CN101699514B - Immune clone quantum clustering-based SAR image segmenting method - Google Patents
Immune clone quantum clustering-based SAR image segmenting method Download PDFInfo
- Publication number
- CN101699514B CN101699514B CN2009102186555A CN200910218655A CN101699514B CN 101699514 B CN101699514 B CN 101699514B CN 2009102186555 A CN2009102186555 A CN 2009102186555A CN 200910218655 A CN200910218655 A CN 200910218655A CN 101699514 B CN101699514 B CN 101699514B
- Authority
- CN
- China
- Prior art keywords
- population
- antibody
- sar image
- affinity degree
- max
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an immune clone quantum clustering-based SAR image segmenting method, which relates to the technical field of image processing, and mainly solves the problem of limitation on the application of the conventional quantum clustering technology in a large-scale data set. The immune clone quantum clustering-based SAR image segmenting method is implemented by the following steps: 1) extracting features of an SAR image to be segmented; 2) initializing an antibody population and coding antibodies; 3) calculating antibody affinity according to quantum mechanical characteristics, and dividing the antibody population into an elite population and a general population; 4) designing different immune clone optimization operators for the elite population and the general population respectively, and performing a cloning operation, a normal cloud model-based adaptive mutation operation, a uniform hypermutation operation, a clonal selection operation and a hypercube interlace operation orderly; and 5) outputting an SAR image segmentation result. The immune clone quantum clustering-based SAR image segmenting method has high iteration optimization speed and high stability, can effectively segment the SAR image which contains large-scale data volume, and is suitable for object detection and identification of the SAR image.
Description
Technical field
The invention belongs to technical field of image processing, relate to the SAR image segmentation, can be used for Radar Targets'Detection and Target Recognition.
Background technology
Cluster is meant under without any the priori situation about sample, utilize the method research of mathematics and the classification of processing special object, there is not one the sample of classification mark to be divided into several subclass according to certain criterion, make similar sample be classified as a class as far as possible, and dissimilar sample is divided in the different classifications as far as possible.Cluster analysis is a kind of of multivariate statistical analysis, also is an important branch of non-supervised recognition.Existing clustering algorithm roughly can be divided into based on the cluster of dividing, based on the cluster of level, based on the cluster of density, based on the cluster of grid, based on the cluster of model, and with fuzzy theory, graph theory, the clustering technique that combines of calculating association area naturally.
Image segmentation is divided into several significant parts according to certain homogeneity or conforming principle with image exactly, interested object is extracted from the ten minutes complicated background, so that further analyze.Therefore, can utilize clustering method that the discontinuous part of one or more features in the image is divided into a sub regions separately, raw information is converted into compact more form, make higher level graphical analysis and understanding become possibility.Wherein, along with synthetic-aperture radar SAR technology is widely used in national defense construction and civil area, the SAR image segmentation has been subjected to increasing attention in recent years as the committed step of SAR image processing process.SAR image partition method based on clustering technique comprises hierarchical cluster again and divides cluster.
As a kind of quantum clustering based on the nonparametric clustering technique of dividing, it can overcome traditional clustering algorithm initial value and noise-sensitive, cluster classification number are wanted defective such as in advance given.But the method that quantum clustering descends by gradient is carried out iteration and is easy to be absorbed in local extremum, and simultaneously, iteration speed has limited it at the large-scale dataset especially application in image segmentation field slowly.Although also there are some improvement technology, for example, the improvement estimated of adjusting the distance, improvement that scale parameter is estimated etc. all fail fundamentally to solve above bottleneck problem.
Summary of the invention
The objective of the invention is to overcome the deficiency of above-mentioned prior art, a kind of image partition method based on immune clone quantum clustering has been proposed, the immune clone Choice Theory is combined with quantum clustering, with the susceptibility that reduces scale parameter is changed, effectively the SAR image that comprises large-scale data is cut apart.
