, delete the j rule.Repetition training until the network output error reaches requirement, finishes training.

Again obtained the check data collection by above-mentioned by model, be used for the performance of check OSFNN dynamic fuzzy neural network, and respectively to R _bf, R _baAnd S _bMake root-mean-square error, respectively as Fig. 4, shown in 5,6.

During ship's navigation, the R that object ship (may enter this ship security fields) is obtained _bf, R _baAnd S _bNumerical value, build new ship domain.

The above; only be the better embodiment of the present invention; but protection scope of the present invention is not limited to this; anyly be familiar with those skilled in the art in the technical scope that the present invention discloses; be equal to replacement or changed according to technical scheme of the present invention and inventive concept thereof, within all should being encompassed in protection scope of the present invention.

Claims (6)

1. discrimination method based on the ship domain model of online Self-organized Fuzzy Neural Network has following steps:

-select the safety of ship regional model, determine function, input variable and the desired output of this model;

-foundation comprises the dynamic fuzzy neural network of input layer, subordinate function layer, T-norm layer and output layer;

-use the training dataset of the input variable and the output valve that comprise described model, described dynamic fuzzy neural network is trained until reach accuracy requirement;

-safety of ship regional model after the navigational parameter of this ship is complete as input variable input training obtains the navigation safety zone of this ship.

2. the discrimination method of a kind of ship domain model based on online self organizing neural network according to claim 1, be further characterized in that: described safety of ship regional model is the cross-sectional area model: this model is approximate to be put together by former and later two semiellipses, and the function of this model is shown below:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=L+(1+s)T_{90}U\\ R_{\mathrm{ba}}=L+T_{90}U\\ S_b=B+(1+t)T_{90}{D}_T\end{array}\right.$

Wherein, R _bf, R _baAnd S _bRepresent respectively radius and the transversal radius of the forward and backward ellipse in zone, T ₉₀Be ship turning 90 required time, the D of degree _TFor tactical diameter, s and t are environmental parameter;

Input variable is: P ^k=[L ^k, B ^k, U ₁ ^k, U ₂ ^k, α ^k], desired output is:

K=1,2.....n, U ₂ ^kRepresent the object ship ship's speed; α ^kRepresent two ship angles.

3. the discrimination method of a kind of ship domain model based on online self organizing neural network according to claim 1 and 2, be further characterized in that: described dynamic fuzzy neural network specifically comprises:

Input layer: have a plurality of nodes, each node represents the linguistic variable of an input;

The subordinate function layer: have a plurality of nodes, each node represents a subordinate function, and described subordinate function uses Gaussian function to be expressed as follows:

${\mathrm{\&mu;}}_{\mathrm{ij}}=(x_i,c_{\mathrm{ij}},{{\mathrm{\&sigma;}}_{\mathrm{ij}}}^L,{{\mathrm{\&sigma;}}_{\mathrm{ij}}}^R)=\left\{\begin{array}{c}\exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_{\mathrm{ij}}}^R\right)}^2}],x_i\&GreaterEqual;c_{\mathrm{ij}}\\ \exp [-\frac{{(x_i-c_{\mathrm{ij}})}^2}{{\left({{\mathrm{\&sigma;}}_{\mathrm{ij}}}^L\right)}^2}],x_i<c_{\mathrm{ij}}\end{array}\right.$

Wherein, i=1,2......r, j=1,2.....u, wherein μ _ijBe x _iJ subordinate function, c _ijX _iThe center of j Gaussian function, σ _j ^L, σ _j ^RRepresent respectively x _iThe left and right width of j subordinate function, r is the input variable number, u is the total regular number of system, x represents the linguistic variable of an input, x=[x ₁, x ₂... ..x _i], x _iThe numerical value of expression x on the i dimension;

T-norm layer: have a plurality of nodes, each node represents the IF-part of a possible fuzzy rule, and j rule is output as:

Wherein

${\mathrm{\&sigma;}}_j\left(x_i\right)=\left\{\begin{array}{c}{{\mathrm{\&sigma;}}_j}^R,x_i\&GreaterEqual;c_{\mathrm{ij}}\\ {{\mathrm{\&sigma;}}_j}^L,x_i<c_{\mathrm{ij}}\end{array}\right.$

Output layer: have at least a node, each node in this layer represents respectively an output variable, and this output is the stack of all input signals:

Y is output variable, ω _jThe THEN-part, for TSK model: ω _j=α _0j+ α _1jx ₁+ α _2jx ₂... .. ,+α _rjx _r, j=1,2 ... .., u.

