CN103745117A - Decision probability transformation method for target identification - Google Patents

Abstract

The invention belongs to the information processing technology field, and specifically relates to a decision probability transformation method for target identification. The decision probability transformation method includes the steps of determining an identification framework and obtaining the basic probability assignment of the identification targets under the framework, calculating the decision probability P1(Xi), i=1,2,..., n, of each identification target, calculating the decision probability P2(Xj), j=1,2,..., n, of each identification target, and fusing P1(Xi) and P2(Xj) according to a D-S evidence fusion rule to obtain the final decision probability P(Xp) of each identification target Xp. The decision probability transformation method is capable of distributing uncertain information by virtue of the reliability ratio and the likelihood ratio of a single-target proposition in a multi-target proposition; the distribution basis is quite objective; the uncertain information can be distributed rationally. The decision probabilities calculated based on the reliability ratio and the likelihood ratio are fused according to the D-S evidence fusion rule, so that the transformation attitude is neither optimistic or conservative, and the obtained result is more rational and effectively. The decision probability transformation method does not involve with complex high-order operation, and is low in calculation amount and convenient to operate.

Description

A kind of decision probability conversion method for target identification

Technical field

The invention belongs to technical field of information processing, be specifically related to a kind of decision probability conversion method for target identification.

Background technology

In high-tech war, enemy is attacked to effective detection and the early warning of target, to the high-resolution reconnaissance and surveillance of Research on Target, all require armament systems to identify round-the-clock, at a distance, rapidly and accurately target, for decision system provides strong decision support.Target identification technology has huge application value in army in modern war, has become the important support technology of current intellectual weapon system, and deep research is all carried out at a large amount of talent and the funds of input in countries in the world.The main method of target identification is to utilize the prospecting tools such as radar sensor, infrared sensor to survey identified region, by detecting target signature, obtain identifying some basic probability assignments (Basic Probability Assignment of target, BPA), then according to BPA, differentiate target, for aid decision making.But often sensor can not accurately be identified the target of feature similarity, this will obtain some BPA that contain multiple target propositions.So, in order can reasonably to make a policy, the BPA in multiple goal proposition need to be distributed to single goal proposition, to obtain suitable decision probability.Therefore, need to carry out decision probability conversion to BPA.

At present, BPA being carried out in the method for decision probability conversion, some methods are that the BPA in multiple goal proposition is all assigned in single goal proposition, easily cause information loss, are unfavorable for decision-making.Some methods have only been considered the impact of likelihood function on transfer process, and do not consider the impact of belief function, and this is incomplete.Some methods can be selected suitable conversion method according to different conversion attitudes, and the scope of application is wider, and the selection of still changing attitude lacks objective basis.

Summary of the invention

The decision probability conversion method for target identification that the object of the present invention is to provide a kind of decision maker of making to make a policy more rationally, accurately and rapidly.

The object of the present invention is achieved like this:

(1) determine identification framework, obtain the basic probability assignment of identifying target under framework;

By the definite identification framework Θ={ X of radar sensor ₁, X ₂..., X _n, X ₁, X ₂..., X _nfor the identification target that radar sensor observes, under identification framework, there is k basic probability assignment m ₁(), m ₂() ..., m _k(): 2 ^Θ→ [0,1], and meet $\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{m}_i(\·)=1;$

(2) calculate each identification target X _idecision probability P ₁(X _i), i=1,2 ..., n:

$P_1\left(X_i\right)=m\left(X_i\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{\mathrm{Bel}\left(X_i\right)}{\mathrm{\ΣBel}\left(X_i\right)}m\left(Y\right)$

Wherein, m (X _i) be identification target X _ibasic probability assignment; Bel (X _i) be identification target X _ireliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target;

(3) calculate each identification target X _jdecision probability P ₂(X _j), j=1,2 ..., n:

$P_2\left(X_j\right)=m\left(X_j\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{{\mathrm{Pl}}^{\′}\left(X_j\right)}{\mathrm{\Σ}{\mathrm{Pl}}^{\′}\left(X_j\right)}m\left(Y\right)$

Wherein, m (X _j) be identification target X _jbasic probability assignment; Bel (X _j) be identification target X _jreliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target; Pl'(X _j) be identification target X _jonly the likelihood score in Y, establishes Θ={ X ₁, X ₂, X ₃, m (Y)=m (X ₁∪ X ₂), X _j=X ₁,

Pl'(X ₁)＝m(X ₁)+m(X ₁∪X ₂)；

(4) according to D-S evidence fusion rule to P ₁(X _i) and P ₂(X _j) merge, draw each identification target X _pfinal decision probability P (X _p), p=1,2 ..., n:

$P\left(X_p\right)=\frac{\underset{X_i\∩X_j=X_p}{\mathrm{\Σ}}{P}_1\left(X_i\right)P_2\left(X_j\right)}{1-K}$

Wherein, K is P ₁(X _i) and P ₂(X _j) between conflict coefficient.

