CN107967449A - A kind of multispectral image unknown object recognition methods based on broad sense evidence theory - Google Patents
A kind of multispectral image unknown object recognition methods based on broad sense evidence theory Download PDFInfo
- Publication number
- CN107967449A CN107967449A CN201711135254.4A CN201711135254A CN107967449A CN 107967449 A CN107967449 A CN 107967449A CN 201711135254 A CN201711135254 A CN 201711135254A CN 107967449 A CN107967449 A CN 107967449A
- Authority
- CN
- China
- Prior art keywords
- msub
- wave band
- sky
- maximum
- cloud
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/257—Belief theory, e.g. Dempster-Shafer
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention is based on multispectral image, there is provided a kind of to realize that unknown object knows method for distinguishing under the sky background of face, is related to target identification, image processing field with broad sense evidence theory.The present invention establishes triangle fuzzy model to target, cloud, sky, will classify after model broadening to pixel, determines whether unknown object according to pixel classification results, and update current Triangular Fuzzy Number model with pixel classification results.The present invention classifies pixel using broad sense evidence theory, this sorting technique has preferably merged the image information of different-waveband, can realize the differentiation of unknown object, has the advantages that calculating is simple, real-time is good;Random disturbances method for removing proposed by the present invention, eliminates erroneous judgement caused by random disturbances well.
Description
Technical field
The present invention relates to target identification, image processing field, is that one kind is realized under the sky background of face based on broad sense evidence theory
Multispectral image unknown object knows method for distinguishing.
Background technology
Under battlefield surroundings, the main object of target identification is the Aircraft Targets for rushing and invading by mistake China territorial sky, comprising non-
Cooperative target and hostile target.From the point of view of machine learning, when tupe identifies problem, researcher is needed to being possible to
The target classification of appearance establishes masterplate database, and chooses wherein all target class and be very originally trained.However, non-cooperative target
Mark and hostile target can not establish its full database.When Aircraft Targets are the unknown objects outside database, that is, it is non-
It is imperfect due to its database when cooperative target or hostile target, easily by the target be mistaken for database other known to
Target.By mistake by enemy's aircraft be determined as our aircraft or our aircraft is mistaken for enemy's aircraft all can be after bringing on a disaster property
Fruit, therefore studying unknown object recognition methods has important military value.
Information fusion technology is that collaboration utilizes multi-source information, to obtain more objective to things or target, more essential understanding
Informix treatment technology, is one of key technology of intelligence science research.In many information fusion models and method, extensively
Adopted evidence theory (Generalized Evidence Theory, GET) algorithm is one of maximally efficient algorithm.Broad sense evidence
Theoretical starting point is an open world based on the environment residing for emerging system, uses empty setTo model open world, give upLimitation.In broad sense evidence theoryIt is not traditional empty set, it is outside framework of identification that it is corresponding
Proposition.Broad sense evidence theory proposes broad sense Basic probability assignment function (Generalized Basic Probability
Assignment, GBPA), and propose the fusion that broad sense composite formula realizes two GBPA.When system is in closed world,
Exactly meetWhen, broad sense evidence theory just deteriorates to the evidence theory of classics, from this view point, broad sense evidence reason
By being that classical evidence theory is simple and intuitively promote.
There has been no the research that unknown object identification is realized based on multispectral image at present.Broad sense evidence theory has many excellent
Point, applied has important military value in the identification of multispectral image unknown object.
The content of the invention
In order to realize that unknown object identifies, the present invention is based on multispectral image, there is provided one kind is realized with broad sense evidence theory
Unknown object knows method for distinguishing under the sky background of face, and the identification of hostile unknown object is realized while current goal is tracked.Use
The unknown object identification that this method is realized has important military value.