For achieving the above object, the present invention includes following steps:
(1) be that 256 * 256 SAR image extracts its gray level co-occurrence matrixes, wavelet transformation and Contourlet transform characteristics respectively to size to be split, all N=35 dimensional feature vectors that extract are combined by row, obtain 65536 * 35 input sample of data;
(2) a picked at random k sample is as the initial antibodies population from above data sample, and k is the classification number of SAR image segmentation;
(3) the initial antibodies population is encoded, each antibody length is N * k, and wherein, first length is that the field of N is represented first cluster centre, and second length is represented second cluster centre for the N field, and the rest may be inferred;
(4) antibody behind the coding is calculated affinity degree value according to following steps:
4a) two sample point x of definition
iAnd x
jBetween distance function D
Ij=|| x
i-x
j||, find the solution schrodinger equation
The computing formula that obtains potential-energy function V is
Wherein, H is a Hamiltonian,
E is the energy feature value of Hamiltonian, and d is the intrinsic dimensionality of data set, and x is the data sample point, and σ is a scale parameter,
Be Laplace operator;
4b) determine K cluster centre { c according to the minimal point of the potential-energy function value that calculates
i, i=1 ... K}, and according to sample point and all kinds of division set O of the nearest principle acquisition of each cluster centre Euclidean distance
i, n
iBe O
iThe number of middle sample point is to each set O
iIn all sample points average, obtain new cluster centre and be
According to the new cluster centre that calculates, the computing formula that obtains affinity degree function is
Wherein, || || for asking for the operational symbol of Euclidean distance;
(5) calculate the affinity degree value of all antibody in the antibody population after, they are sorted from high to low, pM antibody getting the front is as the elite population, remainder is called common population, wherein, the ratio value of p for setting, M is the population size;
(6) the elite population is carried out clone operations according to the clone's scale that is directly proportional with affinity degree value, common population is carried out clone operations according to fixing clone's scale;
(7) after the clone operations, the elite population being carried out self-adaptation mutation operation based on normal cloud model, is 1 the equally distributed exclusive-OR function of hypermutation at random that satisfies to the common population probability that makes a variation;
(8) from the sub-population of antibody behind process clone operations and the mutation operation, select outstanding antibody to form new elite population and common population according to probability, the affinity degree mxm. of the elite population that this is new is fbest1, and the affinity degree mxm. of new common population is fbest2;
(9) the affinity degree mxm. with the affinity degree mxm. of new elite population and new common population compares, if fbest1 〉=fbest2, pM just that the antibody of elite population and common population affinity degree is minimum antibody carries out hypercube and intersects; If fbest1<fbest2, pM just that common population affinity degree is the highest antibody carries out hypercube with elite population antibody and intersects;
(10) through after the interlace operation, return step (4) and carry out iteration optimization again, repeat N
MaxInferior;
(11) will be through N
MaxThe affinity degree mxm. fbest2 of affinity degree mxm. fbest1 of the elite population that inferior iteration obtains at last and common population compares again, with the final cluster centre of the pairing antibody representative of the higher value after the comparison, and be divided in the different classifications according to this cluster centre each pixel with the SAR image, obtain final segmentation result.
The present invention has the following advantages compared with prior art:
1) the present invention has introduced the Immune Clone Selection optimizing process with biology immune mechanism, can effectively overcome the defective that is absorbed in local extremum in the quantum clustering iterative process easily;
2) the present invention can effectively overcome existing quantum clustering technology in the limitation of handling on the large-scale data ability by the new affinity degree function calculation formula of design, can directly cut apart the SAR image;
3) the existing relatively quantum clustering technology of the present invention by proposing self-adaptation mutation operation and the hypercube interlace operation based on normal cloud model, can be accelerated the splitting speed of SAR image, and obtain segmentation effect preferably;
The simulation experiment result shows that the immune clone quantum clustering of proposition can be effectively applied to the SAR image segmentation, and further is applied to Radar Targets'Detection and Target Recognition.
Description of drawings
Fig. 1 is realization flow figure of the present invention;
Fig. 2 is the self-adaptation mutation operation synoptic diagram based on normal cloud model of the present invention;
Fig. 3 is a hypercube interlace operation synoptic diagram of the present invention;
Fig. 4 is the contrast and experiment that the present invention and existing FCM clustering algorithm are applied to the SAR image segmentation.