Beneficial effect of the present invention is:

1. the present invention utilizes reliability ratio and the likelihood score ratio of single goal proposition in multiple goal proposition to distribute uncertain information, distributes according to more objective, can reasonably distribute uncertain information.

By D-S evidence fusion rule to based on reliability than and likelihood score than the decision probability calculating, merge, make to change attitude both pessimistic also conservative, the result obtaining is more reasonable effective.

3. the present invention does not relate to complicated high exponent arithmetic(al), and calculated amount is little, convenient operation.

Accompanying drawing explanation

Fig. 1 is calculation flow chart of the present invention.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described further.

For above-mentioned existing methodical deficiency, a kind of decision probability conversion method for target identification has been proposed.Because the reliability of proposition is the support that proposition is be sure of, the likelihood score of proposition is to the potential support of the maximum of proposition, so the present invention utilizes the reliability of proposition and the feature of likelihood score, proposed respectively based on reliability than and likelihood score than the method for switch decision probability.Consider based on reliability than the method for switch decision probability, attitude during conversion is too optimistic, easily increases risk of policy making; Based on likelihood score, than the method for switch decision probability, attitude during conversion is too conservative again, is unfavorable for decision-making.So, in order to bring into play better the advantage of the two, make up the deficiency of the two, by D-S evidence fusion rule, the two is merged, draw final decision probability, make transformation result more rationally effectively.The method can reasonably be distributed uncertain information, reduces uncertainty, reduces risk of policy making, is convenient to decision maker rationally, make a policy accurately and rapidly.

The technical solution adopted in the present invention is a kind of decision probability conversion method for target identification, and the method comprises the following steps:

(1) determine identification framework, obtain the basic probability assignment of identifying target under framework.Suppose the identification framework Θ={ X definite by radar sensor ₁, X ₂..., X _n, X ₁, X ₂..., X _nfor the identification target that radar sensor observes, under this framework, there is k basic probability assignment m ₁(), m ₂() ..., m _k(): 2 ^Θ→ [0,1], and meet

(2) basis, based on reliability than the method for switch decision probability, calculates each identification target X _idecision probability P ₁(X _i), i=1,2 ..., n, computing formula is as follows:

$P_1\left(X_i\right)=m\left(X_i\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{\mathrm{Bel}\left(X_i\right)}{\mathrm{\ΣBel}\left(X_i\right)}m\left(Y\right)---\left(1\right)$

Wherein, m (X _i) be identification target X _ibasic probability assignment; Bel (X _i) be identification target X _ireliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target.

(3) basis, based on likelihood score than the method for switch decision probability, calculates each identification target X _jdecision probability P ₂(X _j), j=1,2 ..., n, computing formula is as follows:

$P_2\left(X_j\right)=m\left(X_j\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{{\mathrm{Pl}}^{\′}\left(X_j\right)}{\mathrm{\Σ}{\mathrm{Pl}}^{\′}\left(X_j\right)}m\left(Y\right)---\left(2\right)$

Wherein, m (X _j) be identification target X _jbasic probability assignment; Bel (X _j) be identification target X _jreliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target; Pl'(X _j) be identification target X _jonly the likelihood score in Y, for example, suppose Θ={ X ₁, X ₂, X ₃, m (Y)=m (X ₁∪ X ₂), X _j=X ₁, Pl'(X ₁)=m (X ₁)+m (X ₁∪ X ₂).

(4) according to D-S evidence fusion rule to P ₁(X _i) and P ₂(X _j) merge, draw each identification target X _pfinal decision probability P (X _p), p=1,2 ..., n, computing formula is as follows:

$P\left(X_p\right)=\frac{\underset{X_i\∩X_j=X_p}{\mathrm{\Σ}}{P}_1\left(X_i\right)P_2\left(X_j\right)}{1-K}---\left(4\right)$

Wherein, K is P ₁(X _i) and P ₂(X _j) between conflict coefficient.