The technical solution adopted by the present invention to solve the technical problems includes the following steps:
Step 1:Input each wave band cloud (C) under a pattern sky multispectral image and current environment, sky (S) and current
Minimum gray value, intermediate value and the maximum of target (T) imaging, according to the gray scale of the cloud, sky and current goal of input imaging most
Small value, intermediate value, maximum, establish corresponding Triangular Fuzzy Number model, and the current goal is the tracking target of current interest
(such as Defensive Target), framework of identification areC represents cloud in framework of identification, and S represents sky, and T represents current mesh
Mark,Represent unknown object, C, S, the method for T Triangular Fuzzy Number model foundations is:
The minimum gray value Cmin that wave band i clouds are imagedi, intermediate value CaveiAnd maximum CmaxiRespectively as wave band i clouds three
The minimum value of angle Fuzzy Math Model, intermediate value, maximum, then the Triangular Fuzzy Number of wave band i clouds be
The minimum gray value Smin that wave band i skies are imagedi, intermediate value SaveiAnd maximum SmaxiRespectively as wave band i sky Triangle Modules
The minimum value of paste exponential model, intermediate value, maximum, the then Triangular Fuzzy Number that wave band i skies are established are
The minimum gray value Tmin that wave band i current goals are imagedi, intermediate value TaveiAnd maximum of T maxiRespectively as the current mesh of wave band i
The minimum value of mark Triangular Fuzzy Number model, intermediate value, maximum, the then Triangular Fuzzy Number that wave band i current goals are established are
Step 2:The fuzzy number broadening that step 1 is obtained, i.e., reduce the upper bound increase of fuzzy number, lower bound, remember respectively
Fuzzy number after cloud, sky, current goal broadening isThe method for widening is:
Step 3:Classify to any pixel point p in input picture, classification results may be cloud, sky, current mesh
Mark, unknown object:
1) to any classified pixels point p, its wave band i gray values G is takeni, use GiWithIt is basic to generate broad sense
Probability distribution function mi, it is describedRespectively foregoing wave band i clouds, sky, the Triangular Fuzzy Number of current goal,
The broad sense Basic probability assignment function is defined as the subset A, m (A) for belonging to Θ to any one in broad sense evidence theory
∈ [0,1], and meetThen m is 2ΘOn broad sense Basic probability assignment function, wherein 2ΘFor the power of framework of identification
Collection,The broad sense Basic probability assignment function mi
Generation method is:By GiWith fuzzy numberThe high point of intersection point is assigned to the reliability of corresponding single subset elements, by GiWith
Fuzzy numberThe secondary high point of intersection point is assigned to the reliability of corresponding double subset elements, by GiWith fuzzy numberThe low spot of intersection point is assigned to the reliability of corresponding more subset elements, wherein single subset elements refer to walking
The subset { C } of framework of identification Θ, { S } or { T }, double subset elements refer to framework of identification Θ in step 1 in rapid one
Subset { C, S }, { C, T } or { S, T }, more subset elements refer to the subset { C, S, T } of framework of identification Θ in step 1;Note
The sum of reliability of generation is Sum, if Sum is not less than 1, the reliability of generation is normalized to obtain mi;If Sum is less than 1, by 1-
Sum is assigned to empty setReliability;
2) by the broad sense Basic probability assignment function m of 25 wave band generationsiMerge to obtain m using average fusion method, institute
Stating average fusion method is:Whereinmi(i=1,2 ..., 25) it is to be generated in step 2
Broad sense Basic probability assignment function;
3) m after fusion is converted into probability point using Pignistic probability transformation methods
Cloth P, the conversion method are: Wherein
4) classify according to obtained probability distribution P to pixel p, take P ({ C }), P ({ S }), P ({ T }),In most
The corresponding classification of big probability as pixel p classify as a result, C represents cloud in classification results, S represents sky, and T represents current
Target,Represent unknown object;
Step 4:ELIMINATION OF ITS INTERFERENCE is carried out to the region for being categorized as unknown object, to choose real movement unknown object, is done
The principle for disturbing exclusion is realized using the position correlation of consecutive frame zone of ignorance:
If knowing caused by random disturbances by mistake, then the change in location of consecutive frame zone of ignorance is larger, sets rational threshold value
Random disturbances can be excluded, threshold value can be chosen according to target speed and (such as elect 1.5 times of target single frames displacements as);If static
Target, after 5-10 frames its total displacement be able to can equally be excluded close to 0;
Step 5:Obscured according to the triangle of the pixel classification results of present frame renewal previous frame current goal, cloud, sky
Exponential model, it is described using the model after renewal as next two field picture cloud, sky, the other Fuzzy Math Model of three species of current goal
Model update method is:
For all pixels for being identified as cloud, the maximum max and minimum value min of these pixel wave bands i are obtained,
If max>Cmaxi, willThe upper bound is updated to max, if min<Cmini, willLower bound is updated to min;It is identified as day for all
Empty pixel, obtains the maximum max and minimum value min of these pixel wave bands i, if max>Smaxi, willThe upper bound updates
For max, if min<Smini, willLower bound is updated to min;For all pixels for being identified as current goal, these are obtained
The maximum max and minimum value min of pixel wave band i, if max>Tmaxi, willThe upper bound is updated to max, if min<Tmini, willLower bound is updated to min.