4. the discrimination method of a kind of ship domain model based on online self organizing neural network according to claim 3 is further characterized in that:

Three variablees for the cross-sectional area model build three independently online self organizing neural networks:

$\left\{\begin{array}{c}{R}_{\mathrm{bf}}=f_{R_{\mathrm{bf}}}(L,B,U_1,U_2,\mathrm{\&alpha;})\\ R_{\mathrm{ba}}=f_{R_{\mathrm{ba}}}(L,B,U_1,U_2,\mathrm{\&alpha;})\\ S_b=f_{S_b}(L,B,U_1,U_2,\mathrm{\&alpha;})\end{array}\right..$

5. the discrimination method of a kind of ship domain model based on online self organizing neural network according to claim 3 is further characterized in that described dynamic fuzzy neural network training comprises following steps:

-with the degree of membership of asymmetric Gaussian function computational data to each Gauss center, expectation quality k _eWith can hold effective radius k _d, k _e=max{e _maxβ ^k-1, e _min, k _d=max{d _maxγ ^k-1, d _min;

-computing system error: || e ^k||=|| t ^k-y ^k||; Calculate mahalanobis distance:

${\mathrm{disk}}_{\mathrm{jk}}\left(X^k\right)=\sqrt{(X^k-C_j)T{\mathrm{\&Sigma;}}_j\left(X^k\right)(X^k-C_j)},$

Wherein

And find minimum mahalanobis distance d _min, find out from the nearest GEBF node of this sample;

When (1) || e _i||＞k _e, d _min＞k _dThe time, need to increase a fuzzy rule;

(2) || e _i||≤k _e, d _min≤ k _dThe time, the broad sense self organizing neural network can hold this data fully;

(3) || e _i||≤k _e, d _min＞k _dThe time, the broad sense self organizing neural network has preferably generalization ability only need adjust result parameter;

(4) || e _i||＞k _e, d _min≤ k _dThe time, adjust the RBF node that covers these data and upgrade result parameter;

-definition error rate of descent:

Use equation of linear regression T=Ψ A+E and QR to decompose Ψ=PQ, wherein T: the desired output vector; A: weight vector; Ψ: regression matrix; The E error vector; P: orthogonal matrix, P ∈ R ^n*vQ: upper triangular matrix, Q ∈ R ^v*v

${\mathrm{err}}_i=\frac{{\left(p_i^{T}T\right)}^2}{p_i^T{p}_i{T}^{T}T},\mathrm{err}\&Element;R^{1*(r+1)v};$ Err is converted to: ERR=[ρ ₁, ρ ₂... ρ _u], ERR ∈ R ^{(u+1) * u}

Define and calculate the sensitivity of every rule:

${\mathrm{sig}}_j=\sqrt{\frac{{\mathrm{\&rho;}}_j^T{\mathrm{\&rho;}}_j}{r+1}},j=\mathrm{1,2}.....u;$

Judgement: if sig _j＞k _s, delete the j rule, wherein k _sBe the predefine parameter, be the parameter of judgement sensitivity;

-repetition training, until the network output error reaches requirement, finish training, when having barrier or other boats and ships to enter into the resulting safety of ship of the present invention field, the data that the deck officer can obtain according to the present invention judge the ship trajectory that this ship is new, avoid bumping.

6. the discrimination method of a kind of ship domain model based on online self organizing neural network according to claim 5 is further characterized in that: after training obtains dynamic fuzzy neural network, and the performance of service test data set check dynamic neural network: respectively to R _bf, R _baAnd S _bMake root-mean-square error, represent the performance of this dynamic fuzzy neural network.

Images