Embodiment
With reference to Fig. 1, specific implementation process of the present invention is as follows:
SAR image not only data volume is big, in imaging process different atural object have different back to emission and scattering properties, thereby have abundant texture information.And the intrinsic coherent speckle noise of image directly exerts an influence for segmentation performance.Therefore, be necessary before image segmentation, the SAR image to be carried out texture analysis, extract effective textural characteristics and carry out cluster.SAR image texture characteristic extracting method commonly used comprises the gray level co-occurrence matrixes that extracts image and wavelet character etc.Studies show that the multi-scale geometric analysis theory can effectively remedy the defective of wavelet analysis, in texture feature extraction, can fully excavate the image internal information.For example, profile and texture information abundant in the image can be effectively extracted in the Contourlet conversion.Therefore, we attempt these feature extraction directions are merged, to obtain better segmentation result.
On above analysis foundation, earlier SAR image to be split is extracted its gray level co-occurrence matrixes, wavelet transformation and Contourlet transform characteristics respectively, carry out Feature Fusion then.Wherein, extract the angle second order distance of its 4 directions and homogeneity district totally 8 dimensional features for gray level co-occurrence matrixes, the moving window size is 17 * 17; For wavelet transformation and Contourlet conversion, to extract 3 layers of wavelet decomposition 10 dimensional feature and 2 layers of Contourlet respectively and decompose 17 dimensional features, the moving window size is 16 * 16.These proper vectors are combined by row, are 256 * 256 SAR image for size, just constitute the input data that 65536 * 35 data set is used for immune clone quantum clustering.
(2.1) a picked at random k sample point is as the initial antibodies population from input data set, and k is the classification number of SAR image segmentation;
(2.2) antagonist is encoded, and each antibody is a string real number, for the N dimension space, and k cluster centre, antibody length is N * k.Wherein, first length is that the field of N is represented first cluster centre, and second field represented second cluster centre, and the rest may be inferred.For example, consider 2 dimension data collection with 3 cluster centres, the length of antibody is exactly 2 * 3, and 3 sample points of picked at random are as initial center, for example, (1,3), (2,4) and (5,6), antibody just is encoded into 1-3-2-4-5-6, and each group is respectively represented an initial cluster center.
Step 3. calculating antibody affinity degree.
(3.1) two sample point x of definition
iAnd x
jBetween distance function D
Ij=|| x
i-x
j||, find the solution schrodinger equation
The computing formula that obtains potential-energy function V is
Wherein, H is a Hamiltonian,
E is the energy feature value of Hamiltonian, and d is the intrinsic dimensionality of data set, and x is the data sample point, and σ is a scale parameter,
Be Laplace operator;
(3.2) determine K cluster centre { c according to the minimal point of the potential-energy function value that calculates
i, i=1 ... K}, and according to sample point and all kinds of division set O of the nearest principle acquisition of each cluster centre Euclidean distance
i, n
iBe O
iThe number of middle sample point is to each set O
iIn all sample points average, obtain new cluster centre and be
According to the new cluster centre that calculates, the computing formula that obtains affinity degree function is
Wherein, || || for asking for the operational symbol of Euclidean distance.Above-mentioned affinity degree function calculation formula only need calculate on the distance matrix basis between sample point, thereby can effectively overcome existing quantum clustering technology in the limitation of handling on the large-scale data ability, and have stability preferably, can be directly used in the SAR image segmentation;
The operation operator that step 4. design immune clone is optimized.
(4.1) clone operations
Calculate the affinity degree value of all antibody in the antibody population and ordering from high to low, pM higher antibody of ordering back affinity degree value is referred to as the elite population, and remainder is called common population, wherein, the ratio value of p for setting, M is the population size, for the elite population, single antibody a
iClone's scale according to affinity degree value f (a
i) be calculated as follows:
Wherein, n
cBe total clone's scale, Int () expression rounds operation, clones according to the clone's scale that calculates then;
For common population, directly to a plurality of outstanding antibody according to fixing clone's scale N
cClone.
(4.2) mutation operation
Adopt self-adaptation to make a variation to the elite population, self-adaptation variation probability P based on normal cloud model
mComputing formula as follows:
Wherein, η is the constant relevant with solving precision, and t is an iterations, f, f
MaxAnd f
AveBe respectively affinity degree value, affinity degree maximal value and the affinity degree average of each iteration, P
MaxAnd P
MinBe default variation probability bound, A and B are that weight coefficient is taken as A=B=0.5.