Classified the beneficial effects of the present invention are the present invention using broad sense evidence theory to pixel, can be according to pixel
Classification results select current goal and it are tracked, and realize the differentiation of unknown object, therefore sorting technique proposed by the present invention
The preferable image information for having merged different-waveband, has the advantages that calculating is simple, real-time is good;Triangle Module proposed by the present invention
Number modeling method is pasted, the problem of representation for solving fuzzy message well;Broad sense Basic probability assignment function proposed by the present invention
Generation method, realizes the processing to fuzzy message well;Random disturbances method for removing proposed by the present invention, excludes well
Erroneous judgement caused by random disturbances.
Brief description of the drawings
The general flow chart that Fig. 1 present invention realizes.
Fig. 2 is the other fuzzy number of 1 three species of wave band.
Fig. 3 is the fuzzy number after 1 three kinds of classification broadenings of wave band.
Embodiment
The present invention is further described with example below in conjunction with the accompanying drawings.
Step 1:Input each wave band cloud (C), sky under the face sky multispectral image and current environment of 25 wave band of a frame
(S) and current goal (T) imaging minimum gray value, intermediate value and maximum, according to the cloud, sky and current goal of input be imaged
Minimum gray value, intermediate value, maximum, establish corresponding Triangular Fuzzy Number model, the current goal is current interest
Target (such as Defensive Target) is tracked, framework of identification isC represents cloud in framework of identification, and S represents sky, T tables
Show current goal,Represent unknown object, C, S, the method for T Triangular Fuzzy Number model foundations is:
Minimum gray value 131, intermediate value 172.5 and the maximum 214 of 1 cloud of wave band imaging are inputted, respectively as 1 cloud three of wave band
The minimum value of angle Fuzzy Math Model, intermediate value, maximum, then the Triangular Fuzzy Number of 1 cloud of wave band beWill
Minimum gray value 91, intermediate value 140 and the maximum 189 of 1 sky of wave band imaging are respectively as 1 sky Triangular Fuzzy Number model of wave band
Minimum value, intermediate value, maximum, then 1 sky of wave band establish Triangular Fuzzy Number beWave band 1 is current
The minimum gray value 69 of target imaging, 109.5 maximum 150 of intermediate value are respectively as 1 current goal Triangular Fuzzy Number model of wave band
Minimum value, intermediate value, maximum, then 1 current goal of wave band establish Triangular Fuzzy Number be
Step 2:The fuzzy number broadening that step 1 is obtained, the fuzzy number after cloud, sky, current goal broadening are remembered respectively
ForThe method for widening is:
WithBroadening exemplified by, the result after broadening is:
Step 3:Classify to any pixel point p in input picture, classification results may be cloud, sky, current mesh
Mark, unknown object:
1) to any classified pixels point p, its wave band i gray values G is takeni, use GiWithIt is basic to generate broad sense
Probability distribution function mi, it is describedRespectively foregoing wave band i clouds, sky, the Triangular Fuzzy Number of current goal,
The broad sense Basic probability assignment function is defined as the subset A, m (A) for belonging to Θ to any one in broad sense evidence theory
∈ [0,1], and meetThen m is 2ΘOn broad sense Basic probability assignment function, wherein 2ΘFor the power of framework of identification
Collection,The broad sense Basic probability assignment function mi
Generation method is:By GiWith fuzzy numberThe high point of intersection point is assigned to the reliability of corresponding single subset elements, by GiWith
Fuzzy numberThe secondary high point of intersection point is assigned to the reliability of corresponding double subset elements, by GiWith fuzzy numberThe low spot of intersection point is assigned to the reliability of corresponding more subset elements, wherein single subset elements refer to walking
