CN106408587B - Regard SAR image segmentation method and device more - Google Patents
Regard SAR image segmentation method and device more Download PDFInfo
- Publication number
- CN106408587B CN106408587B CN201610833053.0A CN201610833053A CN106408587B CN 106408587 B CN106408587 B CN 106408587B CN 201610833053 A CN201610833053 A CN 201610833053A CN 106408587 B CN106408587 B CN 106408587B
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- calculate
- gamma
- weight
- 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.)
- Expired - Fee Related
Links
Classifications
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Image Analysis (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
SAR image segmentation method and device are regarded the present invention provides a kind of, wherein this method includes more：It reads to be split mostly regarding SAR image；The double weight w of initialization；Repeat following step：Calculate Gamma distribution scale parameters β；Calculate the probability p (z | w) of above-mentioned image；Calculate the distribution function p (w) of above-mentioned double weight w；Calculate merit functions L；Above-mentioned double weight w are updated according to gradient method；Updated double weight w are substituted into merit functions L；Until | L(t+1)‑L(t)| when being less than preset threshold epsilon, stops executing above-mentioned steps, the classification in above-mentioned image belonging to each pixel is determined according to current double weight w；Segmentation result is exported according to the classification belonging to each pixel.The noiseproof feature that the present invention divides image is preferable, and segmentation result accidentally divides phenomenon less, and partitioning boundary fitting is accurate.
Description
Technical field
The present invention relates to technical field of image segmentation, and SAR image segmentation method and dress are regarded in particular to a kind of more
It sets.
Background technology
Synthetic aperture radar (Synthetic Aperture Radar, SAR) is a kind of high-resolution imaging radar, is passed through
It receives the electromagnetic wave of target scattering and is converted into image to record in the form of atural object, the intrinsic spot caused by distinctive image-forming mechanism
Spot noise brings huge difficulty to image segmentation.Although multiple-look technique can reduce partial noise, in practical application
In mostly still have a large amount of speckle noises depending on SAR image, therefore, regard SAR image segmentation method noise immunity and accuracy one
It is directly the hot issue of research.
Currently, the method mostly depending on SAR image segmentation mainly has：Threshold method, boundary method, clustering procedure and statistical model method etc..
Wherein, most widely used is statistic law, the complex distributions situation of generally use mixed model picture engraving.In mixed model
Most common is gauss hybrid models (Gaussian Mixture Model, GMM), assumes the gray value clothes of pixel in image
From Gaussian Profile, still, weight coefficient in Traditional GM M be with vector indicate only with single weight of cluster correlation, and SAR figures
As obeying Gamma distributions, it is not accurate enough that the above problem can cause Traditional GM M to model SAR image.
The problem that SAR image segmentation method noiseproof feature is poor and segmentation result is undesirable is regarded for above-mentioned, at present not yet more
It is proposed effective solution scheme.
Invention content
In view of this, regarding SAR image segmentation method and device the purpose of the present invention is to provide a kind of, figure can be improved more
As the noiseproof feature and enhancing segmentation precision of segmentation.
To achieve the goals above, technical solution used in the embodiment of the present invention is as follows：
In a first aspect, an embodiment of the present invention provides one kind regarding SAR image segmentation method more, including：
It reads to be split mostly regarding SAR image；It is above-mentioned to be split mostly to be indicated using Characteristic Field z depending on SAR image：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice point position
It sets, D is image area；
The double weight w of initialization：
wi=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number；
Repeat following step：Calculate Gamma distribution scale parameters β：
Wherein, α is Gamma distributional pattern parameters；
Calculate the probability p (z | w) of above-mentioned image；
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Calculate the distribution function p (w) of above-mentioned double weight w：
Wherein, A is normalization coefficient, and η is neighborhood function coefficient, NiFor with (xi,yi) centered on 8 neighborhood territory pixel set,
And meet i ' ∈ Ni,i'≠i；
Calculate merit functions L, above-mentioned merit functions L is probability p (z | w) and distribution function p (w) joint probability distribution letters
Several logarithmic functions；
Above-mentioned double weight w are updated according to gradient method：
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)It is as follows for gradient：
Updated double weight w are substituted into merit functions L；
Until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop executing above-mentioned steps, it is true according to current double weight w
Classification in fixed above-mentioned image belonging to each pixel；
Segmentation result is exported according to the classification belonging to each pixel.