The variation probability that successively decreases has been considered evolutionary generation and antibody affinity degree simultaneously, all keeps a relatively large variation probability at all antibody of evolution initial stage, helps keeping the population diversity.Along with population is constantly evolved, equilibrium regulation and control lower variation probability at evolutionary generation and antibody affinity degree proportion reduces gradually, can guarantee that algorithm is deep into around the Local Extremum, approach optimum solution, thereby obtain SAR image segmentation effect preferably with higher precision.
Cloud model is the qualitativing concept represented with natural language and the uncertain transformation model between its quantificational expression, mainly reflects the ambiguity and the randomness of notion in objective world or the human cognitive.Wherein, normal cloud model has randomness and steady tendency, the cloud model of the followed normal distribution stochastic distribution rule that characterizes with expectation value Ex, entropy En and super entropy He.Apply it in the mutation operation, make the randomness of cloud variation can keep the population diversity to avoid search to be absorbed in local extremum, steady tendency has then been protected defect individual preferably and has been carried out overall situation location.As shown in Figure 2, for one dimension normal cloud model C (0, En, He), be that Fig. 2 (a) is C (0 near 0 the true origin in expectation value, 0.5,0.1) time normal cloud model, Fig. 2 (b) is C (0,0.5,0.2) time normal cloud model, Fig. 2 (c) is C (0,1,0.1) normal cloud model the time, Fig. 2 (d) is C (0,1,0.2) normal cloud model the time.As seen, En is big more, and the water dust coverage is big more, and He is big more, and the water dust dispersion degree is big more.Expectation value has embodied the stability of variation, and entropy has embodied the range of variation, and super entropy has embodied the precision of variation.
Self-adaptation mutation operator based on normal cloud model is described below:
1) calculates initial degree of certainty: μ=P according to affinity degree value
Max-(P
Max-P
Min) (f
Max-f)/(f
Max+ f
Min), wherein, f
MinAffinity degree minimum value for each iteration;
2) make Ex get the preceding antibody of variation, En=0.1 σ (σ is the standard deviation of each dimension data variable), He=0.1En;
3) produce expectation and be En, variance is the normal distribution random number of He: En '=En+Herandn;
4) between (0,1), produce a random number r, if satisfy r<P
m, then the antibody after the variation is
For common population, employing variation probability is 1 the equally distributed random variation of satisfying.Produce a random number δ between (0,1), then corresponding gene position λ makes a variation into λ+2 δ-1.Adopt even random variation not have directivity, can produce variation away from initial antibodies.Thereby, keeping the multifarious while of population, make common population possess ability of searching optimum under the different probability of hypermutation, improve the segmentation effect of SAR image.
(4.3) Immune Clone Selection operation
From forming new population through the outstanding antibody of selection the sub-population of antibody behind clone operations and the mutation operation.For the optimum antibody b in each sub-population
i(k)={ a
Ij(k) | f (a
Ij(k))=max f (A
m(k)), j=1,2 ..., q
i, b
i(k) replace initial antibodies A
i(k) probability is:
Wherein, β>0 is a parameter relevant with the antibody population diversity, is worth greatly more, and diversity is good more.
(4.4) hypercube interlace operation
Through after the Immune Clone Selection operation, elite population and common population affinity degree mxm. are respectively fbest1 and fbest2, when fbest1 〉=fbest2, elite population antibody are carried out the hypercube intersection with pM minimum antibody of common population affinity degree; When fbest1<fbest2, pM the antibody that common population affinity degree is the highest carries out hypercube with elite population antibody and intersects.Shown in Fig. 3 (a), under the one-dimensional case, the search volume that hypercube intersects outwards is extended to line segment CD with line segment AB; Shown in Fig. 3 (b), under the two-dimensional case, the search volume is the expansion to the plane; Shown in Fig. 3 (c), under the three-dimensional situation, the search volume is made of jointly inner rectangular parallelepiped and outside continuation space; Three-dimensional above situation, the search volume then is to the continuation of hypercube according to corresponding dimension.