The subset { C } of framework of identification Θ, { S } or { T }, double subset elements refer to framework of identification Θ in step 1 in rapid one
Subset { C, S }, { C, T } or { S, T }, more subset elements refer to the subset { C, S, T } of framework of identification Θ in step 1;Note
The sum of reliability of generation is Sum, if Sum is not less than 1, the reliability of generation is normalized to obtain mi;If Sum is less than 1, by 1-
Sum is assigned to empty setReliability;
The pixel p of input is 71 in the gray value of wave band 1, then m1Generation method is as follows:
Such as Fig. 3, gray value 71 withThere is an intersection point 0.3636, as the reliability of current goal (T).Sum at this time
=T=0.3636, unknown object is distributed to by 1-Sum i.e. 0.6364Reliability.Obtain m1:
m1{ T }=0.3636,
With same method, m is obtained2~m25It is as follows:
m2{ T }=0.5417, m2{ S, T }=0.0471,
m3{ T }=0.3925,
m4{ T }=0.4,
m5{ T }=0.5495, m5{ S, T }=0.1059,
m6{ T }=0.3667,
m7{ T }=0.3524,
m8{ T }=0.3420,
m9{ T }=0.4609, m9{ S, T }=0.3833,
m10{ T }=0.3939, m10{ S, T }=0.0175,
m11{ T }=0.1739,
m12{ T }=0.4608, m12{ S, T }=0.0833,
m13{ T }=0.3770, m13{ S, T }=0.0804,
m14{ T }=0.3489, m14{ S, T }=0.1429,
m15{ T }=0.4186, m15{ S, T }=0.0545,
m16{ T }=0.3904, m16{ S, T }=0.0294,
m17{ T }=0.5552, m17{ C, T }=0.4448
m18{ T }=0.4775, m18{ S, T }=0.0706,
m19{ T }=0.3738,
m20{ T }=0.4587,
m21{ T }=0.5794,
m22{ T }=0.5536,
m23{ T }=0.4587,
m24{ T }=0.5287,
m25{ T }=0.4800,
2) by the broad sense Basic probability assignment function m of 25 wave band generationsiMerge to obtain m using average fusion method, institute
Stating average fusion method is:Whereinmi(i=1,2 ..., 25) it is what is generated in step 2
Broad sense Basic probability assignment function;
Fusion results are:M { T }=0.4319, m { S, T }=0.0584,
3) m after fusion is converted into probability point using Pignistic probability transformation methods
Cloth P, the conversion method are:
Wherein
The result that m is converted to probability distribution P is
P { T }=0.4611, P { S }=0.0292,
4) classify according to obtained probability distribution P to pixel p, take P ({ C }), P ({ S }), P ({ T }),In most
The corresponding classification of big probability as pixel p classify as a result, C represents cloud in classification results, S represents sky, and T represents current
Target,Represent unknown object;
According to probability distribution P, maximum probability classification isTherefore pixel p is classified as unknown object.
Step 4:ELIMINATION OF ITS INTERFERENCE is carried out to the region for being categorized as unknown object, to choose real movement unknown object, is done
The principle for disturbing exclusion is realized using the position correlation of consecutive frame zone of ignorance:
If knowing caused by random disturbances by mistake, then the change in location of consecutive frame zone of ignorance is larger, sets rational threshold value
Random disturbances can be excluded, threshold value can be chosen according to target speed and (such as elect 1.5 times of target single frames displacements as);If static
Target, after 5-10 frames its total displacement be able to can equally be excluded close to 0;
Step 5:Obscured according to the triangle of the pixel classification results of present frame renewal previous frame current goal, cloud, sky
Exponential model, it is described using the model after renewal as next two field picture cloud, sky, the other Fuzzy Math Model of three species of current goal
Model update method is:
For all pixels for being identified as cloud, the maximum max and minimum value min of these pixel wave bands i are obtained,
If max>Cmaxi, willThe upper bound is updated to max, if min<Cmini, willLower bound is updated to min;It is identified as day for all
Empty pixel, obtains the maximum max and minimum value min of these pixel wave bands i, if max>Smaxi, willThe upper bound updates
For max, if min<Smini, willLower bound is updated to min;For all pixels for being identified as current goal, these are obtained
The maximum max and minimum value min of pixel wave band i, if max>Tmaxi, willThe upper bound is updated to max, if min<Tmini, willLower bound is updated to min.
