A method and system for information fusion based on DS evidence theory

Abstract

The application discloses a kind of information fusion method and system based on DS evidence theory, it is related to information fusion technical field;First, DS identification evidence is acquired and uncertainty is measured;Then, the DS identification evidence is re-distributed and screened with credibility, then the DS identification evidence is combined according to compatible mutual exclusion history identification evidence, and finally the identity attribute identification decision is output.This scheme realizes the unified quantization of the heterogeneous information of multiple sensors through a series of improvements on the existing DS evidence theory fusion method, ensures the effective perception of the uncertainty or error of multiple identification processes, ensures the objectivity and correctness of the identification evidence measurement, effectively solves the conflict fusion problem of the independence and compatible mutual exclusion of different conflict intensity evidence fusion and identification evidence, and improves the effectiveness and correctness of the fusion result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information fusion technology, and in particular to an information fusion method and system based on DS evidence theory. BACKGROUND

[0002] Dempster-Shafer (DS) evidence theory is the main theoretical method for fusion processing at present, and mainly faces two technical key problems: one is how to effectively establish DS recognition evidence, that is, how to calculate the belief function of the DS recognition framework subset; and the other is how to effectively solve the evidence conflict problem in DS recognition evidence fusion.

[0003] For the first type of technical key point, the problem is where the evidence comes from when applying DS evidence theory. At present, the main method for establishing DS recognition evidence is that experts generate a basic belief assignment on the recognition framework according to their own knowledge and experience. After obtaining multiple belief assignments, the belief functions are synthesized according to the DS evidence combination rule to obtain the decision result. However, the method based on experts is limited by the limitations of the knowledge and experience of experts, and the DS evidence belief function evaluation method established by experts in a certain field is not applicable to other fields, and lacks generality.

[0004] For the second type of technical key point, a series of improvement methods have been formed for the three elements of the recognition framework, the recognition evidence belief function and the combination rule of the DS evidence fusion theory, but the main problems include that the existing DS conflict evidence screening method mainly measures the evidence fusion weight from the distance measurement between single-period evidences, does not fully consider the consistency of time series, and the conflict fusion mechanism only starts from the weight correction, does not consider the reliability of the sensor itself where the recognition evidence comes from, and is easy to lead to invalid fusion results; at the same time, the fusion of historical recognition evidence only starts from the improvement research on the combination method of the mathematical combination rule between evidences, does not consider the logical compatibility or conflict between various types of evidence to utilize the fusion of multi-period discrete evidence, and is easy to lead to unreasonable fusion results.

[0005] The existing DS evidence theory fusion technology lacks the ability to objectively measure the uncertainty of recognition evidence and effectively quantify it, and converts sensor information into DS recognition evidence propositions;

[0006] The existing DS evidence theory fusion technology measures the evidence fusion weight from the distance measurement between single-period evidences, does not fully consider the consistency of time series, and the conflict fusion mechanism only starts from the weight correction, does not consider the reliability of the sensor itself where the recognition evidence comes from;

[0007] The existing DS evidence theory fusion technology only improves the combination mode from the mathematical combination rule between evidences, and does not consider the logical compatibility or conflict between various evidences to utilize the fusion of multi-period discrete evidences. SUMMARY

[0008] The technical problem to be solved by the present application is to provide an information fusion method capable of objectively measuring the uncertainty of identification evidence and effectively solving the conflict of identification evidence, and improve the correctness and reliability of identification.

[0009] The present application is implemented by the following technical solutions:

[0010] The present application provides an information fusion method based on DS evidence theory, comprising:

[0011] Obtaining DS identification evidence and measuring uncertainty, wherein the DS identification evidence comprises identification evidence of an encrypted cooperative sensor, identification evidence of a non-cooperative sensor and identification evidence based on target position information;

[0012] The DS identification evidence is subjected to belief reassignment and screening: the identity attribute of the DS identification evidence is subjected to multiplicative weighting in combination with the identification stability of the sensor; and in the case of conflict of the DS identification evidence, the identity attribute after multiplicative weighting is subjected to conflict processing, and a corresponding fusion method is dynamically selected to complete conflict information fusion;

[0013] The DS identification evidence is subjected to historical identification evidence combination based on the compatibility and exclusivity of the DS identification evidence, and finally outputs an identity attribute identification decision.