With reference to first aspect, an embodiment of the present invention provides the first possible embodiments of first aspect, wherein meter
Calculating merit functions L includes：
Calculating probability p (z | w) and distribution function p (w) joint probability distribution function p (z, w)：
P (z, w)=p (z | w) p (w)；
Logarithm is taken to the joint probability distribution function：
L (w)=log p (z, w)=log p (z | w)+log p (w).
With reference to first aspect, an embodiment of the present invention provides second of possible embodiments of first aspect, wherein really
Determining the classification belonging to above-mentioned pixel includes：
Calculate the maximum value of double weights：
ci=arg max { wil, l=1 ..., k }；
Wherein, ciFor the label of ith pixel generic；
Using above-mentioned maximum value as the classification belonging to above-mentioned pixel.
With reference to first aspect, an embodiment of the present invention provides the third possible embodiments of first aspect, wherein presses
Include according to the classification output segmentation result belonging to each pixel：With the mean value of the intensity of all pixels in same category in image
The intensity for updating all pixels in the same category obtains mean value image；The mean value image is shown by display device.
With reference to first aspect, an embodiment of the present invention provides the 4th kind of possible embodiment of first aspect, further include：
Constant is set, and above-mentioned constant includes：Class number k, Gamma distributional pattern parameter alpha, normaliztion constant A and neighborhood function coefficient η.
Second aspect, the embodiment of the present invention additionally provide one kind and regarding SAR image segmenting device more, including：Read module is used
It is to be split mostly regarding SAR image in reading；It is above-mentioned to be split mostly to be indicated using Characteristic Field z depending on SAR image：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice point position
It sets, D is image area；
Double weights initialisation modules, for initializing double weight w：
wi=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number；
Weight parameter iteration update module, for repeating following calculating：Calculate Gamma distribution scale parameters β：
Wherein, α is Gamma distributional pattern parameters；
Calculate the probability p (z | w) of image；
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Calculate the distribution function p (w) of above-mentioned double weight w：
Wherein, A is normalization coefficient, and η is neighborhood function coefficient, NiFor with (xi,yi) centered on 8 neighborhood territory pixel set,
And meet i ' ∈ Ni,i'≠i；
Calculate merit functions L, above-mentioned merit functions L is probability p (z | w) and distribution function p (w) joint probability distribution letters
Several logarithmic functions；
Above-mentioned double weight w are updated according to gradient method：
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)It is as follows for gradient：
Updated double weight w are substituted into merit functions L；
Until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop executing above-mentioned calculating；
Category determination module, for determining the classification in image belonging to each pixel according to current double weight w；
Output module exports segmentation result for being split to image according to the classification belonging to each pixel.
In conjunction with second aspect, an embodiment of the present invention provides the first possible embodiments of second aspect, wherein on
Stating weight parameter iteration update module includes：First computing unit, for calculating probability p (z | w) and distribution function p's (w)
Joint probability distribution function p (z, w)：
P (z, w)=p (z | w) p (w)；
Second computing unit, for taking logarithm to above-mentioned joint probability distribution function：
L (w)=log p (z, w)=log p (z | w)+log p (w).
In conjunction with second aspect, an embodiment of the present invention provides second of possible embodiments of second aspect, wherein class
Other determining module includes：Maximum value calculation unit, the maximum value for calculating double weights：
ci=arg max { wil, l=1 ..., k }；
Wherein, ciFor the label of ith pixel generic；
Classification determination unit, for using above-mentioned maximum value as the classification belonging to pixel.