By taking 1 model modification of wave band as an example:For all pixels for being identified as cloud, these pixel wave bands 1 are obtained
Maximum is 214, minimum value 131,Upper bound lower bound does not update;For all pixels for being identified as sky, obtain
The maximum of these pixel wave bands 1 is 189, minimum value 87, therefore willLower bound is updated to 87,The upper bound does not update;For
All pixels for being identified as current goal, the maximum for obtaining these pixel wave bands 1 are 150, minimum value 65,65<
69, therefore willLower bound is updated to 69,The upper bound does not update.
Claims (1)
1. a kind of multispectral image unknown object method of discrimination based on broad sense evidence theory, it is characterised in that including following steps
Suddenly:
Step 1:Input each wave band cloud (C), sky (S) and current goal under a pattern sky multispectral image and current environment
(T) minimum gray value, intermediate value and the maximum of imaging, the minimum gray value being imaged according to the cloud, sky and current goal of input,
Intermediate value, maximum, establish corresponding Triangular Fuzzy Number model, and the current goal is (such as anti-for the tracking target of current interest
Imperial target), framework of identification isC represents cloud in framework of identification, and S represents sky, and T represents current goal,Table
Show unknown object, C, S, the method for T Triangular Fuzzy Number model foundations is:
The minimum gray value Cmin that wave band i clouds are imagedi, intermediate value CaveiAnd maximum CmaxiRespectively as wave band i cloud Triangle Modules
Paste exponential model minimum value, intermediate value, maximum, then the Triangular Fuzzy Number of wave band i clouds beWill
The minimum gray value Smin of wave band i skies imagingi, intermediate value SaveiAnd maximum SmaxiObscured respectively as wave band i sky triangles
The minimum value of exponential model, intermediate value, maximum, the then Triangular Fuzzy Number that wave band i skies are established are
The minimum gray value Tmin that wave band i current goals are imagedi, intermediate value TaveiAnd maximum of T maxiRespectively as the current mesh of wave band i
The minimum value of mark Triangular Fuzzy Number model, intermediate value, maximum, the then Triangular Fuzzy Number that wave band i current goals are established are
Step 2:The fuzzy number broadening that step 1 is obtained, i.e., reduce the upper bound increase of fuzzy number, lower bound, remember cloud, day respectively
Fuzzy number after empty, current goal broadening isThe method for widening is:
<mrow>
<mi>E</mi>
<msub>
<mover>
<mi>C</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>Cmin</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>10</mn>
<mo>,</mo>
<msub>
<mi>Cave</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>Cmax</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>10</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>E</mi>
<msub>
<mover>
<mi>S</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>Smin</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>10</mn>
<mo>,</mo>
<msub>
<mi>Save</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>Smax</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>10</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
<mrow>
<mi>E</mi>
<msub>
<mover>
<mi>T</mi>
<mo>~</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>Tmin</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mn>20</mn>
<mo>,</mo>
<msub>
<mi>Tave</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>Tmax</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mn>20</mn>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
Step 3:Classify to any pixel point p in input picture, classification results may be cloud, sky, current goal, not
Know target:
1) to any classified pixels point p, its wave band i gray values G is takeni, use GiWithGenerate broad sense elementary probability
Partition function mi, it is describedRespectively foregoing wave band i clouds, sky, the Triangular Fuzzy Number of current goal, it is described
Broad sense Basic probability assignment function be defined as in broad sense evidence theory to any one belong to Θ subset A, m (A) ∈ [0,
1], and meetThen m is 2ΘOn broad sense Basic probability assignment function, wherein 2ΘFor the power set of framework of identification,The broad sense Basic probability assignment function miGeneration
Method is:By GiWith fuzzy numberThe high point of intersection point is assigned to the reliability of corresponding single subset elements, by GiWith obscuring