[0014] The working principle of the present application is that the present application provides an information fusion method and system based on DS evidence theory, which realizes the unified quantization of heterogeneous information of multiple sensors through a series of improvements on the existing DS evidence theory fusion method, guarantees the effective perception of the uncertainty or error of multiple identification processes, guarantees the objectivity and correctness of the identification evidence measurement, effectively solves the conflict fusion problem of the independence and compatibility of different conflict intensity evidence fusion and identification evidence, and improves the effectiveness and correctness of the fusion result.

[0015] Further, the acquisition method of the identification evidence of the encrypted cooperative sensor comprises:

[0016] generating the identification evidence of the encryption collaborative sensor according to whether the encryption collaborative sensor obtains the valid identification result;

[0017] The collaborative sensor makes the attribute of the target be directly judged through the information interaction with the target by means of the time correction, signal format, encryption mode and the like agreed by both parties. The identification evidence generation process of the collaborative sensor is expressed by the production rule of "if…then…", that is, if the encryption collaborative sensor obtains the valid identification result, the current target identification evidence of the encryption collaborative sensor is directly generated. The uncertainty measure of the evidence is calculated based on the probability estimation method, and the single identification confidence is estimated by using the global parameter of the identification probability.

[0018] The identification system makes a decision or classification according to the target information S, and the possible result set of the decision or classification is Ω′2={ω0,ω1,...,ω M-1}; and D j is the real identity of the target.

[0019] The uncertainty measure of the identification evidence of the encryption collaborative sensor is the confidence expression C(k) of the decision result ω i (i=0,1,...,M-1) of the identification system under the input of the target information S.

[0020] C(k)=E(P(D j =ω i ))=E(P(ω i |D j )),(i,j=0,1,...,M-1); wherein D j is the actual identity attribute set of the target, and the number of the set members is M; ω i is the decision target identity attribute; P(ω i |D j ) represents the conditional probability that the target with the attribute D j is decided to have the attribute ω i ; and P(D j =ω i ) represents the probability that the identification decision is correct.

[0021] A further optimization scheme is that the identification evidence of the non-collaborative sensor is obtained by the following method:

[0022] The basic information of the non-collaborative sensor is obtained based on the target radiation source, and the basic information includes the target type and model. The mapping relationship is constructed based on the basic information to obtain the identification evidence of the non-collaborative sensor.

[0023] The identification evidence uncertainty measure of the non-cooperative sensor is a target identification confidence report of the sensor.

[0024] The target friend-or-foe attribute identification evidence is obtained based on the mapping relationship of the model-nation-attribute of friend or foe according to the target type / model information provided by the non-cooperative sensor.

[0025] The further optimization scheme is that the identification evidence based on the target position information is obtained by the following method:

[0026] The fuzzy membership functions of the yaw, the heading and the height of the target are respectively established by decomposing the elements of the navigation plan to describe the degree of the legality of the target position; the target attribute direction matching the navigation plan is given as the identification evidence based on the target position information by the legality threshold;

[0027] The target attribute direction uncertainty is calculated according to the fuzzy membership degree.

[0028] The further optimization scheme is that the identification evidence based on the target position information is obtained by the following method:

[0029] The distance d is the yaw distance, the yaw membership μ d is:

[0030]

[0031] Wherein, ε1 is the yaw distance defined by the navigation plan, and ε2 is the maximum yaw distance.

[0032] The heading membership μ c is:

[0033]

[0034] Wherein, θ i is the current heading of the target, and θ n is the defined heading of the navigation plan.

[0035] The height membership μ h is:

[0036]

[0037] Wherein, h is the height of the target, h' is the defined height of the navigation plan, and h ε is the maximum height deviation defined by the navigation plan.

[0038] The average membership reflecting the current legality of the target is obtained by the comprehensive calculation based on the yaw membership, the heading membership and the height membership, the legality evaluation of the current target navigation is obtained based on the average of the multiple time series memberships, and finally the threshold decision is made to form the attribute direction judgment supporting the behavior situation layer information:

[0039]

[0040] wherein μ t is the membership of the target position at time t to the compliance of the navigation plan, μ is the average membership of N times μ t , and σ is the legality threshold;

[0041] When the average membership μ exceeds the legality threshold σ, the identification evidence of the target position information points to {I} or {C} according to the target attribute of the current navigation plan;

[0042] When the average membership μ does not exceed the legality threshold σ, the identification evidence of the target position information points to {S}.

[0043] Further optimization scheme is that the DS identification evidence credibility reassignment method comprises:

[0044] The consistency of the DS identification evidence pointing is calculated from the time sequence, and the consistency evaluation result C L is taken as a multiplicative weighting parameter to adjust the uncertainty quantization result m(·) of the DS identification evidence;

[0045] When the identification result of a sensor to a target is stable, the credibility of the sensor is higher; and when the consistency of the identification result of a sensor to a target in each identification cycle is poor, the credibility of the sensor is lower.

