CN104021392B - A kind of conflicting evidence fusion method based on vector metric - Google Patents
A kind of conflicting evidence fusion method based on vector metric Download PDFInfo
- Publication number
- CN104021392B CN104021392B CN201410039585.8A CN201410039585A CN104021392B CN 104021392 B CN104021392 B CN 104021392B CN 201410039585 A CN201410039585 A CN 201410039585A CN 104021392 B CN104021392 B CN 104021392B
- Authority
- CN
- China
- Prior art keywords
- evidence
- vector
- factor
- burnt
- fusion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention discloses a kind of conflicting evidence fusion method based on vector metric, the measurement to conflict spectrum between evidence in Multi-source Information Fusion can be realized, the weight factor for determining fusion evidence is weighed by the conflict spectrum between evidence, and fusion evidence is modified, it can obtain the reliable target identification result of decision using the fusion of Dempster rules of combination.The present invention program is compared with traditional algorithm, from the angle of evidence vector, consider the conflict spectrum between evidence vector diversity factor and similarity common metrics evidence, the weight factor of fusion evidence is determined by the evidences conflict degree factor, and revised evidence is merged one by one using Dempster rules of combination after being modified to evidence, the rational result of decision is obtained, can be very good to be applied in field of target recognition, there is important theory significance and application value.
Description
Technical field
The present invention relates to a kind of multi-sources Information Fusion Method, more particularly to a kind of conflicting evidence fusion based on vector metric
Method.
Background technology
The military information abundant with civil area acquisition that develops into of sensor technology provides hardware supported, due to sensing
Device is influenceed in complex environment by factors such as external interference or artificial origins, and there is not for its identification target information exported
Certainty and ambiguity, in some instances it may even be possible to it is contradiction, if can not effectively handle high conflicting evidence fusion, practical application
System can not just carry out effective decision-making, greatly affected the decision-making performance of emerging system, and traditional algorithm is often tightly considered
The influence of evidence vector diversity factor, fail to consider other decision factors, so judging that there is certain error to target identification.
The content of the invention
, can be effectively to identification it is an object of the invention to provide a kind of conflicting evidence fusion method based on vector metric
Target makes correct decisions.
The present invention uses following technical proposals:
A kind of conflicting evidence fusion method based on vector metric, including following steps:
A, assigned by obtaining the confidence of the burnt member of the corresponding evidence of multiple sensor measurement informations, each evidence is seen
Make a vector, the vector m of i-th of evidencei=(mi(θ1),…,mi(θr),…,mi(θk))TRepresent, wherein i=1,
2 ..., n, n are the sum of evidence vector, and k is burnt first number in framework of identification Θ;
B, to above-mentioned each evidence vector miDiversity factor, Similarity Measure are carried out respectively:And record any i-th of evidence to
Measure miWith j-th of evidence vector mjBetween the diversity factor factor be Dif (mi-mj), any i-th of evidence vector miWith j-th of card
According to vectorial mjBetween the similarity factor be S (mi,mj), wherein j=1,2 ..., n;
C, by any evidence vector miAnd mjBetween diversity factor factor D if (mi-mj) and similarity factor S (mi,mj) logical
Cross formula:Calculate any evidence vector miAnd mjBetween conflict journey
Spend factor conf (mi,mj);
D, by any i-th of evidence vector miWith j-th of evidence vector mjBetween conflict spectrum factor conf (mi,mj)
Pass through formula: Try to achieve i-th of card
According to total conflict spectrum factor conf (m with other n-1 evidencei) and i-th evidence and other n-1 evidences relative support
Degree factor t ruf (mi), and utilize relative degree of support factor truf maximum in n evidencemaxWith i evidence and other
Total conflict spectrum factor conf (m of n-1 evidencei) obtain weight factor ωi;
E, burnt first θ in i-th of evidencerConfidence appointment mi(θr) represent, wherein r=1,2 ..., k, revised i-th
Burnt first θ in individual evidencerConfidence assigned value useRepresent, according to weight factor ωiPass through formula:
The evidence of fusion is modified:Then Dempster groups are used
Normally revised evidence is merged one by one, burnt member corresponding to the confidence appointment m (A) of the A after fusion maximum is
Identification target, as decision-making final result corresponding to the result of decision of target identification.