NumberThe secondary high point of intersection point is assigned to the reliability of corresponding double subset elements, by GiWith fuzzy number
The low spot of intersection point is assigned to the reliability of corresponding more subset elements, wherein single subset elements refer to framework of identification in step 1
The subset { C } of Θ, { S } or { T }, double subset elements refer to the subset { C, S } of framework of identification Θ in step 1, C,
T } or { S, T }, more subset elements refer to the subset { C, S, T } of framework of identification Θ in step 1;Remember generation reliability it
With for Sum, if Sum is not less than 1, the reliability of generation is normalized to obtain mi;If Sum is less than 1,1-Sum is assigned to empty setReliability;
2) by the broad sense Basic probability assignment function m of 25 wave band generationsiMerge to obtain m using average fusion method, it is described average
Fusion method is:Whereinmi(i=1,2 ..., 25) it is the generalized base generated in step 2
This probability distribution function;
3) m after fusion is converted into probability distribution P using Pignistic probability transformation methods,
The conversion method is:
Wherein
4) classify according to obtained probability distribution P to pixel p, take P ({ C }), P ({ S }), P ({ T }),It is middle maximum
The corresponding classification of probability as pixel p classify as a result, C represents cloud in classification results, S represents sky, and T represents current mesh
Mark,Represent unknown object;
Step 4:ELIMINATION OF ITS INTERFERENCE is carried out to the region for being categorized as unknown object, to choose real movement unknown object, interference row
The principle removed is realized using the position correlation of consecutive frame zone of ignorance:
If knowing caused by random disturbances by mistake, then the change in location of consecutive frame zone of ignorance is larger, sets rational threshold value
Random disturbances are excluded, threshold value can be chosen according to target speed and (such as elect 1.5 times of target single frames displacements as);If static mesh
Mark, after 5-10 frames its total displacement be able to can equally be excluded close to 0;
Step 5:The Triangular Fuzzy Number mould of previous frame current goal, cloud, sky is updated according to the pixel classification results of present frame
Type, using the model after renewal as next two field picture cloud, sky, the other Fuzzy Math Model of three species of current goal, the model
Update method is:
For all pixels for being identified as cloud, the maximum max and minimum value min of these pixel wave bands i are obtained, if
max>Cmaxi, willThe upper bound is updated to max, if min<Cmini, willLower bound is updated to min;It is identified as sky for all
Pixel, the maximum max and minimum value min of these pixel wave bands i are obtained, if max>Smaxi, willThe upper bound is updated to
Max, if min<Smini, willLower bound is updated to min;For all pixels for being identified as current goal, these pictures are obtained
The maximum max and minimum value min of vegetarian refreshments wave band i, if max>Tmaxi, willThe upper bound is updated to max, if min<Tmini, will
Lower bound is updated to min.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711135254.4A CN107967449B (en) | 2017-11-13 | 2017-11-13 | A kind of multispectral image unknown object recognition methods based on broad sense evidence theory |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711135254.4A CN107967449B (en) | 2017-11-13 | 2017-11-13 | A kind of multispectral image unknown object recognition methods based on broad sense evidence theory |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107967449A true CN107967449A (en) | 2018-04-27 |
CN107967449B CN107967449B (en) | 2019-07-05 |
Family
ID=62001106
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711135254.4A Active CN107967449B (en) | 2017-11-13 | 2017-11-13 | A kind of multispectral image unknown object recognition methods based on broad sense evidence theory |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107967449B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109961012A (en) * | 2019-02-20 | 2019-07-02 | 博雅工道(北京)机器人科技有限公司 | A kind of underwater target tracking recognition methods |
CN111563532A (en) * | 2020-04-07 | 2020-08-21 | 西北工业大学 | Unknown target identification method based on attribute weight fusion |
CN111563596A (en) * | 2020-04-22 | 2020-08-21 | 西北工业大学 | Uncertain information reasoning target identification method based on evidence network |