[0046] The target identification confidence reports of the sensor in several identification cycles are R1, R2, …, R L , the number of elements of the identification framework is S, and then the sensor consistency C N of N identification cycles is:

[0047]

[0048] wherein C k,k+1 is the sensor identification consistency of adjacent two cycles, C T is an intermediate result of the sensor consistency calculation, and C N is the final result after normalization calculation based on C T .

[0049] Further optimization scheme is that the DS identification evidence screening method comprises:

[0050] The pointing consistency of the DS identification evidence is judged;

[0051] The DS identification evidence with the pointing consistency is measured: the DS identification evidence with the pointing consistency is processed in parallel; the DS identification evidence with the pointing inconsistency is processed in layers.

[0052] The layer fusion processing includes the following method: the identification evidence of the non-cooperative sensor is combined with the identification evidence based on the target position information, and the target attribute is judged, and then the target attribute judgment result is combined with the identification evidence of the encrypted cooperative sensor.

[0053] Further optimization scheme is that the pointing consistency judgment method between the DS identification evidence includes:

[0054] The two DS identification evidences E1 and E2 of the identification framework Θ correspond to the basic trust allocation functions m1 and m2, and the supported attribute pointing is A i and B j Whether the pointing of the two DS identification evidences is consistent is determined by the following formula:

[0055]

[0056] If the intersection of the subset A and the subset B is empty, the pointing of the two DS identification evidences is inconsistent; otherwise, the pointing of the two DS identification evidences is consistent, wherein m(*) represents the evidence quality of the subset (*).

[0057] In the case of pointing inconsistency, the pointing consistency is measured by G c (E1, E2).

[0058]

[0059] Wherein, Con(E1, E2) describes the conflict between the evidences, and H(E1, E2) describes the consistency between the identification evidences. The pointing consistency G c (E1, E2) is described by the proportion of the conflict confidence between the two identification evidences.

[0060] Suppose that σ is the conflict intensity judgment threshold, when G c (E1, E2) > σ, it indicates that the pointing consistency is not up to standard, and when the conflict intensity G c (E1, E2) ≤ σ, it indicates that the pointing consistency is up to standard.

[0061] In the case of pointing consistency up to standard, the two-by-two identification evidence combination processing is performed according to the processing criteria of the DS evidence theory. Assuming that the N-1th evidence quality is m1, and the Nth evidence quality is m2, then the combination calculation is performed in the following formula.

[0062]

[0063] Before all DS identification evidence participation combination is completed, the processing result of the above two two combinations does not carry out attribute pointing decision based on the highest confidence component, assuming that the highest confidence result of the identification evidence combination is D1, the pointing is Z1, the second highest confidence result is D2, and the pointing is Z2, then the fusion identification evidence participating in the next round of two two combinations is

[0064] In the case of not meeting the pointing consistency, first, the identification evidence of the non-collaborative sensor is combined with the identification evidence based on the target position information, the combination calculation method is consistent with the above combination method, but the decision processing is carried out, that is, according to the above assumption, given σ1, σ2 are two decision thresholds, the decision criterion is: if D1> σ1 and |D1-D2|> σ2, then the identification result is the highest confidence result; Otherwise, identify as unknown. The obtained target attribute decision result is combined with the encrypted collaborative identification evidence, and the combination method is consistent with the above combination method.

[0065] A further optimization scheme is that the historical identification evidence combination method comprises:

[0066] S1, first determine whether the DS identification evidence is independent, if yes, then change the attribute pointing of the DS identification evidence according to the consistency and exclusivity of the DS identification evidence, so that the DS identification evidence conflicts, and then enter step S3; Otherwise, determine whether the joint of the current DS identification evidence and the cache evidence is independent, if yes, then enter S3; Otherwise, enter S2;

[0067] S2, cache the current non-independent evidence, and change the attribute pointing of the DS identification evidence to unknown, and then enter S3;

[0068] S3, combine the current decision result with the historical identification result.

[0069] The independence of the evidence refers to that the identity attribute decision conclusion of the target can be formed only by relying on a certain identification evidence, and on the contrary, certain evidence needs the joint processing of other identification evidence to form a decision conclusion, and a single evidence cannot form an effective conclusion for identification; The consistency and exclusivity of the evidence refers to the conflict between multiple identification evidences in the fusion processing. Two identification evidences are not conflict is called consistent, and conflict is called exclusive. The independent consistency classification of the evidence can be referred to Table 1.