For intersecting parent antibody x
kAnd y
k, make l
Min=min (x
k, y
k), l
Max=max (x
k, y
k), δ=l
Max-l
Min, be x then through the filial generation antibody after the hypercube interlace operation
K+1And y
K+1, wherein:
Wherein, r is a random number, P
cBe crossover probability, unifrnd () is even value at random within the specific limits, and α=0.2 is the space continuation coefficient of hypercube.
Through after the interlace operation, return clone operations and carry out iteration optimization again, repeat N
MaxInferior;
Step 5. output segmentation result.
Will be through N
MaxElite population affinity degree mxm. fbest1 that inferior iteration obtains at last and common population affinity degree mxm. fbest2 compare, with the final cluster centre of the pairing antibody representative of the higher value after the comparison, be divided in the different classifications according to cluster centre each pixel then, obtain final segmentation result the SAR image.
Effect of the present invention can be verified by following emulation experiment figure:
Fig. 4 (a) width of cloth is near the Ku wave band SAR image of the Rio Grande the Albuquerque area, New Mexico, and size is 256 * 256, and we wish it is divided into three classes, i.e. farmland, vegetation and river.
Fig. 4 (b) is a width of cloth ERS-SAR image, and it comprises two class atural objects, i.e. crops and forest.
Fig. 4 (c) uses the result that the FCM cluster is cut apart to Fig. 4 (a).
Fig. 4 (d) uses the result that the FCM cluster is cut apart to Fig. 4 (b).
Fig. 4 (e) uses the result that the present invention is cut apart to Fig. 4 (a).
Fig. 4 (f) uses the result that the present invention is cut apart to Fig. 4 (b).
Can find that from the square frame interior zone of Fig. 4 (c) the bridge target disconnects in the segmentation result of FCM cluster, and segmentation result of the present invention can detect complete bridge target effectively shown in Fig. 4 (e).
Can find that from the elliptic region inside of Fig. 4 (d) segmentation result of FCM cluster contains more spot, and segmentation result of the present invention shown in Fig. 4 (f) has obtained regional preferably consistance, contained spot is less.
Claims (3)
1. SAR image partition method based on immune clone quantum clustering may further comprise the steps:
(1) be that 256 * 256 SAR image extracts its gray level co-occurrence matrixes, wavelet transformation and Contourlet transform characteristics respectively to size to be split, all N=35 dimensional feature vectors that extract are combined by row, obtain 65536 * 35 input sample of data;
(2) a picked at random k sample is as the initial antibodies population from input sample of data, and k is the classification number of SAR image segmentation;
(3) the initial antibodies population is encoded, each antibody length is N * k, and wherein, first length is that the field of N is represented first cluster centre, and second length is represented second cluster centre for the N field, and the rest may be inferred;
(4) antibody behind the coding is calculated affinity degree value according to following steps:
4a) two sample point x of definition
iAnd x
jBetween distance function D
Ij=|| x
i-x
j||, find the solution schrodinger equation
The computing formula that obtains potential-energy function V is:
Wherein, H is a Hamiltonian,
E is the energy feature value of Hamiltonian, and d is the intrinsic dimensionality of data set, and x is the data sample point, and σ is a scale parameter,
Be Laplace operator, i=1 ..., k, j=1 ..., n, wherein n is the number of input sample of data;
4b) determine K cluster centre { c according to the minimal point of the potential-energy function value that calculates
i, i=1 ... K}, and according to sample point and all kinds of division set O of the nearest principle acquisition of each cluster centre Euclidean distance
i, n
iBe O
iThe number of middle sample point is to each set O
iIn all sample points average, obtain new cluster centre and be:
According to the new cluster centre that calculates
Obtaining affinity degree function calculation formula is:
Wherein, || || for asking for the operational symbol of Euclidean distance;
(5) calculate the affinity degree value of all antibody in the antibody population after, they are sorted from high to low, pM antibody getting the front is as the elite population, remainder is called common population, wherein, the ratio value of p for setting, M is the population size;
(6) the elite population is carried out clone operations according to the clone's scale that is directly proportional with affinity degree value, common population is carried out clone operations according to fixing clone's scale;
(7) after the clone operations, the elite population being carried out self-adaptation mutation operation based on normal cloud model, is 1 the equally distributed exclusive-OR function of hypermutation at random that satisfies to the common population probability that makes a variation;
(8) from the sub-population of antibody behind process clone operations and the mutation operation, select outstanding antibody to form new elite population and common population according to probability, the affinity degree mxm. of the elite population that this is new is fbest1, and the affinity degree mxm. of new common population is fbest2;
(9) the affinity degree mxm. with the affinity degree mxm. of new elite population and new common population compares, if fbest1 〉=fbest2, pM just that the antibody of elite population and common population affinity degree is minimum antibody carries out hypercube and intersects; If fbest1<fbest2, pM just that common population affinity degree is the highest antibody carries out hypercube with elite population antibody and intersects;
(10) through after the interlace operation, return step (4) and carry out iteration optimization again, repeat N
MaxInferior;
(11) will be through N
MaxThe affinity degree mxm. fbest2 of affinity degree mxm. fbest1 of the elite population that inferior iteration obtains at last and common population compares again, with the final cluster centre of the pairing antibody representative of the higher value after the comparison, and be divided in the different classifications according to this cluster centre each pixel with the SAR image, obtain final segmentation result.
2. method according to claim 1, the described self-adaptation mutation operation based on normal cloud model of step (7) wherein, carry out according to following steps:
2a) calculate self-adaptation variation probability P according to following formula
m:
Wherein, η is the constant relevant with solving precision, and t is an iterations, P
MaxAnd P
MinBe respectively the upper and lower bound of default variation probability, f, f
MaxAnd f
AveBe respectively affinity degree value, affinity degree maximal value and the affinity degree average of each iteration, A and B are weight coefficient, are taken as A=B=0.5;
2b) produce several r between one (0,1) at random, if satisfy r<P
m, then the antibody after the variation is
Wherein, Ex is the preceding antibody of variation, and En '=En+Herandn, En=0.1 σ, σ are the standard deviation of each dimension data variable, and He=0.1En, randn represent to satisfy the random number of standardized normal distribution, μ=P
Max-(P
Max-P
Min) (f
Max-f)/(f
Max+ f
Min), f
MinAffinity degree minimum value for each iteration.
3. method according to claim 1, the wherein described hypercube interlace operation of step (9), carry out according to following steps:
For two parent antibody x that intersect
kAnd y
k, make l
Min=min (x
k, y
k), l
Max=max (x
k, y
k), δ=l
Max-l
Min, be x then through the filial generation antibody after the hypercube interlace operation
K+1And y
K+1, wherein:
Wherein, r is the random number between (0,1), P
cBe crossover probability, the even at random within the specific limits value of unifrnd () expression, α=0.2 is the space continuation coefficient of hypercube.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009102186555A CN101699514B (en) | 2009-10-30 | 2009-10-30 | Immune clone quantum clustering-based SAR image segmenting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009102186555A CN101699514B (en) | 2009-10-30 | 2009-10-30 | Immune clone quantum clustering-based SAR image segmenting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101699514A CN101699514A (en) | 2010-04-28 |
CN101699514B true CN101699514B (en) | 2011-09-21 |
Family
ID=42147973
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2009102186555A Expired - Fee Related CN101699514B (en) | 2009-10-30 | 2009-10-30 | Immune clone quantum clustering-based SAR image segmenting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101699514B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101847263B (en) * | 2010-06-04 | 2012-02-08 | 西安电子科技大学 | Unsupervised image division method based on multi-target immune cluster integration |
CN101887517B (en) * | 2010-06-13 | 2012-09-26 | 重庆理工大学 | Immune cloned finger venous image characteristic extraction method based on linear weighted function |
CN101866490B (en) * | 2010-06-30 | 2012-02-08 | 西安电子科技大学 | Image segmentation method based on differential immune clone clustering |
CN101908213B (en) * | 2010-07-16 | 2012-10-24 | 西安电子科技大学 | SAR image change detection method based on quantum-inspired immune clone |
CN102374936B (en) * | 2010-08-23 | 2014-03-05 | 太原理工大学 | Mechanical failure diagnostic method based on complex immune network algorithm |
CN102073984B (en) * | 2011-01-10 | 2012-07-04 | 武汉工程大学 | Image II type Schrodinger transformation method |
CN102496156B (en) * | 2011-11-17 | 2013-11-20 | 西安电子科技大学 | Medical image segmentation method based on quantum-behaved particle swarm cooperative optimization |
CN108896971B (en) * | 2018-05-10 | 2022-03-22 | 西安电子科技大学 | Simulation method for echoes of small targets floating on sea surface |
CN109559782B (en) * | 2018-11-08 | 2022-11-25 | 武汉科技大学 | DNA sequence coding method based on multi-target genetic algorithm |
CN111426323B (en) * | 2020-04-16 | 2022-04-12 | 南方电网科学研究院有限责任公司 | Routing planning method and device for inspection robot |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101271572A (en) * | 2008-03-28 | 2008-09-24 | 西安电子科技大学 | Image segmentation method based on immunity clone selection clustering |
-
2009
- 2009-10-30 CN CN2009102186555A patent/CN101699514B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101271572A (en) * | 2008-03-28 | 2008-09-24 | 西安电子科技大学 | Image segmentation method based on immunity clone selection clustering |
Non-Patent Citations (5)
Title |
---|
Paul R. Kersten,et al..Unsupervised Classification of Polarimetric Synthetic Aperture Radar Images Using Fuzzy Clustering and EM Clustering.《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》.2005,第43卷(第3期), * |
李志华,等.一种基于量子机制的分类属性数据模糊聚类算法.《***仿真学报》.2008,第20卷(第8期), * |
田小林,等.基于ICA优化空间信息PCM的SAR图像分割.《电子与信息学报》.2008,第30卷(第7期), * |
马文萍,等.免疫克隆SAR图像分割算法.《电子与信息学报》.2009,第31卷(第7期), * |
马文萍,等.基于量子克隆优化的SAR图像分类.《电子学报》.2007,第35卷(第12期), * |
Also Published As
Publication number | Publication date |
---|---|
CN101699514A (en) | 2010-04-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101699514B (en) | Immune clone quantum clustering-based SAR image segmenting method | |
Sameen et al. | Application of convolutional neural networks featuring Bayesian optimization for landslide susceptibility assessment | |
CN112052754B (en) | Polarization SAR image ground object classification method based on self-supervision characterization learning | |
CN102800093B (en) | Based on the multiple-target remote sensing image dividing method decomposed | |
CN103839261A (en) | SAR image segmentation method based on decomposition evolution multi-objective optimization and FCM | |
CN101847263B (en) | Unsupervised image division method based on multi-target immune cluster integration | |
CN103914705B (en) | Hyperspectral image classification and wave band selection method based on multi-target immune cloning | |
CN104794730B (en) | SAR image segmentation method based on super-pixel | |
CN101980298A (en) | Multi-agent genetic clustering algorithm-based image segmentation method | |
CN105184314B (en) | Wrapper formula EO-1 hyperion band selection methods based on pixel cluster | |
CN112508936A (en) | Remote sensing image change detection method based on deep learning | |
Feng et al. | Embranchment cnn based local climate zone classification using sar and multispectral remote sensing data | |
CN103886335A (en) | Polarized SAR image classifying method based on fuzzy particle swarms and scattering entropy | |
Taib et al. | Data clustering using hybrid water cycle algorithm and a local pattern search method | |
CN111222575B (en) | KLXS multi-model fusion method and system based on HRRP target recognition | |
Günen | Adaptive neighborhood size and effective geometric features selection for 3D scattered point cloud classification | |
CN111611960A (en) | Large-area ground surface coverage classification method based on multilayer perceptive neural network | |
CN113516019B (en) | Hyperspectral image unmixing method and device and electronic equipment | |
Mohamadzadeh et al. | Classification algorithms for remotely sensed images | |
Zhang et al. | Semantic segmentation of spectral LiDAR point clouds based on neural architecture search | |
Nie et al. | Semantic category balance-aware involved anti-interference network for remote sensing semantic segmentation | |
Althobaiti et al. | Intelligent deep data analytics-based remote sensing scene classification model | |
Zhang et al. | A novel combinational forecasting model of dust storms based on rare classes classification algorithm | |
CN101866483B (en) | Texture image segmentation method based on Lamarck multi-target immune algorithm | |
Prasad et al. | A novel decision tree approach for the prediction of precipitation using entropy in sliq |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110921 Termination date: 20181030 |
|
CF01 | Termination of patent right due to non-payment of annual fee |