CN112232375A (en) * | 2020-09-21 | 2021-01-15 | 西北工业大学 | Unknown type target identification method based on evidence theory |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102819838A (en) * | 2012-07-17 | 2012-12-12 | 北京市遥感信息研究所 | Hyperspectral remote sensing image change detection method based on multisource target characteristic support |
CN105513041A (en) * | 2015-10-28 | 2016-04-20 | 深圳大学 | Large-scale remote sensing image sea-land segmentation method and system |
CN105551031A (en) * | 2015-12-10 | 2016-05-04 | 河海大学 | Multi-temporal remote sensing image change detection method based on FCM and evidence theory |
-
2017
- 2017-11-13 CN CN201711135254.4A patent/CN107967449B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102819838A (en) * | 2012-07-17 | 2012-12-12 | 北京市遥感信息研究所 | Hyperspectral remote sensing image change detection method based on multisource target characteristic support |
CN105513041A (en) * | 2015-10-28 | 2016-04-20 | 深圳大学 | Large-scale remote sensing image sea-land segmentation method and system |
CN105551031A (en) * | 2015-12-10 | 2016-05-04 | 河海大学 | Multi-temporal remote sensing image change detection method based on FCM and evidence theory |
Non-Patent Citations (2)
Title |
---|
蒋雯: ""基于模糊特征属性参数最优融合的目标识别"", 《计算机仿真》 * |
邓勇: ""广义证据理论的基本框架"", 《西安交通大学学报》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109961012A (en) * | 2019-02-20 | 2019-07-02 | 博雅工道(北京)机器人科技有限公司 | A kind of underwater target tracking recognition methods |
CN111563532A (en) * | 2020-04-07 | 2020-08-21 | 西北工业大学 | Unknown target identification method based on attribute weight fusion |
CN111563532B (en) * | 2020-04-07 | 2022-03-15 | 西北工业大学 | Unknown target identification method based on attribute weight fusion |
CN111563596A (en) * | 2020-04-22 | 2020-08-21 | 西北工业大学 | Uncertain information reasoning target identification method based on evidence network |
CN111563596B (en) * | 2020-04-22 | 2022-06-03 | 西北工业大学 | Uncertain information reasoning target identification method based on evidence network |
CN112232375A (en) * | 2020-09-21 | 2021-01-15 | 西北工业大学 | Unknown type target identification method based on evidence theory |
Also Published As
Publication number | Publication date |
---|---|
CN107967449B (en) | 2019-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114202696B (en) | SAR target detection method and device based on context vision and storage medium | |
CN107967449A (en) | A kind of multispectral image unknown object recognition methods based on broad sense evidence theory | |
CN112347861B (en) | Human body posture estimation method based on motion feature constraint | |
CN103020992B (en) | A kind of video image conspicuousness detection method based on motion color-associations | |
CN107767406B (en) | A kind of multispectral image Dim target tracking method based on DS evidence theory | |
CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
CN107833239A (en) | A kind of searching of optimal matching method for tracking target based on weighted model constraint | |
CN106709938A (en) | Multi-target tracking method based on improved TLD (tracking-learning-detected) | |
CN110992378A (en) | Dynamic update visual tracking aerial photography method and system based on rotor flying robot | |
CN110334584A (en) | A kind of gesture identification method based on the full convolutional network in region | |
Zhang et al. | A survey on instance segmentation: Recent advances and challenges | |
Quan et al. | Research on human target recognition algorithm of home service robot based on fast-RCNN | |
CN114495170A (en) | Pedestrian re-identification method and system based on local self-attention inhibition | |
Li et al. | Research on target detection algorithm based on deep learning technology | |
CN104008374B (en) | Miner's detection method based on condition random field in a kind of mine image | |
Wei | Small object detection based on deep learning | |
Xiang et al. | Double-branch fusion network with a parallel attention selection mechanism for camouflaged object detection | |
CN109117852B (en) | Unmanned aerial vehicle image adaptation area automatic extraction method and system based on sparse representation | |
CN115496675A (en) | Shadow removal method based on Neighborwood attention mechanism | |
CN113902947A (en) | Method for constructing anti-network of natural image and generating infrared image of empty target | |
Li et al. | YOLOv7-based land and underwater target detection and recognition | |
Wang et al. | Gait Recognition based on lightweight CNNs | |
Qinghe et al. | Research on small target detection based on improved Faster R-CNN depth network method | |
CN108320305A (en) | A kind of multispectral image Dim target tracking method based on DS evidence theories | |
Zeng et al. | Study of YOLOX target detection method based on stand-alone self-attention |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |