CN109446894A - The multispectral image change detecting method clustered based on probabilistic segmentation and Gaussian Mixture - Google Patents
The multispectral image change detecting method clustered based on probabilistic segmentation and Gaussian Mixture Download PDFInfo
- Publication number
- CN109446894A CN109446894A CN201811087966.8A CN201811087966A CN109446894A CN 109446894 A CN109446894 A CN 109446894A CN 201811087966 A CN201811087966 A CN 201811087966A CN 109446894 A CN109446894 A CN 109446894A
- Authority
- CN
- China
- Prior art keywords
- image
- gaussian mixture
- pixel
- algorithm
- probability
- 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
- 230000008859 change Effects 0.000 title claims abstract description 69
- 239000000203 mixture Substances 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000011218 segmentation Effects 0.000 title claims abstract description 33
- 238000001514 detection method Methods 0.000 claims abstract description 64
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 9
- 230000003595 spectral effect Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000004458 analytical method Methods 0.000 claims description 7
- 230000007547 defect Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 abstract description 2
- 238000004088 simulation Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012163 sequencing technique Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the method or apparatus technical fields that applying electronic equipment is identified, disclose a kind of multispectral image change detecting method clustered based on probabilistic segmentation and Gaussian Mixture；Firstly, two original multispectral images of the input with region different time, construct mix variance image HDS using CVA and SAM；Differential image is mapped to super-pixel space secondly, merging algorithm using statistical regions and carrying out multi-scale division to differential image；Finally, the shortcomings that overcoming it easily to converge on locally optimal solution using K-means algorithm initialization gauss hybrid models, is fitted the probability statistical distribution of super-pixel feature space, variation testing result is obtained using the Bayesian Decision rule based on minimal error rate.The present invention preferably utilizes the amplitude change information and angle change information of spectrum vector, and the partial structurtes feature of image can be obtained using super-pixel segmentation, and the detection for effectively increasing region of variation in SAR image is accurate.
Description
Technical Field
The invention belongs to the technical field of methods or devices for identification by using electronic equipment, and particularly relates to a multispectral image change detection method based on probability segmentation and Gaussian mixture clustering.
Background
Currently, the current state of the art commonly used in the industry is such that: with the rapid development of remote sensing technology in recent years, the quantity of remote sensing data is increasing day by day, and the remote sensing data is widely used in the fields of environmental monitoring, atmospheric analysis, urban planning and the like. The change detection of the multispectral remote sensing image is applied to the fields of military affairs and civilian use, and mainly relates to positioning of natural disaster areas such as flood, fire and earthquake, urban expansion condition analysis and evaluation of striking effect in military application, so that the change detection of the multispectral remote sensing image has important significance in research. The change detection of the multispectral remote sensing image is to determine whether the ground feature changes according to the difference of the spectral characteristics or the spatial structure of the image. The general remote sensing image change detection process comprises two steps: 1) constructing a difference image; 2) and analyzing the difference image to obtain a detection result. The method is essentially a two-classification problem, namely that all pixel points of the difference image are classified into a variation class and a non-variation class. In the past decade, a large number of change detection methods have been proposed, which can be classified into a pixel-based method and an object-based method, depending on the basic unit of image analysis. The pixel-based method generally uses the gray features of the difference pixels for classification, and uses a clustering method to obtain a classification result as good as possible. However, as the resolution of the remote sensing image is continuously improved, the spatial cross correlation between the pixel and the neighborhood of the pixel is stronger and stronger, and a single pixel feature does not have a semantic feature, so that the semantic gap problem exists, so that the object-based analysis method, namely, the image is divided into homogeneous regions to obtain semi-semantic information, becomes a key technology for image understanding and identification. The object-based change detection method takes superpixels as basic units of analysis and contains rich information, such as: spectrum, texture, shape, spatial context, etc. For example, an SAR image change detection method based on superpixel segmentation and multi-method fusion firstly introduces an SLIC superpixel segmentation method, and obtains a superpixel segmentation result conforming to the actual ground object boundary by performing joint segmentation on a main image and an auxiliary image; meanwhile, 3 pixel-based change detection methods are used for obtaining an initial change detection result; and then, performing mode voting of two levels by using the super-pixel segmentation result and the initial change detection result, and removing false alarms caused by noise in the detection result and holes in a connected domain to obtain a final change detection result. However, in the method, original double-temporal remote sensing images are respectively segmented, and then distance measurement values between super pixels are calculated, and due to the difference of the shape and the structure of a target between the double-temporal images, a boundary needs to be registered before calculating a similar distance, so that the geometric information of the target is damaged.