In conjunction with second aspect, an embodiment of the present invention provides the third possible embodiments of second aspect, wherein on
Stating output module includes：Equal value cell, it is same that the mean value of the intensity for all pixels in same category in image updates this
The intensity of all pixels in classification obtains mean value image；Display unit shows the mean value image for passing through display device.
In conjunction with second aspect, an embodiment of the present invention provides the 4th kind of possible embodiments of second aspect, further include：
Constant setup module, for constant to be arranged, above-mentioned constant includes：Class number k, Gamma distributional pattern parameter alpha, normaliztion constant A and
Neighborhood function coefficient η.
It is provided in an embodiment of the present invention to regard SAR image segmentation method and device more, using the double power indicated based on generic field
Weight Gamma mixed models model characteristics of image field；In order to introduce neighborhood relationships, it is based on Markov random field model
Center pixel generic weight is considered as mean value in conjunction with error sum of squares theory, constructs it by (Markov Random Filed, MRF)
Sum of squares function with neighborhood territory pixel generic weight difference is characterized with describing the otherness of pixel in neighborhood window with error sum of squares
Difference can make full use of generic information, enhance the segmentation precision of image.Double weights in estimating Gamma mixed models simultaneously
When, the embodiment of the present invention solves double weights using gradient method, and iteration speed is fast, is less prone to local optimum problem.
Description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of flow chart more regarding SAR image segmenting device that the embodiment of the present invention is provided；
Fig. 2 shows a kind of structural schematic diagrams more regarding SAR image segmenting device that the embodiment of the present invention is provided.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
Middle attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
It is a part of the embodiment of the present invention, instead of all the embodiments.The present invention being usually described and illustrated herein in the accompanying drawings is real
Applying the component of example can be arranged and designed with a variety of different configurations.Therefore, below to the reality of the present invention provided in attached drawing
The detailed description for applying example is not intended to limit the range of claimed invention, but is merely representative of the selected implementation of the present invention
Example.Based on the embodiment of the present invention, what those skilled in the art were obtained without making creative work is all
Other embodiment shall fall within the protection scope of the present invention.
The problem of noiseproof feature and accuracy difference existing for SAR image segmentation method is regarded in view of in the prior art more,
SAR image segmentation method and device are regarded an embodiment of the present invention provides a kind of, which may be used corresponding software and hard more
Part is realized.It is described below by embodiment.
Embodiment 1
Fig. 1 shows the flow diagram provided in an embodiment of the present invention for more regarding SAR image segmentation method.It below will be to figure
The detailed process of method shown in 1 is described in detail.
Step S110 is read to be split mostly regarding SAR image.
This is to be split mostly to be indicated depending on SAR image using Characteristic Field z：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice point position
It sets, D is image area.
Step S120 initializes double weight w.
In the present embodiment, double weight w are defined and indicate relationship between pixel and generic, it is specific as follows：
wi=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number.Wherein, initial double
Weight generates at random, and row indicates that pixel, row indicate classification, and element value range is 0~1 in matrix, and meet per a line and
It is 1.
Step S130 calculates Gamma distribution scale parameters β.
In the present embodiment, Gamma distribution scale parameters β is definedlFor about generic weight wilFunction, it is specific as follows：
Wherein, α is Gamma distributional pattern parameters, and for regarding SAR image, α is equal to it and regards number more.Based on double weights, in conjunction with
Gamma mixed model defined features field probability distribution, wherein multiplying for scale parameter and distributional pattern parameter is distributed in mixed model
Product is mean value.
Step S140 calculates the probability p (z | w) of above-mentioned image.
Assuming that each pixel characteristic value is mutual indepedent in image, defining above-mentioned probability is：
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Step S150 calculates the distribution function p (w) of above-mentioned double weight w.
Specifically, based on error sum of squares theory, center pixel and its neighborhood territory pixel otherness are defined in conjunction with MRF,
And then the distribution function p (w) for obtaining double weight w is：
Wherein, A is normalization coefficient, control cluster scale；η is neighborhood function coefficient, characterizes neighbourhood effect intensity；NiFor
With (xi,yi) centered on 8 neighborhood territory pixel set, and meet i ' ∈ Ni,i’≠i。
Step S160, calculates merit functions L, and merit functions L is that probability p (z | w) combines generally with distribution function p (w)
The logarithmic function of rate distribution function.