Diversity factor, which calculates, in described step B passes through following formulaObtain any i-th of evidence
Vectorial miWith j-th of evidence vector mjBetween diversity factor factor D if (mi-mj), < m in formulai,mj> represents two evidence vectors
miAnd mjInner product, the difference and the vector σ of positive definite coefficient matrix D compositions that corresponding burnt first confidence is assigned between two evidences
(mi-mj) represent, vectorial σ (mi-mj) average and variance use E (m respectivelyi-mj) and D (mi-mj) represent.
Similarity is drawn by calculating the included angle cosine value between vector in described step B, especially by following
FormulaIt is calculated.
Conflicting evidence fusion method according to claim 2 based on vector metric, it is characterised in that:It is described two
The difference and the vectorial σ (m of positive definite coefficient matrix D compositions that corresponding burnt first confidence is assigned between evidencei-mj)=(mi-mj)TD, its
Middle mi-mj=(mi(θ1)-mj(θ1),…,mi(θr)-mj(θl),…,mi(θk)-mj(θk))T, r, l=1,2 ..., k.
Described Dempster rules of combination are:
Wherein, m (A) represents that burnt first A confidence is assigned, r, l=1,2 ..., k,For empty set.
The present invention by multisensor measure based on target identification for application background, the information that sensor is provided converts
For evidence, and regard evidence as vector, the conflict spectrum between evidence is weighed from the angle research of evidence vector, so that right
The evidence of fusion is modified.The present invention program has considered the diversity factor and similar of evidence vector compared with traditional algorithm
Conflict spectrum between the parameter measure evidences such as degree, the weight factor of fusion evidence is determined on this basis, and to merging evidence
Be modified, finally using Dempster rules of combination to revised evidence carry out one by one fusion make it is last to target identification
Decision-making, there is important theory significance and application value.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention.
Embodiment
As shown in figure 1, a kind of conflicting evidence fusion method based on vector metric, including following steps:
A, assigned by obtaining the confidence of the burnt member of the corresponding evidence of multiple sensor measurement informations, each evidence is seen
Make a vector, the vector m of i-th of evidenceiRepresent, wherein i=1,2 ..., n, n is the sum of evidence vector, and k is
Burnt first number in framework of identification Θ;It is first that the different sensor of the multiple properties got is corresponding to the identification information of target
Multiple evidences are converted into, and regard the evidence of each fusion (process of i.e. corresponding conversion) as a vector.Assuming that obtain n
Evidence is respectively m1,m2,…,mn, it is assumed that the burnt member in framework of identification Θ is θ1,θ2,…,θk, burnt member corresponding to i-th of evidence puts
It is respectively m that letter, which is assigned,i(θ1),mi(θ2) ..., mi(θk), regard evidence as vector, then element corresponding to i evidence vector is successively
For:mi(θ1),mi(θ2),…,mi(θk)。
During acquisition multisensor observation information of the present invention, also had according to the different modes obtained of actual conditions
Institute is different, such as:Imaging sensor observes shooting to the aerial target in the range of monitoring, according to ATL and extraction image border
Not bending moment as target signature, calculate the Euclidean distance between the moment characteristics stored in realtime graphic moment characteristics and ATL,
Choose with distance value minimum in each class target several posture figures as the distance of figure and the target in real time, reconstruct mapping and ask
The similarity of real-time target and framework of identification target is taken, finally brief inference is normalized place using the definition of mass functions
Reason obtains the evidence according to figure construction in real time, is then converted into evident information and carries out vector representation, further by evident information
The confidence assigned value of middle burnt member is indicated as vector element.