In order to solve the problems, the invention adopts a multispectral image change detection method based on probability segmentation and Gaussian mixture clustering. In order to solve the problems that a change detection method based on a single pixel does not have semantic information and is high in calculation complexity, the image is mapped to a superpixel space from a pixel space by adopting superpixel segmentation, and regional characteristics of a change target can be better reflected; meanwhile, the Gaussian mixture model is adopted to cluster the superpixels, so that the purposes of eliminating the edge effect and improving the detection precision can be achieved, and a clearer physical significance is provided for solving the problem of change detection.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a multispectral image change detection method based on probability segmentation and Gaussian mixture clustering.
The invention is realized by the following steps: firstly, inputting two original multispectral images in the same region and at different time, and constructing a mixed difference image HDS by using a CVA and an SAM; secondly, performing multi-scale segmentation on the difference image by utilizing a statistical region merging algorithm of a dynamic sequencing mode to map the difference image to a super-pixel space; and finally, initializing a Gaussian mixture model by adopting a K-means algorithm to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, fitting the probability statistical distribution of the super-pixel feature space, and obtaining a change detection result by utilizing a Bayes judgment rule based on the minimum error rate.
Further, the multispectral image change detection method based on probability segmentation and Gaussian mixture clustering specifically comprises the following steps:
(1) inputting two multispectral remote sensing images X in different time same regions1And X2；
(2) To X1And X2The two images respectively obtain the angle change mapping theta and the change vector CV of the two images by using SAM and CVA, and a new mixed difference space HDS is constructed by combining the two change characteristics;
(3) merging the pixel points in the mixed difference image DI by using a statistical region merging algorithm to complete the conversion of the difference image from a pixel space to a super pixel space to obtain a merged image DT;
(4) the pixel mean values of each super pixel in the statistical merged image DT form a set X ═ Xn1 ≦ N, where N represents the total number of superpixels;
(5) the mixed probability model classification adopts a K-means algorithm to initialize a Gaussian mixture model to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, and the probability statistical distribution of the super-pixel feature space is fitted:
(6) and obtaining a final detection result of each super pixel by adopting a Bayes discrimination rule based on a minimum error rate, wherein the posterior probability is calculated as follows according to a Bayes formula:
then the bayesian decision flow rule based on the minimum error rate is as follows:
if it isX is thennIs determined to be changed;
if it isX is thennIs determined to be unchanged.
Further, the (5) specifically includes:
a) all data elements in the set are assumed to have a gaussian mixture distribution with two models:
wherein,the parameters to be solved are: mean and variance of Gaussian functions of varying and non-varying classesAnd α1Or α2；
b) Solving parameter values by adopting an EM algorithm: defining the number of components k to 2, and setting a parameter mu for each component kk、σkAnd αkWherein a k-means algorithm is used to calculate the cluster center as μkThe initial value of the parameter is used for avoiding the problem that the EM algorithm is easy to be trapped in a local optimal solution;
c) the EM algorithm finds a rough value of the parameter to be estimated:
according to the current muk，σkAnd αkCalculating the posterior probability gamma (z)nk)：
d) The EM algorithm maximizes the likelihood function using the values of the first step:
according to gamma (z)nk) Calculating model parameter mu of new iterationk (t+1)，σk (t+1)，αk (t+1)：
e) Calculating a log-likelihood function of the Gaussian mixture model:
f) judging whether the likelihood function is converged: if converging, then the parameters are outputAnd α1And α2(ii) a If not, returning to the step c) to execute until meeting the convergence condition.
Another objective of the present invention is to provide a spectral image processing system applying the multispectral image change detection method based on probability segmentation and gaussian mixture clustering.
Another objective of the present invention is to provide an environment monitoring spectral image processing system applying the multispectral image change detection method based on probability segmentation and gaussian mixture clustering.
Another objective of the present invention is to provide an atmospheric analysis spectral image processing system applying the multispectral image change detection method based on probability segmentation and gaussian mixture clustering.
The invention further aims to provide a city planning spectral image processing system applying the multispectral image change detection method based on probability segmentation and Gaussian mixture clustering.
In summary, the advantages and positive effects of the invention are: in the process of generating the difference image, a mixed difference image is constructed by combining the CVA and the SAM; compared with the prior art, the method for generating the difference image by using a difference method or only using CVA and the like is adopted; amplitude change information and angle change information of the multispectral image can be extracted simultaneously, and the precision of the difference image is improved;
the mixed difference image is subjected to multi-scale segmentation by adopting a statistical region merging algorithm, the similarity between adjacent pixels and the local structural characteristics of the image are fully utilized, and the difference image is mapped to a super-pixel space from the pixel space, so that a low-dimensional semi-semantic clustering space for change detection is obtained, and the calculation complexity is reduced;
according to the method, the Gaussian mixture model is adopted to perform spatial clustering on the superpixels so as to obtain a final change detection result, and aiming at the problem of initial value sensitivity of the Gaussian mixture model algorithm, the Gaussian mixture model initialized based on K-means is introduced to avoid a local optimal phenomenon and improve the convergence speed of the model. Compared with the classical clustering methods such as K-means and FCM, the Gaussian mixture model can achieve the purposes of eliminating the edge effect and improving the detection precision.
Drawings
Fig. 1 is a flowchart of a multispectral image change detection method based on probability segmentation and gaussian mixture clustering according to an embodiment of the present invention.
Fig. 2 is a flowchart of an implementation of the multispectral image change detection method based on probability segmentation and gaussian mixture clustering according to the embodiment of the present invention.
Fig. 3 is a diagram illustrating the results of detecting changes in the seian river channel data set according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating the result of performing change detection on a west amp dam building data set according to an embodiment of the present invention.
FIG. 5 is a diagram of a conventional M provided by an embodiment of the present invention2C2And a change detection result graph of the VA algorithm and the AFS algorithm on the Xian river channel data set.
FIG. 6 is a diagram of a conventional M provided by an embodiment of the present invention2C2And (3) detecting a result graph of the change of the data set of the west-safe dam building by using the VA algorithm and the AFS algorithm.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problem of low detection accuracy of a change region in the existing multispectral image; the invention provides a multispectral image change detection method based on probability segmentation and Gaussian mixture clustering, which is used for realizing detection of a change area in a multispectral image and improving the detection accuracy.
The following detailed description of the principles of the invention is provided in connection with the accompanying drawings.
As shown in fig. 1, the multispectral image change detection method based on probability segmentation and gaussian mixture clustering provided by the embodiment of the present invention includes the following steps:
s101: the capabilities that the CVA and the SAM can describe difference information from two different angles of the size and the direction of two spectral vectors are utilized, and a mixed difference image is constructed by combining the two methods;
s102: carrying out multi-scale segmentation on the difference image by utilizing a statistical region merging algorithm of a dynamic sequencing mode to map the difference image to a super-pixel space;
s103: a Gaussian mixture model is initialized by adopting a K-means algorithm to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, probability statistical distribution of a super-pixel feature space is fitted, and a change detection result is obtained by utilizing a Bayes judgment rule based on the minimum error rate.
The multispectral image change detection method based on probability segmentation and Gaussian mixture clustering provided by the embodiment of the invention specifically comprises the following steps:
(1) inputting two multispectral remote sensing images X in different time same regions1And X2；
(2) To X1And X2The two images respectively obtain the angle change mapping theta and the change loss CV of the two images by using SAM and CVA, and a new mixed difference space HDS is constructed by combining the two change characteristics;
(3) merging the pixel points in the mixed difference image DI by using a statistical region merging algorithm to complete the conversion of the difference image from a pixel space to a super pixel space to obtain a merged image DT;
(4) the pixel mean values of each super pixel in the statistical merged image DT form a set X ═ Xn1 ≦ N, where N represents the total number of superpixels;
(5) the mixed probability model classification adopts a K-means algorithm to initialize a Gaussian mixture model to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, and the probability statistical distribution of the super-pixel feature space is fitted:
5a) all data elements in the set are assumed to have a gaussian mixture distribution with two models:
wherein,the parameters to be solved are: mean and variance of Gaussian functions of varying and non-varying classesAnd α1Or α2；
5b) Solving parameter values by adopting an EM algorithm: defining the number of components k to 2, and setting a parameter mu for each component kk、σkAnd αkWherein a k-means algorithm is used to calculate the cluster center as μkThe initial value of the parameter is used for avoiding the problem that the EM algorithm is easy to be trapped in a local optimal solution;
5c) the first step of the EM algorithm, finds the rough values of the parameters to be estimated:
according to the current muk，σkAnd αkCalculating the posterior probability gamma (z)nk)：
5d) The second step of the EM algorithm, using the values of the first step to maximize the likelihood function:
according to gamma (z)nk) Calculating model parameter mu of new iterationk (t+1)，σk (t+1)，αk (t+1)：
5e) Calculating a log-likelihood function of the Gaussian mixture model:
5f) judging whether the likelihood function is converged: if converging, then the parameters are outputAnd α1And α2(ii) a If not, returning to the step 5c) to execute until meeting the convergence condition.
(6) And obtaining a final detection result of each super pixel by adopting a Bayesian discrimination rule based on the minimum error rate. According to the Bayesian formula, the posterior probability is calculated as follows:
then the bayesian decision flow rule based on the minimum error rate is as follows:
if it isX is thennIs determined to be changed;
if it isX is thennIs determined to be unchanged.
The application of the principles of the present invention will now be described in further detail with reference to the accompanying drawings.
As shown in fig. 2, the multispectral image change detection method based on probability segmentation and gaussian mixture clustering provided by the embodiment of the present invention specifically includes the following steps:
step one, inputting two multispectral remote sensing images X in different time same regions1And X2；
Step two, for X1And X2The two images respectively obtain the angle mapping theta and the variation loss CV by using SAM and CVA, and a new mixed difference space HDS is constructed by combining two variation characteristics:
2a) order S1＝(x1 1，x1 2，...x1 L) And S2＝(x2 1，x2 2，...x2 L) Representing a set of multi-spectral images X1And X2The intermediate coordinate is a spectral vector of (i, j), and the multispectral image contains four bands of red, green, blue and near infrared, so that L is 4, and the angle mapping θ is solved:
wherein,representing a spectral vector S1The b-th component.
2b) Calculating the change vector CV:
2c) the mixed difference vector HDS is constructed by combining two variation features:
wherein, theta*α θ (α max (cv)/k), k being a constant, k being set to 45 herein.
2d) Obtaining a mixed difference image DI by one-dimensional mapping the mixed difference space:
DI＝||HDS||2；
combining pixel points in the mixed difference image DI by using a statistical region combination algorithm to complete the conversion of the difference image from a pixel space to a region space to obtain a combined image DT;
3a) a similarity weight is calculated for each pair of pixels in the mixed difference image DI:
f(p,p')＝|p-p'|；
where p and p' represent the gray values of two adjacent pixels or the pixel mean of two adjacent regions.
3b) And sequencing the similarity weights in an ascending order from small to large, and sequentially selecting the pixel pairs according to the sequencing order to judge whether the pixel pairs are combined:
if it satisfiesThen the merging is carried out, wherein, represents the average pixel value of the R region. R|R|Representing a set of regions with R pixels,the method comprises the following steps that I is an empirical constant, the number of pixels contained in an image is represented by I, and Q represents the complexity of a real scene; otherwise, merging; when each group of pixel pairs completes the judgment process, the merged image DT is obtained.
Step four, the pixel mean value of each super pixel in the statistical merging image DT forms a set X ═ XnAnd |1 is less than or equal to N is less than or equal to N, and N represents the total number of the superpixels. The calculation of the mean pixel value of the superpixel comprises the following steps: and for each super pixel, calculating the sum of pixel values of all pixel points contained in the region of the super pixel, and then dividing the sum by the number of the pixel points to obtain the pixel mean value of the super pixel.
Step five, fitting the probability statistical distribution of the super-pixel feature space by adopting a Gaussian mixture model:
5a) all data elements in the set are assumed to have a gaussian mixture distribution with two models, namely:
wherein,the parameters to be solved are: mean and variance of Gaussian functions of varying and non-varying classesAnd α1And α2；
5b) Solving the parameter values using the EM algorithm:
defining the number of components k to 2, and setting a parameter mu for each component kk、σkAnd αkWherein a k-means algorithm is used to calculate the cluster center as μkInitial values of the parameters to avoid the problem of the EM algorithm being prone to local optimal solution.
The k-means algorithm accepts the parameter k 2, performs clustering with k points in space as the center, and classifies the objects closest to them. And (4) gradually updating the value of each clustering center through an iterative method until the best clustering result is obtained. Assuming that the sample set is to be divided into k classes, the algorithm is described as follows: (1) properly selecting initial centers (2) of k classes, in the ith iteration, calculating the distances from any sample to the k centers, updating the center values (4) of the classes by using methods such as mean values and the like for the class (3) of which the sample is classified to the center with the shortest distance, and ending the iteration if the values are kept unchanged after the updating by using the iteration methods of (2) and (3), otherwise, continuing the iteration. E.g. the set of raw data is (x)1,x2,…,xn) And each xiFor d-dimensional vectors, the purpose of k-means clustering is to separate the raw data into k classes S { S ═ S given the number of classification groups k, where k ≦ n1,S2,…,SkIts objective function is as follows:
5c) the first step of the EM algorithm, finds the rough values of the parameters to be estimated:
according to the current muk，σkAnd αkCalculating the posterior probability gamma (z)nk)：
5d) The second step of the EM algorithm, using the values of the first step to maximize the likelihood function:
according to gamma (z)nk) Calculating model parameter mu of new iterationk (t+1)，σk (t+1)，αk (t+1)：
5e) Calculating a log-likelihood function of the Gaussian mixture model:
5f) judging whether the likelihood function is converged: if converging, then the parameters are outputAnd α1And α2(ii) a If not, returning to the step 5c) to execute until meeting the convergence condition.
And 6, obtaining a final detection result of each super pixel by adopting a Bayes judgment rule based on the minimum error rate. According to the Bayesian formula, the posterior probability is calculated as follows:
then the bayesian decision flow rule based on the minimum error rate is as follows:
if it isX is thennIs determined to be changed;
if it isX is thennIs determined to be unchanged.
The application effect of the present invention will be described in detail with reference to simulation experiments.
1. Simulation conditions
The invention uses MATLAB R2016a to complete the simulation experiment of the invention on a PC with a CPU of Intel (R) Core i5-4590, CPU3.30GHz, RAM 8.00GB and Windows 7 operating system.
2. Content of simulation experiment
In the experiment, two data sets of the Xian river channel and the Xian dam building are selected for verification of change detection results, the size of the image of the Xian river channel is 301 multiplied by 4, and the size of the image of the data set of the Xian 27984 dam building is 134 multiplied by 132 multiplied by 4, wherein the data set of the Xian dam building comprises 4 wave bands of blue, green, red and near infrared.
Simulation 1, the method of the invention is adopted to perform change detection on the data sets of the canal of west ampere and the data sets of the west ampere dam building, and the detection results are shown in fig. 3 and 4, wherein:
(1) FIG. 3(a) shows a pre-change image in the canal dataset of Xian ;
(2) FIG. 3(b) shows a post-alteration image of the canal data set of Xian ;
(3) FIG. 3(c) is a graph showing a standard reference variation in the canal data set of Xian ;
(4) FIG. 3(d) is a graph showing the results of multispectral image change detection using the present invention;
(5) FIG. 4(a) shows a pre-change image in a Seaman dam building data set;
(6) FIG. 4(b) shows a post-change image in a Seisan dam building data set;
(7) FIG. 4(c) shows a standard reference variation in the Seaman dam building data set;
(8) FIG. 4(d) is a graph showing the results of multispectral image change detection using the present invention;
simulation 2, using M2C2The VA algorithm, the AFS algorithm and the method of the present invention perform change detection on two data sets of the west ampere river channel and the west ampere dam building, and the results are shown in fig. 5 and fig. 6, wherein:
(1) FIG. 5(a) shows that the Seaman river channel data set employs M2C2Detecting the result of the VA algorithm;
(2) FIG. 5(b) shows the detection result of the river channel data set of Xian using the AFS algorithm;
(3) FIG. 5(c) shows the results of a detection of the Seaman river channel data set using the method of the present invention;
(4) FIG. 6(a) shows that the Seaman dam building data set employs M2C2Detecting the result of the VA algorithm;
(5) FIG. 6(b) shows the detection result of the Western Ann dam building data set using the AFS algorithm;
(6) FIG. 6(c) shows the results of a detection of the Seaman dam building data set using the method of the present invention.
3. Simulation experiment results and analysis
As can be seen from fig. 3, the method of the present invention can effectively detect the changed regions in the multispectral image. As can be seen from FIG. 4, in comparison to M2C2VA algorithm and AFS algorithm, the method of the invention can effectively improve the detection accuracy of the change area in the multispectral image. The method has the advantages of high detection precision, completeness maintenance of target geometric details and remarkable improvement of the performance of multispectral image change detection.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (7)
1. A multispectral image change detection method based on probability segmentation and Gaussian mixture clustering is characterized by comprising the following steps: firstly, inputting two original multispectral images in the same region and at different time, and constructing a mixed difference image HDS by using a CVA and an SAM; secondly, performing multi-scale segmentation on the difference image by using a statistical region merging algorithm to map the difference image to a super-pixel space; and finally, initializing a Gaussian mixture model by adopting a K-means algorithm to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, fitting the probability statistical distribution of the super-pixel feature space, and obtaining a change detection result by utilizing a Bayes judgment rule based on the minimum error rate.
2. The method according to claim 1, further comprising:
(1) inputting two multispectral remote sensing images X in different time same regions1And X2；
(2) To X1And X2The two images respectively obtain the angle change mapping theta and the change vector CV of the two images by using SAM and CVA, and a new mixed difference space HDS is constructed by combining the two change characteristics;
(3) merging the pixel points in the mixed difference image DI by using a statistical region merging algorithm to complete the conversion of the difference image from a pixel space to a super pixel space to obtain a merged image DT;
(4) the pixel mean values of each super pixel in the statistical merged image DT form a set X ═ Xn1 ≦ N, where N represents the total number of superpixels;
(5) the mixed probability model classification adopts a K-means algorithm to initialize a Gaussian mixture model to overcome the defect that the Gaussian mixture model is easy to converge on a local optimal solution, and the probability statistical distribution of the super-pixel feature space is fitted:
(6) and obtaining a final detection result of each super pixel by adopting a Bayes discrimination rule based on a minimum error rate, wherein the posterior probability is calculated as follows according to a Bayes formula:
then the bayesian decision flow rule based on the minimum error rate is as follows:
if it isX is thennIs determined to be changed;
if it isX is thennIs determined to be unchanged.
3. The method according to claim 2, wherein the step (5) comprises:
a) all data elements in the set are assumed to have a gaussian mixture distribution with two models:
wherein,the parameters to be solved are: mean and variance of Gaussian functions of varying and non-varying classesAnd α1Or α2；
b) Solving parameter values by adopting an EM algorithm: defining the number of components k to 2, and setting a parameter mu for each component kk、σkAnd αkWherein a k-means algorithm is used to calculate the cluster center as μkThe initial value of the parameter is used for avoiding the problem that the EM algorithm is easy to be trapped in a local optimal solution;
c) the EM algorithm finds a rough value of the parameter to be estimated:
according to the current muk，σkAnd αkCalculating the posterior probability gamma (z)nk)：
d) The EM algorithm maximizes the likelihood function using the values of the first step:
according to gamma (z)nk) Calculating model parameter mu of new iterationk (t+1)，σk (t+1)，αk (t+1)：
e) Calculating a log-likelihood function of the Gaussian mixture model:
f) judging whether the likelihood function is converged: if converging, then the parameters are outputAnd α1And α2(ii) a If not, returning to the step c) to execute until meeting the convergence condition.
4. A spectral image processing system using the method for detecting multispectral image change based on probability segmentation and gaussian mixture clustering according to any one of claims 1 to 3.
5. An environment monitoring spectral image processing system applying the multispectral image change detection method based on probability segmentation and Gaussian mixture clustering according to any one of claims 1-3.
6. An atmospheric analysis spectral image processing system applying the multispectral image change detection method based on probability segmentation and Gaussian mixture clustering according to any one of claims 1 to 3.
7. An urban planning spectral image processing system applying the multispectral image change detection method based on probability segmentation and Gaussian mixture clustering according to any one of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811087966.8A CN109446894B (en) | 2018-09-18 | 2018-09-18 | Multispectral image change detection method based on probability segmentation and Gaussian mixture clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811087966.8A CN109446894B (en) | 2018-09-18 | 2018-09-18 | Multispectral image change detection method based on probability segmentation and Gaussian mixture clustering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109446894A true CN109446894A (en) | 2019-03-08 |
| CN109446894B CN109446894B (en) | 2021-10-22 |
Family
ID=65532835
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811087966.8A Active CN109446894B (en) | 2018-09-18 | 2018-09-18 | Multispectral image change detection method based on probability segmentation and Gaussian mixture clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446894B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091054A (en) * | 2019-11-13 | 2020-05-01 | 广东国地规划科技股份有限公司 | Method, system and storage medium for monitoring land type change |
| CN111161225A (en) * | 2019-12-19 | 2020-05-15 | 浙江大华技术股份有限公司 | Image difference detection method and device, electronic equipment and storage medium |
| CN111160300A (en) * | 2019-12-31 | 2020-05-15 | 北京理工大学重庆创新中心 | Deep learning hyperspectral image saliency detection algorithm combined with global prior |
| CN111583274A (en) * | 2020-04-30 | 2020-08-25 | 贝壳技术有限公司 | Image segmentation method and device, computer-readable storage medium and electronic equipment |
| CN112802054A (en) * | 2021-02-04 | 2021-05-14 | 重庆大学 | Mixed Gaussian model foreground detection method fusing image segmentation |
| CN113160146A (en) * | 2021-03-29 | 2021-07-23 | 重庆大学 | Change detection method based on graph neural network |
| CN113791318A (en) * | 2021-09-01 | 2021-12-14 | 上海交通大学 | Partial discharge fault identification method and system based on multispectral detection array |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030214583A1 (en) * | 2002-05-20 | 2003-11-20 | Mokhtar Sadok | Distinguishing between fire and non-fire conditions using cameras |
| CN101634709A (en) * | 2009-08-19 | 2010-01-27 | 西安电子科技大学 | Method for detecting changes of SAR images based on multi-scale product and principal component analysis |
| CN105118065A (en) * | 2015-09-14 | 2015-12-02 | 中国民航大学 | Polari SAR (metric synthetic aperture radar) image variation detection method of wavelet domain polarization distance transformation |
| CN106056577A (en) * | 2016-05-19 | 2016-10-26 | 西安电子科技大学 | Hybrid cascaded SAR image change detection method based on MDS-SRM |
| CN107992891A (en) * | 2017-12-01 | 2018-05-04 | 西安电子科技大学 | Based on spectrum vector analysis multi-spectral remote sensing image change detecting method |
-
2018
- 2018-09-18 CN CN201811087966.8A patent/CN109446894B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030214583A1 (en) * | 2002-05-20 | 2003-11-20 | Mokhtar Sadok | Distinguishing between fire and non-fire conditions using cameras |
| CN101634709A (en) * | 2009-08-19 | 2010-01-27 | 西安电子科技大学 | Method for detecting changes of SAR images based on multi-scale product and principal component analysis |
| CN105118065A (en) * | 2015-09-14 | 2015-12-02 | 中国民航大学 | Polari SAR (metric synthetic aperture radar) image variation detection method of wavelet domain polarization distance transformation |
| CN106056577A (en) * | 2016-05-19 | 2016-10-26 | 西安电子科技大学 | Hybrid cascaded SAR image change detection method based on MDS-SRM |
| CN107992891A (en) * | 2017-12-01 | 2018-05-04 | 西安电子科技大学 | Based on spectrum vector analysis multi-spectral remote sensing image change detecting method |
Non-Patent Citations (5)
| Title |
|---|
| YUANYUAN CHEN: "Change Detection in Multispectral Remote Sensing Images Based on Optimized Fusion of Subspaces", 《2018 INTERNATIONAL CONFERENCE ON INTELLIGENCE SCIENCE AND BIG DATA ENGINEERING》 * |
| 张建龙等: "DSSRM级联分割的SAR图像变化检测", 《遥感学报》 * |
| 王爱斐: "基于混合高斯模型的地板块纹理分类算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 王鑫: "基于高斯混合模型的k均值初始化EM算法的研究", 《商丘师范学院学报》 * |
| 胡雪梅: "基于MST的遥感图像变化检测研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091054A (en) * | 2019-11-13 | 2020-05-01 | 广东国地规划科技股份有限公司 | Method, system and storage medium for monitoring land type change |
| CN111091054B (en) * | 2019-11-13 | 2020-11-10 | 广东国地规划科技股份有限公司 | Method, system and device for monitoring land type change and storage medium |
| CN111161225A (en) * | 2019-12-19 | 2020-05-15 | 浙江大华技术股份有限公司 | Image difference detection method and device, electronic equipment and storage medium |
| CN111161225B (en) * | 2019-12-19 | 2023-05-12 | 浙江大华技术股份有限公司 | Image difference detection method and device, electronic equipment and storage medium |
| CN111160300A (en) * | 2019-12-31 | 2020-05-15 | 北京理工大学重庆创新中心 | Deep learning hyperspectral image saliency detection algorithm combined with global prior |
| CN111160300B (en) * | 2019-12-31 | 2022-06-28 | 北京理工大学重庆创新中心 | Deep learning hyperspectral image saliency detection algorithm combined with global prior |
| CN111583274A (en) * | 2020-04-30 | 2020-08-25 | 贝壳技术有限公司 | Image segmentation method and device, computer-readable storage medium and electronic equipment |
| CN112802054A (en) * | 2021-02-04 | 2021-05-14 | 重庆大学 | Mixed Gaussian model foreground detection method fusing image segmentation |
| CN112802054B (en) * | 2021-02-04 | 2023-09-01 | 重庆大学 | Mixed Gaussian model foreground detection method based on fusion image segmentation |
| CN113160146A (en) * | 2021-03-29 | 2021-07-23 | 重庆大学 | Change detection method based on graph neural network |
| CN113160146B (en) * | 2021-03-29 | 2022-05-17 | 重庆大学 | Change detection method based on graph neural network |
| CN113791318A (en) * | 2021-09-01 | 2021-12-14 | 上海交通大学 | Partial discharge fault identification method and system based on multispectral detection array |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446894B (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109446894B (en) | Multispectral image change detection method based on probability segmentation and Gaussian mixture clustering | |
| CN111091105B (en) | Remote sensing image target detection method based on new frame regression loss function | |
| CN110136154B (en) | Remote sensing image semantic segmentation method based on full convolution network and morphological processing | |
| CN108510532B (en) | Optical and SAR image registration method based on deep convolution GAN | |
| CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
| CN111126482B (en) | Remote sensing image automatic classification method based on multi-classifier cascade model | |
| CN106875395B (en) | Super-pixel-level SAR image change detection method based on deep neural network | |
| CN110969088A (en) | Remote sensing image change detection method based on significance detection and depth twin neural network | |
| CN111899172A (en) | Vehicle target detection method oriented to remote sensing application scene | |
| CN108492298B (en) | Multispectral image change detection method based on generation countermeasure network | |
| CN106096655B (en) | A kind of remote sensing image airplane detection method based on convolutional neural networks | |
| CN109657610A (en) | A kind of land use change survey detection method of high-resolution multi-source Remote Sensing Images | |
| CN110633708A (en) | Deep network significance detection method based on global model and local optimization | |
| Fan et al. | Registration of multiresolution remote sensing images based on L2-siamese model | |
| CN103714148B (en) | SAR image search method based on sparse coding classification | |
| CN111539422B (en) | Flight target cooperative identification method based on fast RCNN | |
| CN107491734A (en) | Semi-supervised Classification of Polarimetric SAR Image method based on multi-core integration Yu space W ishart LapSVM | |
| CN106096658B (en) | Aerial Images classification method based on unsupervised deep space feature coding | |
| CN112288758B (en) | Infrared and visible light image registration method for power equipment | |
| Zhang et al. | Superpixel generation for SAR imagery based on fast DBSCAN clustering with edge penalty | |
| CN106991686A (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
| CN112802054A (en) | Mixed Gaussian model foreground detection method fusing image segmentation | |
| CN115311502A (en) | Remote sensing image small sample scene classification method based on multi-scale double-flow architecture | |
| CN105787505A (en) | Infrared image clustering segmentation method combining sparse coding and spatial constraints | |
| Chen et al. | Heterogeneous images change detection based on iterative joint global–local translation |
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 |