Wherein, merit functions L is calculated to specifically include：
Calculate joint probability distribution function ps (z, w) of the probability p (z | w) with distribution function p (w)：
P (z, w)=p (z | w) p (w)；
Merit functions L is defined as the logarithmic function of p (z, w), i.e., takes logarithm to above formula：
L (w)=log p (z, w)=log p (z | w)+log p (w).
Step S170 updates above-mentioned double weight w according to gradient method.
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)For gradient, which is defined as the derivative of above-mentioned merit functions L
It is as follows：
Updated double weight w are substituted into merit functions L by step S180.
Step S190 repeats above-mentioned steps S130-S180, until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop
Above-mentioned steps are only executed, the classification in image belonging to each pixel is determined according to current double weight w.
Determine that the classification belonging to above-mentioned each pixel specifically includes following steps：
(1) maximum value of above-mentioned double weights is calculated：
ci=arg max { wil, l=1 ..., k }；
Wherein, ciFor the label of ith pixel generic；
(2) using the maximum value as the classification belonging to the pixel.
By the definition of double weights in above-mentioned steps S120 it is found that it is the relational matrix of pixel and generic, in every a line most
It is pixel generic to be worth corresponding row greatly.
Step S200 exports segmentation result according to the classification belonging to each pixel.
It specifically, will be in same class after determining which classification is each pixel of image to be split belong to through the above steps
The intensity (i.e. the gray value of pixel) of all pixels is averaged, and segmentation result is obtained using the mean value as such intensity value.
The intensity of all pixels in the same category is updated with the mean value of the intensity of all pixels in same category in above-mentioned image,
Obtain mean value image；The mean value image is shown by display device.
In the method actual implementation of the present embodiment, before executing above-mentioned steps after inputting image to be split, further include
The step of constant is arranged, which specifically includes：Class number k, Gamma distributional pattern parameter alpha, normaliztion constant A and neighborhood effect
Coefficient η.
It is provided in an embodiment of the present invention to regard SAR image segmentation method more, by using the double weights indicated based on generic field
Gamma mixed models model characteristics of image field；In order to introduce neighborhood relationships, it is based on Markov random field model
Center pixel generic weight is considered as mean value in conjunction with error sum of squares theory, constructs it by (Markov Random Filed, MRF)
Sum of squares function with neighborhood territory pixel generic weight difference is characterized with describing the otherness of pixel in neighborhood window with error sum of squares
Difference can make full use of generic information, enhance the segmentation precision of image；Double weights in estimating Gamma mixed models simultaneously
When, the embodiment of the present invention solves double weights using gradient method.It is specific as follows：
(1) probability distribution that SAR image Characteristic Field is portrayed using Gamma mixed models characterizes pixel and class with generic field
Relationship between category, the relationship are indicated with matrix, and using the generic matrix as double weights of Gamma mixed models, from pixel and are gathered
Two angle collective effects of class in Gamma be distributed, compared in conventional hybrid model with vector indicate only with the Dan Quan of cluster correlation
Beijing South Maxpower Technology Co. Ltd is enough preferably to portray Characteristic Field；
(2) it is based on error sum of squares and markov random file (Markov Random Field, MRF) is theoretical, with center
The quadratic sum of pixel and its 8 neighborhood territory pixel generic weight difference portrays feature difference degree in neighborhood window, and then defines double weights
Distribution function illustrate that pixel characteristic difference is bigger in window, i.e., its probability to belong to a different category is got over when quadratic sum is bigger
Greatly；
(3) joint probability distribution function for utilizing Bayes' theorem defined feature field and double weights, is made with its logarithmic function
For merit functions, and double weights are solved using gradient method, the corresponding best estimate of merit functions is minimized to obtain.