B, to above-mentioned each evidence vector miAnd mjDiversity factor, Similarity Measure are carried out respectively:And record any i-th of card
According to vectorial miWith j-th of evidence vector mjBetween the diversity factor factor be Dif (mi-mj), any i-th of evidence vector miAnd jth
Individual evidence vector mjBetween the similarity factor be S (mi,mj), wherein j=1,2 ..., n;
The diversity factor of evidence vector is calculated by corresponding to difference, the positive definite coefficient square that burnt first confidence is assigned between two evidences
Relation obtains the vectorial difference of evidence between confidence assigns the number being not zero in battle array composition of vector, two evidence vector elements
Degree.Evidence vector diversity factor embodies the otherness between two evidences, and evidence vector diversity factor is bigger, the conflict between evidence
Degree is bigger.Diversity factor, which calculates, in specific step B of the present invention passes through following formulaObtain any i-th of evidence
Vectorial miWith j-th of evidence vector mjBetween diversity factor factor D if (mi-mj), < m in formulai,mi> represents two evidence vectors
miAnd miInner product, < mj,mj> represents two evidence vector mjAnd mjInner product, corresponding burnt first confidence is assigned between two evidences
Difference and positive definite coefficient matrix D composition vector σ (mi-mj) represent, vectorial σ (mi-mj) average and variance use E respectively
(mi-mj) and D (mi-mj) represent.
Similarity is drawn by calculating the included angle cosine value between vector in described step B, especially by following
FormulaAny i-th of evidence vector m is calculatediWith j-th of evidence vector mj
Between similarity factor S (mi-mj), the difference and positive definite coefficient matrix D that corresponding burnt first confidence is assigned between described two evidences
Vectorial σ (the m of compositioni-mj)=(mi-mj)TD, wherein mi-mj=(mi(θ1)-mj(θ1),…,mi(θr)-mj(θl),…,mi(θk)-
mj(θk))T, the element D (θ in matrix Dr,θl) represent, then Wherein r, l=1,2 ..., k.
C, by any vectorial mi(θk) and mj(θh) between diversity factor factor D if (mi-mj) and similarity factor S (mi,mj)
Pass through formula:Calculate any vectorial miAnd mjBetween conflict spectrum
Factor conf (mi,mj);
D, by any i-th of evidence vector miWith j-th of evidence vector mjBetween conflict spectrum factor conf (mi,mj)
Total conflict spectrum factor conf (m of i-th of evidence and other n-1 evidences are tried to achieve by following three formulai) and i-th of card
According to the relative degree of support factor t ruf (m with other n-1 evidencesi) (if the conflict spectrum factor between two evidences is bigger,
Illustrate that the conflict spectrum between two evidences is bigger), and utilize relative degree of support factor truf maximum in n evidencemax
With i evidence and total conflict spectrum factor conf (m of other n-1 evidencei) by following formula 4. obtain weight factor ωi;It is false
If evidence mi,mjConflict spectrum factor conf (m between (i, j=1,2 ..., n)i,mj), i-th of evidence and other n-1 card
According to total conflict spectrum factor conf (mi) represent, the relative degree of support factor of i-th of evidence and other n-1 evidence is used
truf(mi) represent, maximum relative degree of support factor truf in n evidencemaxRepresent, the weight factor of i-th of evidence
Use ωiRepresent.Specifically formula is:
①
②
trufmax=max (truf (m1),…,truf(ml),…,truf(mn)) ③
④
E, burnt first θ in i-th of evidencerConfidence appointment mi(θr) represent, burnt first θ in revised i-th of evidencer's
Confidence assigned value is usedRepresent, according to weight factor ωiPass through formula:
The evidence of fusion is modified:Then Dempster groups are used
Normally merged, described Dempster rules of combination are:
Wherein, m (A) represents that burnt first A confidence is assigned, r, l=1,2 ..., k,For empty set.Burnt first A's after fusion puts
Letter assigns burnt member corresponding to m (A) maximum most to be terminated for identification target, as decision-making corresponding to the result of decision of target identification
Fruit.
The present invention program, from the angle of evidence vector, considers the vectorial difference of evidence compared with traditional algorithm
Spend similarity common metrics evidence between conflict spectrum, by the evidences conflict degree factor determine fusion evidence weight because
Son, and revised evidence is merged one by one using Dempster rules of combination after being modified to evidence, it is reasonable to obtain
The result of decision, can be very good be applied to field of target recognition in, there is important theory significance and application value.