[0070] The present scheme also provides a DS evidence theory-based information fusion system for realizing the above DS evidence theory-based information fusion method, comprising:

[0071] The collection module is used for acquiring DS identification evidence, and the DS identification evidence includes identification evidence of an encryption cooperative sensor, identification evidence of a non-cooperative sensor and identification evidence based on target position information.

[0072] The distribution screening module is used for re-distribution and screening of the credibility of the DS identification evidence: consistency of pointing of the DS identification evidence is calculated from a time sequence, the DS identification evidence meeting the consistency of pointing is processed in parallel fusion, and the DS identification evidence not meeting the consistency of pointing is processed in hierarchical fusion.

[0073] The fusion module is used for historical identification result fusion based on the logical consistency and exclusivity of the DS identification evidence.

[0074] Compared with the prior art, the DS evidence theory-based information fusion method and system has the following advantages and beneficial effects:

[0075] The DS evidence theory-based information fusion method and system provided by the application realizes unified quantization of heterogeneous information of multiple sensors through a series of improvements on the existing DS evidence theory fusion method, guarantees effective perception of multiple process uncertainties or errors, guarantees objectivity and correctness of identification evidence measurement, effectively solves the conflict fusion problems of different conflict strength evidence fusion and identification evidence existence independence and consistency and exclusivity, and improves effectiveness and correctness of fusion results. BRIEF DESCRIPTION OF DRAWINGS

[0076] In order to more clearly illustrate the technical scheme of the exemplary embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be considered as a limitation to the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor. In the drawings:

[0077] Figure 1 It is a flowchart of the DS evidence theory-based information fusion method;

[0078] Figure 2 It is a schematic diagram of the overall information fusion processing flow;

[0079] Figure 3 It is a schematic diagram of the fusion processing flow under the condition that the consistency of pointing of the identification evidence meets the standard;

[0080] Figure 4 It is a schematic diagram of the fusion processing flow under the condition that the consistency of pointing of the identification evidence does not meet the standard;

[0081] Figure 5 It is a schematic diagram of the historical result fusion processing flow. DETAILED DESCRIPTION

[0082] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be given below in combination with embodiments and drawings, and the schematic embodiments and their descriptions are only used to explain the present application, but not to limit the present application.

[0083] Objective measurement of identification evidence uncertainty and effective solution to identification evidence conflict information fusion method, improve the correctness and reliability of identification, how to convert different types of sensor identification information into DS identification evidence proposition, is the key technical problem to be solved by the present application; how to effectively measure the weight of identification evidence, avoid the subjectivity of evidence screening, and on this basis, design a conflict processing method to solve the invalid results caused by parallel fusion of evidence, realize the correct and reliable fusion processing of conflict evidence, is the key technical problem to be solved by the present application; the interpretation method of the same type of features of different types of targets existing under different identification background conditions is different in the evidence conversion process, which will bring identification evidence conflict and even misidentification problem. For this kind of problem, how to judge the source of conflict identification evidence, and solve the conflict caused by information interpretation, is the key technical problem to be solved by the present application; the following embodiments solve the above technical problems:

[0084] Embodiment 1

[0085] The embodiment provides an information fusion method based on DS evidence theory, as shown in Figure 1 , comprising:

[0086] Obtaining DS identification evidence and measuring uncertainty, the DS identification evidence includes: identification evidence of encrypted cooperative sensor, identification evidence of non-cooperative sensor and identification evidence based on target position information;

[0087] The DS identification evidence is subjected to belief reassignment and screening: the identity attribute of the DS identification evidence is subjected to multiplicative weighting in combination with the identification stability of the sensor; and in the case of DS identification evidence conflict, the identity attribute after multiplicative weighting is subjected to conflict processing, and then the corresponding fusion method is dynamically selected to complete conflict information fusion;

[0088] Based on the compatibility and exclusivity of the DS identification evidence, the DS identification evidence is subjected to historical identification evidence combination, and finally the identity attribute identification decision is output.

[0089] As shown in Figure 2As shown, it gives the overall information fusion processing flow of the application, and the main flow includes three steps of DS identification evidence acquisition, identification evidence credibility reassignment and screening, and evidence combination. The DS identification evidence acquisition converts the identification information of different sensors, and quantifies the identification evidence about the identity attribute of the target; the credibility assignment and screening of the DS identification evidence is a weighted processing of the evidence, which is combined with the sensor identification stability evaluation to multiply the initial judgment conclusion of the identity attribute, reflects the uncertainty of the sensor information acquisition stage and the information correlation stage, and in the case of identification evidence conflict, the weighted initial judgment conclusion of the identity attribute is processed, according to the different inconsistency strength of the weighted identification evidence, a corresponding fusion structure is dynamically selected to complete the conflict information fusion; the evidence combination uses the discrete identification evidence on the multi-time sequence, based on the compatibility or mutual exclusivity between the discrete identification evidence, completes the fusion processing combined with the historical identification result, and outputs the final identity attribute judgment result of this processing.

