CN108090913A - A kind of image, semantic dividing method based on object level Gauss-Markov random fields - Google Patents
A kind of image, semantic dividing method based on object level Gauss-Markov random fields Download PDFInfo
- Publication number
- CN108090913A CN108090913A CN201711316006.XA CN201711316006A CN108090913A CN 108090913 A CN108090913 A CN 108090913A CN 201711316006 A CN201711316006 A CN 201711316006A CN 108090913 A CN108090913 A CN 108090913A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msubsup
- msub
- msup
- munder
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 230000011218 segmentation Effects 0.000 claims abstract description 71
- 238000012417 linear regression Methods 0.000 claims abstract description 20
- 238000010276 construction Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000009826 distribution Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 4
- 238000002372 labelling Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005381 potential energy Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000000686 essence Substances 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/143—Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Coloring Foods And Improving Nutritive Qualities (AREA)
Abstract
The present invention proposes a kind of image, semantic dividing method based on object level Gauss Markov random fields, and step is:Initialization over-segmentation is carried out to pixel-level image, obtains object level image and Region adjacency graph, and defines neighborhood system, observational characteristic field and dividing mark field respectively on Region adjacency graph;Mark Label Field and neighborhood system are split according to object level, Gauss Markov modelings are carried out to the feature in each region of observational characteristic field and its feature of neighborhood, construction is for the object level equation of linear regression in each region;Probabilistic Modeling is carried out to Characteristic Field and Label Field respectively, the Posterior distrbutionp of dividing mark field is obtained according to Bayes criterions, and final segmentation result is obtained according to maximum posteriori criterion.The present invention can be used under complicated semantic and high spatial resolution background in carrying out semantic segmentation system to image in bulk, compared to artificial detection, drastically increase work efficiency.
Description
Technical field
The present invention relates to image, semantic segmentation technical field more particularly to it is a kind of based on object level Gauss-Markov with
The image, semantic dividing method on airport.
Background technology
Image, semantic segmentation refers to the pixel in image being grouped according to semantic difference expressed in image, the mistake
Journey is independently carried out by machine.
With the continuous development of modern sensor manufacturing process and imaging technique, handled image spatial resolution is more next
It is higher, and the amount of images obtained is increased with exponential form, if using segmentation by hand, inefficiency.Pervious Pixel-level
Dividing method can not consider larger range of spatial information, cause the waste of bulk information.And the ground based on object level in recent years
Reason analytical technology becomes a kind of hot spot technology for extracting image information, it is applied in image, semantic segmentation, it can be considered that
Larger range of spatial information.But the interaction relationship between provincial characteristics is ignored, segmentation precision has much room for improvement.Cause
This had not only ensured making full use of for spatial information, it is necessary to a kind of image, semantic dividing method, but also it can be considered that between provincial characteristics
It interacts.
The content of the invention
Cannot be taken into account for conventional images semantic segmentation method make full use of ensure spatial information and consider provincial characteristics it
Between the technical issues of interacting, the present invention proposes a kind of image, semantic segmentation based on object level Gauss-Markov random fields
Method, had not only ensured making full use of for spatial information, but also it can be considered that interaction between provincial characteristics.
In order to achieve the above object, the technical proposal of the invention is realized in this way:One kind is based on object level Gauss-
The image, semantic dividing method of Markov random fields, which is characterized in that its step are as follows:
Step 1:Initialization over-segmentation is carried out to the pixel-level image of reading, obtains the object being made of overdivided region
Grade image and corresponding object level Region adjacency graph RAG define the neighborhood system N of the image according to Region adjacency graph RAGO, it is right
As grade observational characteristic field YOWith object level dividing mark field XO;
Step 2:According to object level dividing mark field XOWith neighborhood system NOTo observational characteristic field YOEach region ri's
Feature and its feature of neighborhood carry out Gauss-Markov modelings, and construction is for each region riObject level linear regression side
Journey, i=1 ..., l;
Step 3:Respectively to observational characteristic field YOWith dividing mark field XOProbabilistic Modeling is carried out, and is obtained according to Bayes criterions
To dividing mark field XOPosterior distrbutionp, update iterative segmentation with maximum posteriori criterion and thus solve final segmentation.