The effect of the embodiment of the present invention can be further illustrated by following emulation experiment：
(1) emulation experiment condition
The present embodiment uses on the 7 flagship edition systems of Windows that CPU is Core (TM) i5-3470 3.20GHz
Emulation is realized in MATLAB 2011a software programmings.
It is simulation SAR image to emulate data 1, including 3 homogeneous regions, are by standard form image addition morphological parameters
4, scale parameter is respectively that 2,10,20 Gamma distribution random numbers obtain.Emulation data 2 are true SAR images, including 2 same
Freehandhand-drawing template is considered as standard form by matter region since true remote sensing images are without standard form.The above emulation data image
Size is 128 × 128, image total pixel number n=16384.
(2) the simulation experiment result
In order to prove the validity of the present embodiment algorithm, respectively using template image as standard, to the present embodiment algorithm and right
Confusion matrix is generated than algorithm segmentation result, and calculates its user's precision, Product Precision and Kappa values, with to the present embodiment side
Method carries out quantitative analysis (as shown in table 1), and wherein "-" indicates in image without this region.As shown in Table 1, total essence of the present embodiment
Degree and Kappa values are above comparison algorithm, and the validity of the present embodiment method is accurately demonstrated from digital angle.The present embodiment
Method noiseproof feature it is preferable, segmentation result accidentally divides phenomenon less, and partitioning boundary fitting is accurate；And compare algorithm cannot be effective gram
Complex Noise is taken, accidentally divides pixel more in segmentation result, causes visual effect very poor.
Table 1
Embodiment 2
In conjunction with previous embodiment, present embodiments provides one kind and regarding SAR image segmenting device more, shown in Figure 2 is more
Depending on the structural schematic diagram of SAR image segmenting device, which includes：Read module 301, double weights initialisation modules 302, weight
Parameter iteration update module 303, category determination module 304 and display module 305.
It is specifically described as follows：
Read module 301, it is to be split mostly regarding SAR image for reading.
This is to be split mostly to be indicated depending on SAR image using Characteristic Field z：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice point position
It sets, D is image area.
Double weights initialisation modules 302, for initializing double weight w.
In the present embodiment, double weight w are defined and indicate relationship between pixel and generic, it is specific as follows：
wi=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number.Wherein, initial double
Weight generates at random, and row indicates that pixel, row indicate classification, and element value range is 0~1 in matrix, and meet per a line and
It is 1.
Weight parameter iteration update module 303, for repeating following calculating：
Calculate Gamma distribution scale parameters β.
Define Gamma distribution scale parameters βlFor about generic weight wilFunction, it is specific as follows：
Wherein, α is Gamma distributional pattern parameters, and for regarding SAR image, α is equal to it and regards number more.Based on double weights, in conjunction with
Gamma mixed model defined features field probability distribution, wherein multiplying for scale parameter and distributional pattern parameter is distributed in mixed model
Product is mean value.
Calculate the probability p (z | w) of image.
Assuming that each pixel characteristic value is mutual indepedent in image, defining above-mentioned probability is：
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Calculate the distribution function p (w) of above-mentioned double weight w.
Specifically, based on error sum of squares theory, center pixel and its neighborhood territory pixel otherness are defined in conjunction with MRF,
And then the distribution function p (w) for obtaining double weight w is：
Wherein, A is normalization coefficient, control cluster scale；η is neighborhood function coefficient, characterizes neighbourhood effect intensity；NiFor
With (xi,yi) centered on 8 neighborhood territory pixel set, and meet i ' ∈ Ni,i’≠i。
Calculate merit functions L, above-mentioned merit functions L is probability p (z | w) and distribution function p (w) joint probability distribution letters
Several logarithmic functions.
Above-mentioned weight parameter iteration update module 303 includes：First computing unit, for calculate probability p (z | w) with point
The joint probability distribution function p (z, w) of cloth function p (w)：
P (z, w)=p (z | w) p (w)；
Second computing unit, for taking logarithm to above formula：
L (w)=log p (z, w)=log p (z | w)+log p (w).
Above-mentioned double weight w are updated according to gradient method.
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)For gradient, which is defined as the derivative of above-mentioned merit functions L
It is as follows：
Updated double weight w are substituted into merit functions L；
Until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop executing above-mentioned steps.
Category determination module 304, for determining the classification in image belonging to each pixel according to current double weight w.
Above-mentioned category determination module 304 includes：Maximum value calculation unit, the maximum value for calculating double weights：
ci=arg max { wil, l=1 ..., k }
Wherein, ciFor the label of ith pixel generic；
Classification determination unit, for using above-mentioned maximum value as the classification belonging to pixel.
By the definition of above-mentioned double weights it is found that it is the relational matrix of pixel and generic, maximum value is corresponding in every a line
Row are pixel generic.
Output module 305, for exporting segmentation result according to the classification belonging to each pixel.
Specifically, above-mentioned output module 305 includes：
The mean value of equal value cell, the intensity for all pixels in same category in image updates in the same category
The intensity of all pixels obtains mean value image；
Display unit shows the mean value image for passing through display device.
In the device actual implementation of the present embodiment, further include：Constant setup module, for constant, above-mentioned constant to be arranged
Including：Class number k, Gamma distributional pattern parameter alpha, normaliztion constant A and neighborhood function coefficient η.
What the present embodiment was provided regards the realization principle of SAR image segmenting device and technique effect and the aforementioned reality of generation more
It is identical to apply example, to briefly describe, the present embodiment part does not refer to place, can refer to corresponding contents in previous embodiment.
It, can be with if above-mentioned function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer read/write memory medium.Based on this understanding, technical scheme of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be expressed in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be
People's computer, server or network equipment etc.) execute all or part of step of each embodiment method of the present invention.And it is preceding
The storage medium stated includes：USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory
The various media that can store program code such as (RAM, Random Access Memory), magnetic disc or CD.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be subject to the protection scope in claims.
Claims (10)
1. a kind of regarding SAR image segmentation method more, which is characterized in that including：
It reads to be split mostly regarding SAR image；It is described to be split mostly to be indicated using Characteristic Field z depending on SAR image：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice site, D
For image area；
The double weight w of initialization：
W=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number；
Repeat following step：Calculate Gamma distribution scale parameters βl：
Wherein, α is Gamma distributional pattern parameters；
Calculate the probability p (z | w) of described image；
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Wherein Ga (zi|βl) it is with ziFor independent variable, βlFor the Gamma distribution probability density functions of scale parameter, Γ (α) is with α
For the Gamma functions of parameter；
Calculate the distribution function p (w) of double weight w：
Wherein, A is normalization coefficient, and η is neighborhood function coefficient, NiFor with (xi,yi) centered on 8 neighborhood territory pixel set, and it is full
Sufficient i ' ∈ Ni,i'≠i；
Calculate merit functions L, the merit functions L is the probability p (z | w) and distribution function p (w) joint probability point
The logarithmic function of cloth function；
Double weight w are updated according to gradient method：
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)It is as follows for gradient：
Updated double weight w are substituted into the merit functions L；
Until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop executing above-mentioned steps, it is true according to current double weight w
Determine the classification belonging to each pixel in described image；
Segmentation result is exported according to the classification belonging to each pixel.
2. according to the method described in claim 1, it is characterized in that, calculating merit functions L includes：
Calculate joint probability distribution function ps (z, w) of the probability p (z | w) with distribution function p (w)：
P (z, w)=p (z | w) p (w)；
Logarithm is taken to the joint probability distribution function：
L (w)=logp (z, w)=logp (z | w)+logp (w).
3. according to the method described in claim 1, it is characterized in that, determining that the classification belonging to the pixel includes：
Calculate the maximum value of double weights：
ci=arg max { wil, l=1 ..., k }；
Wherein, ciFor the label of ith pixel generic；
Using the maximum value as the classification belonging to the pixel.
4. according to the method described in claim 1, it is characterized in that, according to the classification output segmentation knot belonging to each pixel
Fruit includes：All pictures in the same category are updated with the mean value of the intensity of all pixels in same category in described image
The intensity of element, obtains mean value image；The mean value image is shown by display device.
5. according to the method described in claim 1, it is characterized in that, further including：
Constant is set, and the constant includes：Class number k, Gamma distributional pattern parameter alpha, normaliztion constant A and neighborhood function coefficient
η。
6. a kind of regarding SAR image segmenting device more, which is characterized in that including：
Read module, it is to be split mostly regarding SAR image for reading；It is described to be split mostly to be indicated using Characteristic Field z depending on SAR image：
Z={ zi(xi,yi):I=1 ..., n }
Wherein, i is pixel index, and n is total pixel number, ziFor the intensity of pixel i, (xi, yi) ∈ D be pixel i lattice site, D
For image area；
Double weights initialisation modules, for initializing double weight w：
W=(wil:L=1 ..., k)
Wherein, wilInclude the generic weight of pixel i for generic l, meetsK is class number；
Weight parameter iteration update module, for repeating following step：Calculate Gamma distribution scale parameters βl：
Wherein, α is Gamma distributional pattern parameters；
Calculate the probability p (z | w) of described image；
Wherein p (zi|wil) it is the z defined with double weight Gamma mixed modelsiProbability density, it is as follows：
Wherein Ga (zi|βl) it is with ziFor independent variable, βlFor the Gamma distribution probability density functions of scale parameter, Γ (α) is with α
For the Gamma functions of parameter；
Calculate the distribution function p (w) of double weight w：
Wherein, A is normalization coefficient, and η is neighborhood function coefficient, NiFor with (xi,yi) centered on 8 neighborhood territory pixel set, and it is full
Sufficient i ' ∈ Ni,i'≠i；
Calculate merit functions L, the merit functions L is the probability p (z | w) and distribution function p (w) joint probability point
The logarithmic function of cloth function；
Double weight w are updated according to gradient method：
w(t+1)=w(t)+ξΔw(t)
Wherein, t is iterations, and ξ is step-length, Δ w(t)It is as follows for gradient：
Updated double weight w are substituted into the merit functions L；
Until | L(t+1)-L(t)| when being less than preset threshold epsilon, stop executing above-mentioned steps；
Category determination module, for determining the classification in described image belonging to each pixel according to current double weight w；
Output module, for exporting segmentation result according to the classification belonging to each pixel.
7. device according to claim 6, which is characterized in that the weight parameter iteration update module includes：
First computing unit, the joint probability distribution function p (z, w) for calculating probability p (z | w) and distribution function p (w)：
P (z, w)=p (z | w) p (w)；
Second computing unit, for taking logarithm to the joint probability distribution function：
L (w)=logp (z, w)=logp (z | w)+logp (w).
8. device according to claim 6, which is characterized in that the category determination module includes：
Maximum value calculation unit, the maximum value for calculating double weights：
ci=arg max { wil, l=1 ..., k }；
Wherein, ciFor the label of ith pixel generic；
Classification determination unit, for using the maximum value as the classification belonging to the pixel.
9. device according to claim 6, which is characterized in that the output module includes：
The mean value of equal value cell, the intensity for all pixels in same category in described image updates in the same category
All pixels intensity, obtain mean value image；
Display unit shows the mean value image for passing through display device.
10. device according to claim 6, which is characterized in that further include：
Constant setup module, for constant to be arranged, the constant includes：Class number k, Gamma distributional pattern parameter alpha, normalization are normal
Number A and neighborhood function coefficient η.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610833053.0A CN106408587B (en) | 2016-09-19 | 2016-09-19 | Regard SAR image segmentation method and device more |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610833053.0A CN106408587B (en) | 2016-09-19 | 2016-09-19 | Regard SAR image segmentation method and device more |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106408587A CN106408587A (en) | 2017-02-15 |
| CN106408587B true CN106408587B (en) | 2018-11-09 |
Family
ID=57996834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610833053.0A Expired - Fee Related CN106408587B (en) | 2016-09-19 | 2016-09-19 | Regard SAR image segmentation method and device more |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106408587B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711016B (en) * | 2018-12-14 | 2020-09-04 | 中国地质大学（武汉） | Method and system for simulating imaging form and dynamically simulating two-sided SAR (synthetic Aperture Radar) satellite |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103426175A (en) * | 2013-08-23 | 2013-12-04 | 西安电子科技大学 | Polarization SAR image segmentation method based on characteristic value measurement spectral clustering |
| CN104598922A (en) * | 2015-01-07 | 2015-05-06 | 河海大学 | Completely-polarized SAR (synthetic aperture radar) image classification method based on fuzzy c-means |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6646593B1 (en) * | 2002-01-08 | 2003-11-11 | Science Applications International Corporation | Process for mapping multiple-bounce ghosting artifacts from radar imaging data |
-
2016
- 2016-09-19 CN CN201610833053.0A patent/CN106408587B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103426175A (en) * | 2013-08-23 | 2013-12-04 | 西安电子科技大学 | Polarization SAR image segmentation method based on characteristic value measurement spectral clustering |
| CN104598922A (en) * | 2015-01-07 | 2015-05-06 | 河海大学 | Completely-polarized SAR (synthetic aperture radar) image classification method based on fuzzy c-means |
Non-Patent Citations (1)
| Title |
|---|
| 基于Voronoi几何划分和EM/MPM算法的多视SAR图像分割;赵泉华 等;《遥感学报》;20130430;第17卷(第4期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106408587A (en) | 2017-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107194872B (en) | Remote sensed image super-resolution reconstruction method based on perception of content deep learning network | |
| CN106485269B (en) | SAR image object detection method based on mixing statistical distribution and multi-part model | |
| CN106991686A (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
| CN104063876A (en) | Interactive image segmentation method | |
| CN110569782A (en) | Target detection method based on deep learning | |
| CN110610143B (en) | Crowd counting network method, system, medium and terminal for multi-task combined training | |
| Li et al. | No-reference image quality assessment using statistical characterization in the shearlet domain | |
| CN105787935A (en) | Fuzzy cluster SAR image segmentation method based on Gamma distribution | |
| CN103955709B (en) | Weighted synthetic kernel and triple markov field (TMF) based polarimetric synthetic aperture radar (SAR) image classification method | |
| CN104240264A (en) | Height detection method and device for moving object | |
| CN110189292A (en) | A kind of cancer cell detection method based on Faster R-CNN and density estimation | |
| US11403807B2 (en) | Learning hybrid (surface-based and volume-based) shape representation | |
| CN113160392B (en) | Optical building target three-dimensional reconstruction method based on deep neural network | |
| CN104346814B (en) | Based on the SAR image segmentation method that level vision is semantic | |
| Limper et al. | Mesh Saliency Analysis via Local Curvature Entropy. | |
| CN109727226A (en) | A kind of position table automatic generation method based on machine learning | |
| CN108564083A (en) | A kind of method for detecting change of remote sensing image and device | |
| CN105678790B (en) | High-resolution remote sensing image supervised segmentation method based on variable gauss hybrid models | |
| CN105787505A (en) | Infrared image clustering segmentation method combining sparse coding and spatial constraints | |
| CN106408587B (en) | Regard SAR image segmentation method and device more | |
| CN104239892A (en) | SAR image mixed model fitting method based on KSVD training dictionary | |
| CN106846377A (en) | A kind of target tracking algorism extracted based on color attribute and active features | |
| CN103584888B (en) | Ultrasonic target motion method for tracing | |
| CN109523590A (en) | A kind of 3D rendering depth information visual comfort appraisal procedure based on sample | |
| CN108401564B (en) | Conspicuousness object detection method based on super-pixel segmentation and image abstraction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181109 Termination date: 20210919 |