[0090] The acquisition method of the identification evidence of the encrypted collaborative sensor includes:

[0091] According to whether the encrypted collaborative sensor obtains an effective identification result, the identification evidence of the encrypted collaborative sensor is generated;

[0092] Such collaborative sensors can directly determine the attribute through information interaction with the target by jointly agreeing on time correction, signal format, encryption method and other measures by the two parties.

[0093] The identification evidence generation process of such sensors is expressed by the production rule of "if…then…", that is, if the encrypted collaborative sensor obtains an effective identification result, the current target identification evidence of {I} is directly generated. The uncertainty measure of the evidence is calculated based on the probability estimation method, and the single identification confidence is estimated by using the global parameter of identification probability.

[0094] The identification system makes a decision or classification according to the target information S, and the possible result set of the decision or classification is D j is the true identity of the target;

[0095] The uncertainty measure of the identification evidence of the encrypted collaborative sensor is that the decision result is ω i (i=0,1,…,M-1) under the input target information S of the identification system, and the confidence expression C(k) of ω

[0096] C(k)=E(P(D j =ω i ))=E(P(ω i |D j), (i, j = 0, 1, …, M-1); wherein, D j is the actual identity attribute set of the target, and the number of set members is M; ω i is the target identity attribute to be determined; P(D i | ω j ) represents the conditional probability that the target with attribute D j is determined to have attribute ω i ; and P(D j = ω i ) represents the probability of correct identification.

[0097] The identification evidence of the non-cooperative sensor, and the acquisition method thereof comprises:

[0098] Based on the basic information of the non-cooperative sensor identified by the target radiation source, the basic information comprises the target type and model, and a mapping relationship is constructed based on the basic information to obtain the identification evidence of the non-cooperative sensor;

[0099] The uncertainty measurement of the identification evidence of the non-cooperative sensor is the target identification confidence report of the sensor.

[0100] The target type / model information given by the non-cooperative sensor is used to obtain the identification evidence of the enemy attribute of the target based on the mapping relationship of the model-nationality-enemy attribute.

[0101] The identification evidence based on the target position information, and the acquisition method thereof comprises:

[0102] By decomposing the elements of the navigation plan, the fuzzy membership functions of the yaw, heading and altitude of the target are respectively established to describe the degree of legitimacy of the target position; the target attribute direction matched with the navigation plan is given as the identification evidence based on the target position information through the legitimacy threshold;

[0103] The uncertainty of the target attribute direction is calculated according to the fuzzy membership degree.

[0104] The identification evidence based on the target position information, and the acquisition method thereof comprises:

[0105] The distance d is the yaw distance, and the yaw membership degree μ d is:

[0106]

[0107] Wherein, ε1 is the yaw distance defined by the navigation plan, and ε2 is the maximum yaw distance.

[0108] The heading membership degree μ c is:

[0109]

[0110] where θ i is the target current heading, θ n is the prescribed heading of the navigation plan;

[0111] height membership μ h is:

[0112]

[0113] where h is the target height, h' is the navigation plan limited height, h ε is the maximum height deviation limited by the navigation plan;

[0114] Based on the yaw membership, the heading membership and the height membership, the average membership reflecting the current legality of the target is calculated, and the legality evaluation of the current target navigation is obtained based on the average of multiple time series memberships. Finally, the threshold decision is made to form the attribute direction judgment of the support behavior situation layer information:

[0115]

[0116] where μ t is the membership of the target position at time t relative to the compliance of the navigation plan, μ is the average membership of N time μ t , and σ is the legality threshold;

[0117] When the average membership μ exceeds the legality threshold σ, the identification evidence of the target position information is directed to {I} or {civilian} according to the target attribute of the current navigation plan;

[0118] When the average membership μ does not exceed the legality threshold σ, the identification evidence of the target position information is directed to {suspected enemy}.