The specific implementation step of the step 1 is as follows:
1) definition of location index set and Pixel-level are carried out to the high spatial resolution triple channel image I (R, G, B) of input
Observational characteristic set defines, it is assumed that image I (R, G, B) resolution ratio is m × n, is obtained:Location index set S={ sxy=(x, y) |
1≤x≤m;1≤y≤n }, Pixel-level observational characteristic collectionWherein,Represent picture at the s of position
The observational characteristic value of vegetarian refreshments,The respectively value of R, G of image, B component, m are the length of image, and n is the width of image
Degree, (x, y) are the position coordinates of pixel in image;
2) over-segmentation processing is carried out to pixel-level image with mean-shift methods according to the minimum area of setting:It will figure
L minimum area is segmented into as s as I (R, G, B) is undueminRegion, each region assigns label, obtains labelling matrix Ls={ ls|
S ∈ S }, wherein, element ls∈{1,…,l},s∈S;Thus the location index set R={ r of object level image are obtained1,r2,…,
rl, wherein, region ri=s | ls=i };
3) handled to obtain object level Region adjacency graph G=(R, E) according to over-segmentation, wherein, location index set R is object
Grade element, each element represent an overdivided region, E={ eij| 1≤i, j≤l } represent syntople, element eijRepresent area
Domain riIn with region rjAdjacent number of pixels, eij≠ 0 and if only if element RiAnd RjIt is adjacent;
4) object level observational characteristic field is defined on Region adjacency graph GAnd object
Grade segmentation Label FieldWherein,Represent region riObservational characteristic, | ri| represent region ri
Interior pixel number;XOIt is a random field,It is a stochastic variable,Wherein, K is segmentation class
Do not gather, k is previously given segmentation classification number;
5) object level neighborhood system is provided according to object level Region adjacency graph G=(R, E):Its
In,
The step 2 is as follows:
1) it can obtain what each overdivided region was included by location index set R in Region adjacency graph G=(R, E)
Number of pixels is the area parameters of object level element, obtains area matrix RS={ RSi| 1≤i≤l }, wherein RSi=| ri|;
2) x is setOIt is object level dividing mark field XOOne realization, according to xOObtain characteristic mean of all categories and feature
Covariance matrix realizes that flow is:
(a) known object grade segmentation Label Field is embodied as xO, calculate the corresponding segmentation of each pixel in original image
Generic, i.e. Pixel-level dividing mark matrixWherein
(b) characteristic mean m={ m are calculated respectivelyi| 1≤i≤k } and Eigen Covariance matrix ∑={ ∑i|1≤i≤k}:
3) for each object level element ri, give its dividing mark and be embodied asAfterwards, equation of linear regression is constructed such as
Under:
Wherein, ei~N (0, ∑h) it is a white Gaussian noise.
The specific method of the step 3 is as follows:
2) for object level observational characteristic field YO, it is not that joint probability modeling is directly carried out to observational characteristic, but to every
One object level element riResidual error item in the object level equation of linear regression constructed carries out joint modeling, obtains Characteristic Field
Likelihood function, i.e.,:
2) object level dividing mark field XOProbabilistic Modeling is carried out, from Markov-Gibbs equivalences, object level segmentation
Label Field meets Gibbs distributions, and the prior distribution for obtaining Label Field is as follows:
Wherein, Z is normalization constant, U (xO) represent that segmentation field is embodied as xOWhen energy, K for segmentation category set, V2
() is group potential-energy function, is provided by Potts models, i.e.,:
3) Posterior distrbutionp that Label Field can be obtained by Bayes formula is:
So the optimum of dividing mark is asked, which to translate into, seeks dividing mark field XOThe maximized problem of Posterior distrbutionp,
I.e.:
By loop iteration, dividing mark is updated, finally obtains segmentation result.
The specific implementation process of the loop iteration is:
5) Pixel-level MRF methods are realized by classical ICM algorithms first, obtain the segmentation generic of each pixel,
That is Pixel-level segmentation field result:xP={ xs| s ∈ S }, and then the Object Segmentation Label Field for obtaining primary iteration is realizedWhereinMode is mode function;
6) realization walked by object level dividing mark field in tIt is corresponded to according to the following formula
The characteristic mean of each classificationAnd Eigen Covariance
7) each object level element r is calculated respectivelyiObject level equation of linear regression:
8) computing object and Characteristic Field probability and Label Field probability, and by the update dividing mark of object, be specially respectively:
Beneficial effects of the present invention:Provide the molding semantic segmentation method to high spatial resolution RGB image;It can use
In the semantic segmentation of batch processing high spatial resolution RGB image, segmentation efficiency far is horizontal higher than traditional-handwork segmentation, also compares
Existing major part object-oriented grade partitioning scheme is efficient;Fixed value is directly assigned for the parameter to be estimated in equation of linear regression
Letter is simple and efficient, and precision is high.
Description of the drawings
It in order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention, for those of ordinary skill in the art, without creative efforts, can be with
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of the present invention.
Fig. 2 is the flow chart that the present invention initializes.
Fig. 3 is the exemplary plot of initialization process of the present invention.
Fig. 4 is the flow chart of equation of linear regression of the present invention construction.
Fig. 5 is the flow chart of present invention joint modeling.
Fig. 6 is experiment simulation figure of the present invention.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other under the premise of not making the creative labor
Embodiment belongs to the scope of protection of the invention.
As shown in Figure 1, a kind of image, semantic dividing method based on object level Gauss-Markov random fields, step is such as
Under:
Step 1:Initialization over-segmentation is carried out to the pixel-level image of reading, obtains the object being made of overdivided region
Grade image and corresponding object level Region adjacency graph RAG define the neighborhood system N of the image according to Region adjacency graph RAGO, it is right
As grade observational characteristic field YOWith object level dividing mark field XO。
In order to carry out object level graphical analysis, efficiency of algorithm is improved, it is necessary to initialization over-segmentation be carried out, to obtain region
Adjacent map RAG.Region adjacency graph RAG is that the spatial relationship between each overdivided region according to object level image obtains.
The method that image carries out initialization over-segmentation is the mean-shift methods of minimum area factor.According to pixel-level image feature,
Object level is obtained using the mean-shift methods for being related to minimum area parameter (contained pixel number i.e. in overdivided region)
Graphical representation finally acquires object level characteristics of image.As shown in Fig. 2, its specific implementation step is as follows:
1) definition of location index set and Pixel-level are carried out to the high spatial resolution triple channel image I (R, G, B) of input
Observational characteristic set defines, it is assumed that image I (R, G, B) resolution ratio is m × n, then can be respectively obtained:Location index set S
={ sxy=(x, y) | 1≤x≤m;1≤y≤n }, Pixel-level observational characteristic collectionWherein,
Represent the observational characteristic value of pixel at the s of position,The respectively value of R, G of image, B component, m are the length of image
Degree, n are the width of image, and (x, y) is the position coordinates of pixel in image.
2) over-segmentation processing is carried out to pixel-level image with mean-shift methods according to the minimum area of setting:It will figure
L minimum area is segmented into as s as I (R, G, B) is undueminRegion, each region assigns label, obtains labelling matrix Ls={ ls|
S ∈ S }, wherein, element ls∈{1,…,l},s∈S.Thus the location index set R={ r of object level image are obtained1,r2,…,
rl, wherein, region ri=s | ls=i }.After handling result gray processing as shown in Fig. 3 (a), the lines in figure are the knots of its segmentation
Fruit.
3) handled to obtain object level Region adjacency graph G=(R, E) according to over-segmentation.Wherein, location index set R is object
Grade element, each element represent an overdivided region.E={ eij| 1≤i, j≤l } represent syntople, element eijRepresent area
Domain riIn with region rjAdjacent number of pixels, eij≠ 0 and if only if element RiAnd RjIt is adjacent.
4) object level observational characteristic field is defined on Region adjacency graph GObject level
Dividing mark fieldWherein,Represent region riObservational characteristic, | ri| represent region riIt is interior
Pixel number.XOIt is a random field,It is a stochastic variable, represents overdivided region riSegmentation classification,Wherein K is segmentation category set, and k is previously given segmentation classification number.
5) object level neighborhood system is provided according to object level Region adjacency graph G=(R, E):Its
In,A part in rectangle frame in Fig. 3 (a) is amplified to obtain Fig. 3 (b), in Fig. 3 (b)
Each area carries out neighbourhood signatures such as Fig. 3 (c).
Step 2:Each region rs of the mark Label Field XO and neighborhood system NO to observational characteristic field YO is split according to object leveli
Feature and its feature of neighborhood carry out Gauss-Markov modelings, construction is for each region riObject level linear regression
Equation, i=1 ..., l.
Object level equation of linear regression use object level element size and boundary length as equation of linear regression
Parameter, to each object level element build equation of linear regression, as shown in figure 4, being as follows:
1) it can obtain what each overdivided region was included by location index set R in Region adjacency graph G=(R, E)
It is considered as the area parameters of object level element by number of pixels, obtains area matrix RS={ RSi| 1≤i≤l }, wherein RSi=|
ri|。
2) x is assumedOIt is object level dividing mark field XOOne realization, according to xOObtain characteristic mean of all categories and spy
Covariance matrix is levied, realizes that flow is:
(a) known object grade segmentation Label Field is embodied as xO, calculate the corresponding segmentation of each pixel in original image
Generic, i.e. Pixel-level dividing mark matrixWherein
(b) characteristic mean m={ mi | 1≤i≤k } and Eigen Covariance matrix ∑={ ∑ are calculated respectivelyi|1≤i≤k}:
3) for each object level element ri, give its dividing mark and be embodied as xiAfter O, equation of linear regression is constructed
It is as follows:
Wherein, for the ease of calculating, it is assumed that ei~N (0, ∑h) it is a white Gaussian noise.
Step 3:Respectively to observational characteristic field YOWith dividing mark field XOProbabilistic Modeling is carried out, and is obtained according to Bayes criterions
To dividing mark field XOPosterior distrbutionp, update iterative segmentation with maximum posteriori criterion and thus solve final segmentation.
Probabilistic Modeling is included by the error term construction observational characteristic field Y in object level equation of linear regressionOMultivariate Normal
It is distributed and using Potts Construction of A Model segmentation mark Label Field XOGibbs distribution.Finally segmentation result is:It is distributed using Gibbs
Sampling update iterative segmentation, final output convergence solution.As shown in figure 5, concrete operations are as follows:
1) for object level observational characteristic field YO, it is not that joint probability modeling is directly carried out to observational characteristic, but to every
One object level element riResidual error item in the object level equation of linear regression constructed carries out joint modeling, obtains Characteristic Field
Likelihood function, i.e.,:
2) object level dividing mark field XOProbabilistic Modeling is carried out, since Label Field has geneva, by Markov-Gibbs
Equivalence understands that Label Field meets Gibbs distributions, and the prior distribution for obtaining Label Field is as follows:
Wherein, Z is normalization constant, U (xO) represent that segmentation field is embodied as xOWhen energy, V2() is group potential energy letter
Number, is provided, i.e., by Potts models:
3) Posterior distrbutionp that Label Field can be obtained by Bayes formula is:
So the optimum of dividing mark is asked, which to translate into, seeks dividing mark field XOThe maximized problem of Posterior distrbutionp,
I.e.:
By loop iteration, dividing mark is updated, finally obtains result.Specific loop iteration process is as follows:
1) Pixel-level MRF (Markov are realized by classical ICM (iteration condition model) algorithm first
Random field) method, obtain the segmentation generic of each pixel, i.e. Pixel-level segmentation field result:xP={ xs|s∈
S }, and then the Object Segmentation Label Field for obtaining primary iteration is realizedWherein
I.e. for overdivided region ri, dividing mark is the mode of its interior pixels point dividing mark.
2) realization walked by object level dividing mark field in tIt is corresponded to according to the following formula
The characteristic mean of each classificationAnd Eigen Covariance
3) each object level element r is calculated respectivelyiObject level equation of linear regression:
4) computing object and Characteristic Field probability and Label Field probability, and by the update dividing mark of object, be specially respectively:
The present invention operation platform be:Core [email protected], RAM:4G, 64 win10 systems, 2015a editions
matlab.(coloured image gray processing) shown in the coloured image of Aerial Images 1024_1 such as Fig. 6 (a1), true segmentation by hand
As shown in Fig. 6 (a2).To image 1024_1 ICM methods, make β=0.5, coloured image such as Fig. 6 of obtained segmentation result
(a3) shown in.To image 1024_1 GMRF methods, make β=0.5, coloured image such as Fig. 6 (a4) institute of obtained segmentation result
Show.It is three layers, β=0.5 with " Haar " wavelet decomposition to image 1024_1 MRMRF methods, the colour of obtained segmentation result
Shown in image such as Fig. 6 (a5).To image 1024_1 OMRF methods, and make s=256, β=0.5, obtained segmentation result
Shown in coloured image such as Fig. 6 (a6).To the image 1024_1 present invention (OGMRF-RC) methods, and make s=256, β=0.5,
Shown in the coloured image of obtained segmentation result such as Fig. 6 (a7).Shown in the coloured image of Aerial Images 1024_2 such as Fig. 6 (b1),
True segmentation by hand is as shown in Fig. 6 (b2).To image 1024_2 ICM methods, make β=0.3, the colour of obtained segmentation result
Shown in image such as Fig. 6 (b3).To image 1024_2 GMRF methods, make β=0.3, the coloured image of obtained segmentation result is such as
Shown in Fig. 6 (b4).It is three layers, β=0.3 with " Haar " wavelet decomposition to image 1024_2 MRMRF methods, obtained segmentation
As a result shown in coloured image such as Fig. 6 (b5).To image 1024_2 OMRF methods, and make s=144, β=0.3 obtains
Shown in the coloured image of segmentation result such as Fig. 6 (b6).To the image 1024_2 present invention (OGMRF-RC) methods, and make s=
144, β=0.3, shown in coloured image such as Fig. 6 (b7) of obtained segmentation result.Aerial Images 1024_1's and image 1024_2
The Kappa coefficients of segmentation result are as shown in table 1, and the overall accuracy OA of segmentation result is as shown in table 2.
The Kappa coefficients of 1 segmentation result of table
The overall accuracy (Overall Accuracy, OA) of 2 segmentation result of table
The segmentation precision of the present invention is best it can be seen from data in Fig. 6 and table 1-2.Aerial Images include more
Texture information, the spectrum value of the subobject in same class differs greatly, and different classes of subobject may have it is similar
Spectrum value.For example, in urban parts, roof and garden have a different spectral values, but the trees of urban parts and forest part
Spectral value be similar.For these reasons, three kinds of methods based on pixel have many mistake classification fine crushing.With based on picture
The method of element is compared, and overdivided region is considered as elementary cell by object-based method, therefore significantly optimizes segmentation precision.
OMRF methods are modeled property field using the probability distribution of the feature of object, and OGMRF-RC methods are then to utilize object level
The probability distribution of residual error item is modeled property field in equation of linear regression.OGMRF-RC methods the advantage of doing so is that, can
With reduce in iterative process it is generic between influence of the spectrum change for segmentation.For example, in the top half of Fig. 6 (a7),
Large-scale bare area and forest are accurately divided into idle part, rather than in Fig. 6 (a6) be divided into house part OMRF that
Sample.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modifications, equivalent replacements and improvements are made should all be included in the protection scope of the present invention god.
Claims (5)
1. a kind of image, semantic dividing method based on object level Gauss-Markov random fields, which is characterized in that its step is such as
Under:
Step 1:Initialization over-segmentation is carried out to the pixel-level image of reading, obtains the object level figure being made of overdivided region
Picture and corresponding object level Region adjacency graph RAG define the neighborhood system N of the image according to Region adjacency graph RAGO, object level
Observational characteristic field YOWith object level dividing mark field XO;
Step 2:According to object level dividing mark field XOWith neighborhood system NOTo observational characteristic field YOEach region riFeature and
The feature of its neighborhood carries out Gauss-Markov modelings, and construction is for each region riObject level equation of linear regression, i=
1,…,l;
Step 3:Respectively to observational characteristic field YOWith dividing mark field XOProbabilistic Modeling is carried out, and is divided according to Bayes criterions
Cut Label Field XOPosterior distrbutionp, update iterative segmentation with maximum posteriori criterion and thus solve final segmentation.
2. the image, semantic dividing method according to claim 1 based on object level Gauss-Markov random fields, special
Sign is that the specific implementation step of the step 1 is as follows:
1) definition of location index set is carried out to the high spatial resolution triple channel image I (R, G, B) of input and Pixel-level is observed
Characteristic set defines, it is assumed that image I (R, G, B) resolution ratio is m × n, is obtained:Location index set S={ sxy=(x, y) | 1≤x
≤m;1≤y≤n }, Pixel-level observational characteristic collectionWherein,Represent pixel at the s of position
Observational characteristic value,The respectively value of R, G of image, B component, m are the length of image, and n is the width of image,
(x, y) is the position coordinates of pixel in image;
2) over-segmentation processing is carried out to pixel-level image with mean-shift methods according to the minimum area of setting:By image I
(R, G, B) is too segmented into l minimum area as sminRegion, each region assigns label, obtains labelling matrix Ls={ ls|s∈
S }, wherein, element ls∈{1,…,l},s∈S;Thus the location index set R={ r of object level image are obtained1,r2,…,
rl, wherein, region ri=s | ls=i };
3) handled to obtain object level Region adjacency graph G=(R, E) according to over-segmentation, wherein, location index set R is object level member
Element, each element represent an overdivided region, E={ eij| 1≤i, j≤l } represent syntople, element eijRepresent region ri
In with region rjAdjacent number of pixels, eij≠ 0 and if only if element RiAnd RjIt is adjacent;
4) object level observational characteristic field is defined on Region adjacency graph GSplit with object level
Label FieldWherein,Represent region riObservational characteristic, | ri| represent region riInterior picture
Vegetarian refreshments number;XOIt is a random field,It is a stochastic variable,Wherein, K is segmentation classification collection
It closes, k is previously given segmentation classification number;
5) object level neighborhood system is provided according to object level Region adjacency graph G=(R, E):Wherein,
3. the image, semantic dividing method according to claim 1 based on object level Gauss-Markov random fields, special
Sign is that the step 2 is as follows:
1) pixel that each overdivided region included can be obtained by location index set R in Region adjacency graph G=(R, E)
Number is the area parameters of object level element, obtains area matrix RS={ RSi| 1≤i≤l }, wherein RSi=| ri|;
2) x is setOIt is object level dividing mark field XOOne realization, according to xOObtain characteristic mean of all categories and feature association side
Poor matrix realizes that flow is:
(a) known object grade segmentation Label Field is embodied as xO, calculate class belonging to the corresponding segmentation of each pixel in original image
Not, i.e. Pixel-level dividing mark matrixWherein
(b) characteristic mean μ={ μ is calculated respectivelyi| 1≤i≤k } and Eigen Covariance matrix ∑={ ∑i|1≤i≤k}:
<mrow>
<msub>
<mi>&mu;</mi>
<mi>h</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
<mo>,</mo>
<mn>1</mn>
<mo>&le;</mo>
<mi>h</mi>
<mo>&le;</mo>
<mi>k</mi>
<mo>,</mo>
</mrow>
<mrow>
<msub>
<mi>&Sigma;</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>h</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>h</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
<mo>,</mo>
<mn>1</mn>
<mo>&le;</mo>
<mi>h</mi>
<mo>&le;</mo>
<mi>k</mi>
<mo>;</mo>
</mrow>
3) for each object level element ri, give its dividing mark and be embodied asAfterwards, it is as follows to construct equation of linear regression:
<mrow>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>=</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>e</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
</mrow>
<mrow>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<msub>
<mi>e</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msub>
<mi>r</mi>
<msup>
<mi>j</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>e</mi>
<mrow>
<msup>
<mi>ij</mi>
<mo>&prime;</mo>
</msup>
</mrow>
</msub>
</mrow>
</mfrac>
<mo>&CenterDot;</mo>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msub>
<mi>r</mi>
<msup>
<mi>j</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<msup>
<mi>j</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
Wherein,It is a white Gaussian noise.
4. the image, semantic dividing method according to claim 1 based on object level Gauss-Markov random fields, special
Sign is that the specific method of the step 3 is as follows:
1) for object level observational characteristic field YO, it is not that joint probability modeling is directly carried out to observational characteristic, but it is right to each
As grade element riResidual error item in the object level equation of linear regression constructed carries out joint modeling, obtains the likelihood letter of Characteristic Field
Number, i.e.,:
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>p</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>|</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mi>&Pi;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<mi>p</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>Y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<munderover>
<mi>&Pi;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<mfrac>
<mn>1</mn>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mi>&pi;</mi>
<mo>)</mo>
</mrow>
<mrow>
<mn>3</mn>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
<msup>
<mrow>
<mo>|</mo>
<msub>
<mi>&Sigma;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>&lsqb;</mo>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>-</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>&Sigma;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>-</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<munderover>
<mi>&Pi;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<mfrac>
<mn>1</mn>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mi>&pi;</mi>
<mo>)</mo>
</mrow>
<mrow>
<mn>3</mn>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
<msup>
<mrow>
<mo>|</mo>
<msub>
<mi>&Sigma;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
</mrow>
</mfrac>
<mi>exp</mi>
<mo>&lsqb;</mo>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<msub>
<mi>e</mi>
<mi>i</mi>
</msub>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>&Sigma;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<msubsup>
<mi>e</mi>
<mi>i</mi>
<mi>T</mi>
</msubsup>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
2) object level dividing mark field XOProbabilistic Modeling is carried out, from Markov-Gibbs equivalences, object level dividing mark field
Meet Gibbs distributions, the prior distribution for obtaining Label Field is as follows:
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>Z</mi>
</mfrac>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mi>U</mi>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>Z</mi>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
</munder>
<mi>U</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>U</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>&Element;</mo>
<mi>R</mi>
<mo>,</mo>
<mi>j</mi>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>V</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<mi>R</mi>
<mo>-</mo>
<mo>{</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>}</mo>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<munderover>
<mi>&Pi;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mi>U</mi>
<mo>(</mo>
<mrow>
<mo>{</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>}</mo>
<mo>&cup;</mo>
<mo>{</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>}</mo>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mi>U</mi>
<mo>(</mo>
<mrow>
<mo>{</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>}</mo>
<mo>&cup;</mo>
<mo>{</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>}</mo>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>:</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>V</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>:</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>V</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>;</mo>
</mrow>
Wherein, Z is normalization constant, U (xO) represent that segmentation field is embodied as xOWhen energy, K for segmentation category set, V2(·)
For group potential-energy function, provided by Potts models, i.e.,:
<mrow>
<msub>
<mi>V</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>&beta;</mi>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>&NotEqual;</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mi>i</mi>
<mo>&NotEqual;</mo>
<mi>j</mi>
<mo>;</mo>
</mrow>
3) Posterior distrbutionp that Label Field can be obtained by Bayes formula is:
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>|</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>|</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
So the optimum of dividing mark is asked, which to translate into, seeks dividing mark field XOThe maximized problem of Posterior distrbutionp, i.e.,:
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>O</mi>
</msup>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>max</mi>
</mrow>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
</munder>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>|</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>max</mi>
</mrow>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
</munder>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>Y</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>y</mi>
<mi>O</mi>
</msup>
<mo>|</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>O</mi>
</msup>
<mo>=</mo>
<msup>
<mi>x</mi>
<mi>O</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
By loop iteration, dividing mark is updated, finally obtains segmentation result.