Claims (5)
- A kind of 1. conflicting evidence fusion method based on vector metric, it is characterised in that:Including following steps:A, assigned by obtaining the confidence of the burnt member of the corresponding evidence of multiple sensor measurement informations, regard each evidence as one Individual vector, the vector m of i-th of evidencei=(mi(θ1),…,mi(θr),…,mi(θk))TRepresent, wherein i=1,2 ..., n, n For the sum of evidence vector, k is burnt first number in framework of identification Θ;B, to above-mentioned each evidence vector miDiversity factor, Similarity Measure are carried out respectively:And record any i-th of evidence vector mi With j-th of evidence vector mjBetween the diversity factor factor be Dif (mi-mj), any i-th of evidence vector miWith j-th of evidence to Measure mjBetween the similarity factor be S (mi,mj), wherein j=1,2 ..., n;C, by any evidence vector miAnd mjBetween diversity factor factor D if (mi-mj) and similarity factor S (mi,mj) pass through public affairs Formula:Calculate any evidence vector miAnd mjBetween conflict spectrum because Sub- conf (mi,mj);D, by any i-th of evidence vector miWith j-th of evidence vector mjBetween conflict spectrum factor conf (mi,mj) pass through public affairs Formula: Try to achieve i-th of evidence and its Total conflict spectrum factor conf (m of his n-1 evidencei) and i-th evidence and other n-1 evidences relative degree of support because Sub- truf (mi), and utilize relative degree of support factor t ruf maximum in n evidencemaxWith i-th of evidence and other n-1 Total conflict spectrum factor conf (m of evidencei) obtain weight factor ωi;E, burnt first θ in i-th of evidencerConfidence appointment mi (θr) represent, wherein r=1,2 ..., k, burnt first θ in revised i-th of evidencerConfidence assigned value useRepresent, root According to weight factor ωiPass through formula:The evidence of fusion is modified, wherein A is burnt member;Then use Dempster rules of combination are merged one by one to revised evidence, and the confidence of the A after fusion assigns m (A) maximum pair Jiao's member answered is identification target, as decision-making final result corresponding to the result of decision of target identification.
- 2. the conflicting evidence fusion method according to claim 1 based on vector metric, it is characterised in that:In described step B Diversity factor, which calculates, passes through following formula Obtain any i-th of evidence vector miWith j-th of evidence vector mjBetween diversity factor factor D if (mi-mj), in formula<mi,mj> Represent two evidence vector miAnd mjInner product, the difference and positive definite coefficient matrix D that corresponding burnt first confidence is assigned between two evidences The vector of composition σ (mi-mj) represent, vectorial σ (mi-mj) average and variance use E (m respectivelyi-mj) and D (mi-mj) represent.
- 3. the conflicting evidence fusion method according to claim 2 based on vector metric, it is characterised in that:Described step Similarity is drawn by calculating the included angle cosine value between vector in B, especially by following formulaIt is calculated.
- 4. the conflicting evidence fusion method according to claim 2 based on vector metric, it is characterised in that:Described two cards The difference and the vectorial σ (m of positive definite coefficient matrix D compositions that corresponding burnt first confidence is assigned betweeni-mj)=(mi-mj)TD, wherein mi-mj=(mi(θ1)-mj(θ1),…,mi(θr)-mj(θl),…,mi(θk)-mj(θk))T, r, l=1,2 ..., k.
- 5. the conflicting evidence fusion method based on vector metric according to claim 1-4 any claims, its feature It is:Described Dempster rules of combination are:Wherein, m (A) represents that burnt first A confidence is assigned, r, l=1,2 ..., k,For empty set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410039585.8A CN104021392B (en) | 2014-01-27 | 2014-01-27 | A kind of conflicting evidence fusion method based on vector metric |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410039585.8A CN104021392B (en) | 2014-01-27 | 2014-01-27 | A kind of conflicting evidence fusion method based on vector metric |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104021392A CN104021392A (en) | 2014-09-03 |
CN104021392B true CN104021392B (en) | 2018-01-19 |
Family
ID=51438136
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410039585.8A Active CN104021392B (en) | 2014-01-27 | 2014-01-27 | A kind of conflicting evidence fusion method based on vector metric |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104021392B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105719064A (en) * | 2016-01-15 | 2016-06-29 | 深圳大学 | Method and system for evaluating confliction degree between targets |