[0119] The belief reassignment method of the DS identification evidence includes:

[0120] The consistency of the DS identification evidence direction is calculated from the time series, and the consistency evaluation result C L is taken as a multiplicative weighting parameter to adjust the uncertainty quantization result m(·) of the DS identification evidence;

[0121] When the identification result of a sensor for a target is stable, the credibility of the sensor is higher; when the consistency of the identification result of a sensor for a target in each identification cycle is poor, the credibility of the sensor is lower.

[0122] The target identification confidence reports obtained by the sensor in several identification cycles are R1, R2, …, R L , the number of elements of the identification framework is S, and the sensor consistency C N of N identification cycles is:

[0123]

[0124] wherein C k,k+1 is the consistency of adjacent two periods of sensor identification, C T is the intermediate result of sensor consistency calculation, C N is the final result after normalization calculation based on C T .

[0125] The screening method of DS identification evidence includes:

[0126] Judging the pointing consistency between DS identification evidences;

[0127] For DS identification evidences with pointing consistency, performing pointing consistency measurement: parallel fusion processing of DS identification evidences with pointing consistency up to standard; hierarchical fusion processing of DS identification evidences with pointing consistency not up to standard;

[0128] The hierarchical fusion processing includes the method: first, combining the identification evidence of non-collaborative sensor with the identification evidence based on target position information, and performing target attribute judgment, and then combining the target attribute judgment result with the identification evidence of encrypted collaborative sensor.

[0129] The pointing consistency judgment method between DS identification evidences includes:

[0130] Two DS identification evidences E1 and E2 of the identification framework Θ, the corresponding basic trust allocation functions are m1 and m2, and the supported attribute pointing is A i and B j , then whether the pointing between the two DS identification evidences is consistent is determined by the following formula:

[0131]

[0132] If the intersection of the subset A and the subset B is empty, the pointing of the two DS identification evidences is inconsistent; otherwise, the pointing of the two DS identification evidences is consistent, wherein m(*) represents the evidence quality of the subset (*).

[0133] In the case of pointing inconsistency, the pointing consistency G c (E1, E2) is measured by G

[0134]

[0135] wherein Con(E1, E2) describes the conflict amount between the evidences, and H(E1, E2) describes the consistency amount between the identification evidences. The pointing consistency G c (E1, E2) is described by the proportion of the conflict belief part between the two identification evidences.

[0136] Assume that σ is the conflict intensity judgment threshold, when G c (E1, E2) > σ indicates that the pointing consistency is not up to standard, when the conflict intensity G c (E1, E2) ≤ σ indicates that the pointing consistency is up to standard.

[0137] In the case of pointing consistency up to standard, the two- two identification evidence combination processing is carried out according to the processing criterion of DS evidence theory. Assuming that the quality of the N-1th evidence is m1, and the quality of the Nth evidence is m2, then the combination calculation is carried out in the following formula manner:

[0138]

[0139] Before all the DS identification evidences are combined, the processing result of the above two- two combination does not carry out the attribute pointing judgment based on the highest confidence component, assuming that the highest confidence result of the identification evidence combination is D1, the pointing is Z1, the second highest confidence result is D2, and the pointing is Z2, then the fusion identification evidence participating in the next round of two- two combination is:

[0140]

[0141] Wherein, wherein, represents the representation of the last round of fusion evidence m(X) and m(Y) according to the DS evidence combination calculation method.

[0142] The processing flow is as shown in Figure 3 .

[0143] In the case of pointing consistency not up to standard, first, the identification evidence of the non- cooperative sensor is combined with the identification evidence based on the target position information, and the combination calculation method is consistent with the above combination method, but the judgment processing is carried out, that is, according to the above assumption, σ1 and σ2 are given as two judgment thresholds, and the judgment criterion is: if D1 > σ1 and |D1-D2| > σ2, then the identification result is the highest confidence result; Otherwise, identify as unknown. The obtained target attribute judgment result is combined with the encrypted cooperative identification evidence, and the combination method is consistent with the above combination method, and the processing flow is as shown in Figure 3 .

[0144] The method for combining historical identification evidences comprises:

[0145] S1, first, judge whether the DS identification evidences are independent, if yes, then according to the compatibility and exclusivity of the DS identification evidences, change the attribute pointing of the DS identification evidences, so that the DS identification evidences conflict, and then enter step S3; Otherwise, judge whether the joint of the current DS identification evidence and the cache evidence is independent, if yes, then enter S3; Otherwise, enter S2;

[0146] S2, cache the current non-independent evidence and change the attribute of the evidence identified by the DS to unknown, and then go to S3;

[0147] S3, combine the current decision result with the historical identification result.