Below by a concrete example, the present invention is described in detail.

(1) determine identification framework, obtain the basic probability assignment of identifying target under framework.Suppose an existing radar sensor, with it, observe aerial airbound target, the target that may observe has F-15, Mig-27 and Boeing-747, identification framework Θ={ F, M, B}, wherein F, M, B, represents respectively F-15, Mig-27 and Boeing-747, and the basic probability assignment of identifying target under this framework is as shown in table 1.

Table 1 basic probability assignment

Subsets	{F}	{M}	{B}	{F∪M}	{F∪M∪B}
						m(·)	0.4	0.2	0.1	0.2	0.1

(2) basis, based on reliability than the method for switch decision probability, calculates each identification target X _idecision probability P ₁(X _i), i=1,2,3, computing formula is as follows:

$P_1\left(X_i\right)=m\left(X_i\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{\mathrm{Bel}\left(X_i\right)}{\mathrm{\ΣBel}\left(X_i\right)}m\left(Y\right)---\left(1\right)$

Wherein, m (X _i) be identification target X _ibasic probability assignment; Bel (X _i) be identification target X _ireliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target.

According to formula (1), calculate respectively the decision probability of F, M and B:

$\begin{array}{c}{P}_1\left(F\right)=m\left(F\right)+\frac{m\left(F\right)}{m\left(F\right)+m\left(M\right)}m(F\∪M)+\frac{m\left(F\right)}{m\left(F\right)+m\left(M\right)+m\left(B\right)}m(F\∪M\∪B)\\ =0.4+0.4*0.2/0.6+0.4*0.1/0.7\\ =0.5905\end{array}$

$\begin{array}{c}{P}_1\left(M\right)=m\left(M\right)+\frac{m\left(M\right)}{m\left(F\right)+m\left(M\right)}m(F\∪M)+\frac{m\left(M\right)}{m\left(F\right)+m\left(M\right)+m\left(B\right)}m(F\∪M\∪B)\\ =0.2+0.2*0.2/0.6+0.2*0.1/0.7\\ =0.2952\end{array}$

$\begin{array}{c}{P}_1\left(B\right)=m\left(B\right)+\frac{m\left(B\right)}{m\left(F\right)+m\left(M\right)+m\left(B\right)}m(F\∪M\∪B)\\ =0.1+0.1*0.1/0.7\\ 0.1143\end{array}$

(3) basis, based on likelihood score than the method for switch decision probability, calculates each identification target X _jdecision probability P ₂(X _j), j=1,2,3, computing formula is as follows:

$P_2\left(X_j\right)=m\left(X_j\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{{\mathrm{Pl}}^{\′}\left(X_j\right)}{\mathrm{\Σ}{\mathrm{Pl}}^{\′}\left(X_j\right)}m\left(Y\right)---\left(2\right)$

Wherein, m (X _j) be identification target X _jbasic probability assignment; Bel (X _j) be identification target X _jreliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target; Pl'(X _j) be identification target X _jthe only likelihood score in Y, for example in this embodiment, as m (Y)=m (X ₁∪ X ₂), X _j=X ₁time,

Pl'(X ₁)＝m(X ₁)+m(X ₁∪X ₂)。

According to formula (2), calculate respectively the decision probability of F, M and B:

$\begin{array}{c}{P}_2\left(F\right)=m\left(F\right)+\frac{m\left(F\right)+m(F\∪M)}{m\left(F\right)+m(F\∪M)+m\left(M\right)+m(F\∪M)}m(F\∪M)\\ +\frac{m\left(F\right)+m(F\∪M)+m(F\∪M\∪B)}{m\left(F\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(M\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(B\right)+m(F\∪M\∪B)}m(F\∪M\∪B)\\ =0.4+0.6*0.2/1+0.7*0.1/1.4\\ =0.5700\end{array}$

$\begin{array}{c}{P}_2\left(M\right)=m\left(M\right)+\frac{m\left(M\right)+m(F\∪M)}{m\left(F\right)+m(F\∪M)+m\left(M\right)+m(F\∪M)}m(F\∪M)\\ +\frac{m\left(M\right)+m(F\∪M)+m(F\∪M\∪B)}{m\left(F\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(M\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(B\right)+m(F\∪M\∪B)}m(F\∪M\∪B)\\ =0.2+0.4*0.2/1+0.5*0.1/1.4\\ =0.3157\end{array}$