5. the image, semantic dividing method according to claim 4 based on object level Gauss-Markov random fields, special
Sign is that the specific implementation process of the loop iteration is:
1) Pixel-level MRF methods are realized by classical ICM algorithms first, obtains the segmentation generic of each pixel, i.e. picture
Plain grade splits field result:xP={ xs| s ∈ S }, and then the Object Segmentation Label Field for obtaining primary iteration is realizedWhereinMode is mode function;
2) realization walked by object level dividing mark field in tIt obtains according to the following formula corresponding each
The characteristic mean of classificationAnd Eigen Covariance
<mrow>
<msubsup>
<mi>&mu;</mi>
<mi>h</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
<mo>,</mo>
</mrow>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mi>h</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>h</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>s</mi>
<mi>P</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>h</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>:</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>=</mo>
<mi>h</mi>
</mrow>
</munder>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>s</mi>
<mo>&Element;</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
3) each object level element r is calculated respectivelyiObject level equation of linear regression:
<mrow>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mo>=</mo>
<munder>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>:</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
</mrow>
</munder>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>&mu;</mi>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msubsup>
<mi>e</mi>
<mi>i</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</msubsup>
</mrow>
4) computing object and Characteristic Field probability and Label Field probability, and by the update dividing mark of object, be specially respectively:
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>max</mi>
</mrow>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>&Element;</mo>
<mi>K</mi>
</mrow>
</munder>
<mi>p</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>Y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>X</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>=</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>y</mi>
<mi>j</mi>
<mi>O</mi>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mi>x</mi>
<mi>j</mi>
<mrow>
<mi>O</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
</mrow>
</msubsup>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msubsup>
<mi>N</mi>
<mi>i</mi>
<mi>O</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711316006.XA CN108090913B (en) | 2017-12-12 | 2017-12-12 | Image semantic segmentation method based on object-level Gauss-Markov random field |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711316006.XA CN108090913B (en) | 2017-12-12 | 2017-12-12 | Image semantic segmentation method based on object-level Gauss-Markov random field |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108090913A true CN108090913A (en) | 2018-05-29 |
CN108090913B CN108090913B (en) | 2020-06-19 |
Family
ID=62173916
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711316006.XA Active CN108090913B (en) | 2017-12-12 | 2017-12-12 | Image semantic segmentation method based on object-level Gauss-Markov random field |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108090913B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108830319A (en) * | 2018-06-12 | 2018-11-16 | 北京合众思壮科技股份有限公司 | A kind of image classification method and device |
CN109615637A (en) * | 2019-01-16 | 2019-04-12 | 中国科学院地理科学与资源研究所 | A kind of improved remote sensing image Hybrid Techniques |
CN110136143A (en) * | 2019-05-16 | 2019-08-16 | 河南大学 | Geneva based on ADMM algorithm multiresolution remote sensing image segmentation method off field |
CN111210433A (en) * | 2019-04-16 | 2020-05-29 | 河南大学 | Markov field remote sensing image segmentation method based on anisotropic potential function |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295236A (en) * | 2013-05-29 | 2013-09-11 | 湘潭大学 | Method for building Markov multi-feature random field model and technology for segmenting brain MR (magnetic resonance) images on basis of model |
CN106600611A (en) * | 2016-12-23 | 2017-04-26 | 西安电子科技大学 | SAR image segmentation method based on sparse triple Markov field |
CN106951830A (en) * | 2017-02-23 | 2017-07-14 | 北京联合大学 | A kind of many object marking methods of image scene constrained based on priori conditions |
CN107180434A (en) * | 2017-05-23 | 2017-09-19 | 中国地质大学(武汉) | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic |
-
2017
- 2017-12-12 CN CN201711316006.XA patent/CN108090913B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295236A (en) * | 2013-05-29 | 2013-09-11 | 湘潭大学 | Method for building Markov multi-feature random field model and technology for segmenting brain MR (magnetic resonance) images on basis of model |