[0148] Figure 4 The step 401 shown is to judge the independence of the current identification evidence, if it is independent, step 403 is executed, if it is not independent, step 402 is executed. Step 402 is to further combine the cached target historical identification evidence with the current identification evidence to jointly judge the independence of the evidence, if it is independent, step 403 is executed, if it is not independent, step 404 is executed. Step 404 is to cache the current non-independent identification evidence for subsequent fusion processing. Step 405 is to correct the identity attribute of the current non-independent SD identification evidence to unknown. Step 403 is to judge the compatibility and exclusivity between the independent identification evidence and the historical identification evidence, if they are compatible, the attribute of the current evidence needs to be changed so that they do not conflict with each other; if they are exclusive, the attribute of the current evidence needs to be changed so that they conflict with each other. Step 406 is to combine the DS between the current evidence and the historical identification evidence, the processing method is the same as that of step 201 of the DS evidence theory. Figure 2 Step 407 is to make a recognition decision based on the DS combination result to obtain the final decision result. After the recognition decision is made, one multi-source information fusion processing is completed.

[0149] The independence of the evidence refers to that the identity attribute of the target can be determined by only relying on a certain identification evidence, and on the contrary, certain evidence needs to be combined with other identification evidence to form a decision conclusion, and a single evidence cannot form an effective conclusion for identification. The compatibility and exclusivity of the evidence refers to the conflict between multiple identification evidences in the fusion processing. Two identification evidences are not in conflict, which is called compatible, and in conflict, which is called exclusive.

[0150] The independence and compatibility of the evidence can be referred to Table 1.

[0151] Table 1 Independence and compatibility of identification evidence

[0152]

[0153] Embodiment 2

[0154] The embodiment provides an information fusion system based on the DS evidence theory, which is used to realize the information fusion method based on the DS evidence theory in the above embodiment, and includes:

[0155] The collection module is configured to acquire DS identification evidence, wherein the DS identification evidence comprises identification evidence of an encryption cooperative sensor, identification evidence of a non-cooperative sensor, and identification evidence based on target position information.

[0156] The distribution screening module is configured to perform credibility redistribution and screening on the DS identification evidence, calculate consistency of the DS identification evidence from a time sequence, perform parallel fusion processing on the DS identification evidence that meets the consistency standard, and perform hierarchical fusion processing on the DS identification evidence that does not meet the consistency standard.

[0157] The fusion module is configured to perform fusion based on historical identification results of the DS identification evidence that are logically compatible and mutually exclusive.

[0158] The above detailed description further describes the purpose, technical solutions, and beneficial effects of the present application. It should be understood that the above description is only a specific implementation of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An information fusion method based on DS evidence theory, characterized in that, include: Obtain DS identification evidence and perform uncertainty measurement. The DS identification evidence includes: identification evidence of encrypted cooperative sensors, identification evidence of non-cooperative sensors, and identification evidence based on target location information. The method for obtaining the identification evidence based on target location information includes: Fuzzy membership functions for the target's yaw, heading, and altitude are established to describe the degree of legitimacy of the target's position; and the target attribute indices that match the flight plan are given as identification evidence based on the target's position information through a legitimacy threshold. The uncertainty of the target attribute is calculated based on the fuzzy membership degree. The method for obtaining the identification evidence based on target location information specifically includes: Let d be the yaw distance, and the yaw membership degree. for: ； in, The yaw distance defined for the navigation plan. This is the maximum yaw distance; Heading membership for: ； in, Current course to the target The prescribed course for the voyage plan; High membership for: ； in, For the target height, To limit the altitude for the flight plan, The maximum altitude deviation specified for the flight plan; The average membership degree, reflecting the current legitimacy of the target, is calculated by comprehensively considering yaw membership degree, heading membership degree, and altitude membership degree. Then, based on the average of membership degrees from multiple time series, the legitimacy evaluation of the target's current navigation is obtained. Finally, a threshold decision is made to form an attribute orientation judgment that supports the behavioral situation layer information. ； in, Let be the membership degree of the target position at time t relative to the conformity of the flight plan. For N time points Average membership degree For legality threshold; When the average membership degree Exceeding the legality threshold At that time, the identification evidence for the target location information is determined to be either {I} or {civilian} based on the target attributes of the current navigation plan; When the average membership degree Not exceeding the legality threshold Then the identification evidence for the target location information points to {suspicious enemy}; Reliability redistribution and screening of DS identification evidence: The identity attributes of DS identification evidence are multiplicatively weighted based on the recognition stability of the sensor; and in the case of conflict of DS identification evidence, the conflict of the multiplicatively weighted identity attributes is processed and the corresponding fusion method is dynamically selected to complete the fusion of conflict information. Based on the compatibility and mutual exclusion of DS identification evidence, the DS identification evidence is combined with historical identification evidence, and finally the identity attribute identification judgment is output.

2. The information fusion method based on DS evidence theory according to claim 1, characterized in that, The method for obtaining the identification evidence from the encrypted collaborative sensor includes: Based on whether the encrypted collaborative sensor obtains a valid recognition result, generate recognition evidence for the encrypted collaborative sensor; The recognition system is based on target information Make a judgment or classification, and the possible set of results for the judgment or classification is as follows: ;by To ascertain the target's true identity; The uncertainty measure of identification evidence for encrypted collaborative sensors is: the target information input into the identification system. The verdict was as follows Confidence expression ; ； in, Let M be the set of the actual identity attributes of the target. The target identity attribute for the judgment; Indicates the attribute is The target was judged as an attribute The conditional probability; represents the probability of a correct identification decision; k represents the target identity option to be judged; and E(*) represents the statistical average.

3. The information fusion method based on DS evidence theory according to claim 1, characterized in that, The identification evidence from the non-cooperative sensor is obtained through the following methods: Based on the basic information of the non-cooperative sensor identified by the target radiation source, including the target type and model, a mapping relationship is constructed to obtain evidence of the non-cooperative sensor identification. The uncertainty measure of identification evidence for non-cooperative sensors is the target identification confidence report of the sensor.

4. The information fusion method based on DS evidence theory according to claim 1, characterized in that, The reliability reassignment methods for DS-identified evidence include: The consistency of the evidence identified by DS is calculated from the time series, and the consistency assessment results are then used. As a multiplicative weighting parameter, it adjusts the uncertainty quantification result of DS identification evidence. ; The target recognition confidence report obtained by the sensor over several recognition cycles is as follows: If the number of elements in the recognition frame is S, then the sensor consistency over N recognition cycles... for: ； in, To ensure consistency in sensor identification between two adjacent cycles, Intermediate results for sensor consistency calculation. Based on The final result after normalization calculation.

5. The information fusion method based on DS evidence theory according to claim 1, characterized in that, The methods for screening evidence identified by DS include: Assess the consistency of the points of reference among DS identification evidence; For DS identification evidence with consistent orientation, a consistency measurement is performed: DS identification evidence that meets the consistency criteria is fused in parallel; DS identification evidence that does not meet the consistency criteria is fused in a layered manner. The layered fusion processing includes the following method: first, combining the identification evidence from non-cooperative sensors with the identification evidence based on target location information, and then performing target attribute judgment processing; and finally, combining the target attribute judgment result with the identification evidence from encrypted cooperative sensors.

6. The information fusion method based on DS evidence theory according to claim 5, characterized in that, Methods for determining the consistency of evidence identified by DS include: Recognition Framework Two DS identification evidence and The corresponding basic trust assignment function is and Supported attribute pointers are and Then, the following formula determines whether the two DS identification pieces of evidence point to the same conclusion: ； If the intersection of subset A and subset B is empty, then the two DS identification evidences point to different things; otherwise, the two DS identification evidences point to the same thing, where m(*) represents the quality of evidence of subset (*).

7. The information fusion method based on DS evidence theory according to claim 1, characterized in that, Methods for combining historical identification evidence include: S1. First, determine whether the DS identification evidence is independent. If so, change the attribute pointing of the DS identification evidence according to the compatibility and mutual exclusion of the DS identification evidence to make the DS identification evidence conflict, and then proceed to step S3. Otherwise, determine whether the combination of the current DS identification evidence and the cached evidence is independent. If so, proceed to S3; otherwise, proceed to S2. S2, cache the current non-independent evidence, change the attribute of the DS-identified evidence to unknown, and then proceed to S3; S3 combines the current judgment result with the historical identification result.

8. An information fusion system based on DS evidence theory, characterized in that, The information fusion method based on DS evidence theory as described in any one of claims 1-7 includes: The acquisition module is used to acquire DS identification evidence, which includes: identification evidence of encrypted cooperative sensors, identification evidence of non-cooperative sensors, and identification evidence based on target location information. The allocation and screening module is used for reliability redistribution and screening of DS identification evidence: it calculates the consistency of the DS identification evidence in terms of time series, merges the DS identification evidence that meets the consistency standard in parallel, and merges the DS identification evidence that does not meet the consistency standard in a hierarchical manner. The fusion module is used to fuse historical identification results based on the logical compatibility and mutual exclusion of evidence identified by DS.

Images