$\begin{array}{c}{P}_2\left(B\right)=m\left(B\right)+\frac{m\left(B\right)+m(F\∪M\∪B)}{m\left(F\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(M\right)+m(F\∪M)+m(F\∪M\∪B)+m\left(B\right)+m(F\∪M\∪B)}m(F\∪M\∪B)\\ =0.1+0.2*0.1/1.4\\ =0.1143\end{array}$

(4) according to D-S evidence fusion rule to P ₁(X _i) and P ₂(X _j) merge, draw each identification target X _pfinal decision probability P (X _p), p=1,2,3, computing formula is as follows:

Wherein, K is P ₁(X _i) and P ₂(X _j) between conflict coefficient.

$P\left(X_p\right)=\frac{\underset{X_i\∩X_j=X_p}{\mathrm{\Σ}}{P}_1\left(X_i\right)P_2\left(X_j\right)}{1-K}---\left(4\right)$

According to formula (3), calculate P ₁(X _i) and P ₂(X _j) between conflict coefficient K:

K＝0.5905*(0.3157+0.1143)+0.2952*(0.57+0.1143)+0.1143*(0.57+0.3157)

＝0.5572

According to formula (4) respectively to P ₁and P (F) ₂(F), P ₁and P (M) ₂(M), P ₁and P (B) ₂(B) merge, draw the final decision probability of F, M, B:

P(F)＝0.5905*0.57/(1-0.5572)

＝0.7601

P(M)＝0.2952*0.3157/(1-0.5572)

＝0.2105

P(B)＝0.1143*0.1143/(1-0.5572)

＝0.0295

From fusion results, can find out, the target that radar sensor is more supported is F-15.

Claims (1)

1. for a decision probability conversion method for target identification, it is characterized in that, comprise the following steps:

(1) determine identification framework, obtain the basic probability assignment of identifying target under framework;

By the definite identification framework Θ={ X of radar sensor ₁, X ₂..., X _n, X ₁, X ₂..., X _nfor the identification target that radar sensor observes, under identification framework, there is k basic probability assignment m ₁(), m ₂() ..., m _k(): 2 ^Θ→ [0,1], and meet $\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{m}_i(\·)=1;$

(2) calculate each identification target X _idecision probability P ₁(X _i), i=1,2 ..., n:

$P_1\left(X_i\right)=m\left(X_i\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{\mathrm{Bel}\left(X_i\right)}{\mathrm{\ΣBel}\left(X_i\right)}m\left(Y\right)$

Wherein, m (X _i) be identification target X _ibasic probability assignment; Bel (X _i) be identification target X _ireliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target;

(3) calculate each identification target X _jdecision probability P ₂(X _j), j=1,2 ..., n:

$P_2\left(X_j\right)=m\left(X_j\right)+\underset{Y\∈2^{\mathrm{\Θ}},\left|Y\right|>1,X_i\⊆Y,\left|X_i\right|=1}{\mathrm{\Σ}}\frac{{\mathrm{Pl}}^{\′}\left(X_j\right)}{\mathrm{\Σ}{\mathrm{Pl}}^{\′}\left(X_j\right)}m\left(Y\right)$

Wherein, m (X _j) be identification target X _jbasic probability assignment; Bel (X _j) be identification target X _jreliability; Y is the proposition that contains multiple identification targets, and m (Y) is the basic probability assignment that contains the proposition of multiple identification target; Pl'(X _j) be identification target X _jonly the likelihood score in Y, establishes Θ={ X ₁, X ₂, X ₃, m (Y)=m (X ₁∪ X ₂), X _j=X ₁,

Pl'(X ₁)＝m(X ₁)+m(X ₁∪X ₂)；

(4) according to D-S evidence fusion rule to P ₁(X _i) and P ₂(X _j) merge, draw each identification target X _pfinal decision probability P (X _p), p=1,2 ..., n:

$P\left(X_p\right)=\frac{\underset{X_i\∩X_j=X_p}{\mathrm{\Σ}}{P}_1\left(X_i\right)P_2\left(X_j\right)}{1-K}$

Wherein, K is P ₁(X _i) and P ₂(X _j) between conflict coefficient.