CN106600611A (en) * | 2016-12-23 | 2017-04-26 | 西安电子科技大学 | SAR image segmentation method based on sparse triple Markov field |
CN106951830A (en) * | 2017-02-23 | 2017-07-14 | 北京联合大学 | A kind of many object marking methods of image scene constrained based on priori conditions |
CN107180434A (en) * | 2017-05-23 | 2017-09-19 | 中国地质大学(武汉) | Polarization SAR image segmentation method based on super-pixel and fractal net work evolution algorithmic |
Non-Patent Citations (2)
Title |
---|
CHEN ZHENG, ET AL.: "Semantic Segmentation of Remote Sensing Imagery Using Object-Based Markov Random Field Model With Region Penalties", 《IEEE JOURNAL OF SELECTED TOPICS IN ALPPLIED EARTH OBSERVATION AND REMOTE SENSING》 * |
YANG YONG, ET AL.: "Segmentation of SAR imagery using the Gaussian Markov random field model", 《PROCEEDINGS 7TH INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108830319A (en) * | 2018-06-12 | 2018-11-16 | 北京合众思壮科技股份有限公司 | A kind of image classification method and device |
CN108830319B (en) * | 2018-06-12 | 2022-09-16 | 北京合众思壮科技股份有限公司 | Image classification method and device |
CN109615637A (en) * | 2019-01-16 | 2019-04-12 | 中国科学院地理科学与资源研究所 | A kind of improved remote sensing image Hybrid Techniques |
CN111210433A (en) * | 2019-04-16 | 2020-05-29 | 河南大学 | Markov field remote sensing image segmentation method based on anisotropic potential function |
CN111210433B (en) * | 2019-04-16 | 2023-03-03 | 河南大学 | Markov field remote sensing image segmentation method based on anisotropic potential function |
CN110136143A (en) * | 2019-05-16 | 2019-08-16 | 河南大学 | Geneva based on ADMM algorithm multiresolution remote sensing image segmentation method off field |
Also Published As
Publication number | Publication date |
---|---|
CN108090913B (en) | 2020-06-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Tyleček et al. | Spatial pattern templates for recognition of objects with regular structure | |
CN108573276B (en) | Change detection method based on high-resolution remote sensing image | |
Alshehhi et al. | Hierarchical graph-based segmentation for extracting road networks from high-resolution satellite images | |
CN104915636B (en) | Remote sensing image road recognition methods based on multistage frame significant characteristics | |
CN103049763B (en) | Context-constraint-based target identification method | |
CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
Ta et al. | Graph-based tools for microscopic cellular image segmentation | |
CN108090913A (en) | A kind of image, semantic dividing method based on object level Gauss-Markov random fields | |
CN103198333B (en) | A kind of automatic semantic marker method of high-resolution remote sensing image | |
CN102496034B (en) | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words | |
CN107909015A (en) | Hyperspectral image classification method based on convolutional neural networks and empty spectrum information fusion | |
Liu et al. | Multiscale road centerlines extraction from high-resolution aerial imagery | |
CN102096821A (en) | Number plate identification method under strong interference environment on basis of complex network theory | |
CN103426158B (en) | The method of two phase Remote Sensing Imagery Change Detection | |
CN110163239A (en) | A kind of Weakly supervised image, semantic dividing method based on super-pixel and condition random field | |
Merabet et al. | Building roof segmentation from aerial images using a line-and region-based watershed segmentation technique | |
CN112950780B (en) | Intelligent network map generation method and system based on remote sensing image | |
CN106846322A (en) | Based on the SAR image segmentation method that curve wave filter and convolutional coding structure learn | |
Masouleh et al. | Fusion of deep learning with adaptive bilateral filter for building outline extraction from remote sensing imagery | |
CN105956610B (en) | A kind of remote sensing images classification of landform method based on multi-layer coding structure | |
CN107358249A (en) | The hyperspectral image classification method of dictionary learning is differentiated based on tag compliance and Fisher | |
Weinmann et al. | A hybrid semantic point cloud classification-segmentation framework based on geometric features and semantic rules | |
CN104346814A (en) | SAR (specific absorption rate) image segmentation method based on hierarchy visual semantics | |
Ouzounis et al. | Interactive collection of training samples from the max-tree structure | |
CN102737232B (en) | Cleavage cell recognition method |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20240117 Address after: Building 43, Zone A, Energy Conservation and Environmental Protection Innovation Park, No. 199 Hongwu Avenue, Tangqiao Town, Zhangjiagang City, Jiangsu Province, 215000 Patentee after: Suzhou Qingchen Technology Co.,Ltd. Address before: No.85, Minglun street, Shunhe District, Kaifeng City, Henan Province Patentee before: Henan University |
|
TR01 | Transfer of patent right |