CN106022366B (en) * | 2016-07-04 | 2019-03-08 | 杭州电子科技大学 | A kind of rotating machinery method for diagnosing faults based on neighbour's evidence fusion |
CN106778847B (en) * | 2016-12-02 | 2019-11-19 | 河南大学 | The fusion method of evidences conflict is measured based on logarithmic function |
CN109766933A (en) * | 2018-12-26 | 2019-05-17 | 中国电子科技集团公司第二十研究所 | A kind of multisource data fusion recognition methods based on evidence fuzzy factor |
CN109977763B (en) * | 2019-02-03 | 2022-10-04 | 河南科技大学 | Aerial small target identification method based on improved evidence trust |
CN110033028B (en) * | 2019-03-19 | 2022-09-16 | 河南大学 | Conflict evidence fusion method based on arithmetic mean closeness |
CN110348504B (en) * | 2019-07-02 | 2021-03-23 | 北京理工大学 | Gas pipeline network leakage grade judgment method based on improved evidence fusion algorithm |
CN111625775B (en) * | 2020-05-28 | 2023-04-07 | 河南大学 | Hellinger distance and reliability entropy based weighted conflict evidence fusion method |
CN113449412B (en) * | 2021-05-24 | 2022-07-22 | 河南大学 | Fault diagnosis method based on K-means clustering and comprehensive correlation |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1645358A (en) * | 2005-01-12 | 2005-07-27 | 河海大学 | Evidence theory information blending decision method based on state vector distance |
CN101556651A (en) * | 2009-04-15 | 2009-10-14 | 北京航空航天大学 | Multi-source data fusion method in clustering wireless sensor network |
CN101996157A (en) * | 2010-10-23 | 2011-03-30 | 山东科技大学 | Multisource information fusion method in evidence high-conflict environment |
-
2014
- 2014-01-27 CN CN201410039585.8A patent/CN104021392B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1645358A (en) * | 2005-01-12 | 2005-07-27 | 河海大学 | Evidence theory information blending decision method based on state vector distance |
CN101556651A (en) * | 2009-04-15 | 2009-10-14 | 北京航空航天大学 | Multi-source data fusion method in clustering wireless sensor network |
CN101996157A (en) * | 2010-10-23 | 2011-03-30 | 山东科技大学 | Multisource information fusion method in evidence high-conflict environment |
Also Published As
Publication number | Publication date |
---|---|
CN104021392A (en) | 2014-09-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104021392B (en) | A kind of conflicting evidence fusion method based on vector metric | |
CN106650785B (en) | Weighted evidence fusion method based on the classification of evidence and measure method for conflict | |
CN110033028B (en) | Conflict evidence fusion method based on arithmetic mean closeness | |
CN105046067B (en) | Multiple sensor information amalgamation method based on evidence similarity | |
CN109446986B (en) | Effective feature extraction and tree species identification method for tree laser point cloud | |
CN105354841B (en) | A kind of rapid remote sensing image matching method and system | |
CN113326735B (en) | YOLOv 5-based multi-mode small target detection method | |
CN106778847A (en) | The fusion method of evidences conflict is weighed based on logarithmic function | |
CN111504509A (en) | Temperature measurement method based on multilayer neural network | |
CN112949468A (en) | Face recognition method and device, computer equipment and storage medium | |
CN115544714B (en) | Time sequence dynamic countermeasure threat assessment method based on aircraft formation | |
CN112232375B (en) | Unknown type target identification method based on evidence theory | |
CN113987789A (en) | Dynamic threat assessment method in unmanned aerial vehicle collaborative air combat | |
CN112561878A (en) | Finger vein image quality evaluation method based on weighted fusion | |
CN111539422A (en) | Flight target cooperative identification method based on fast RCNN | |
CN109034239A (en) | Remote sensing image classification method, and site selection method and device for distributed wind power plant | |
CN109669849B (en) | Complex system health state assessment method based on uncertain depth theory | |
CN113295421B (en) | Engine fault diagnosis method based on improved conflict coefficient and reliability entropy | |
CN103226698B (en) | A kind of method for detecting human face | |
CN114187464A (en) | Multi-cycle target identification method based on laser radar and vision fusion in complex environment | |
CN110008985A (en) | Based on the shipboard aircraft group target identification method for improving D-S evidence theory rule | |
CN112802011A (en) | Fan blade defect detection method based on VGG-BLS | |
CN106980845A (en) | The crucial independent positioning method of face based on structured modeling | |
CN103425979B (en) | Hand authentication method | |
CN116739739A (en) | Loan amount evaluation method and device, electronic equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |