CN105184764A - Image registering method in real number coding based clonal selection algorithm - Google Patents
Image registering method in real number coding based clonal selection algorithm Download PDFInfo
- Publication number
- CN105184764A CN105184764A CN201510229255.XA CN201510229255A CN105184764A CN 105184764 A CN105184764 A CN 105184764A CN 201510229255 A CN201510229255 A CN 201510229255A CN 105184764 A CN105184764 A CN 105184764A
- Authority
- CN
- China
- Prior art keywords
- antibody
- image
- affinity
- collection
- gray level
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an image registering method in a real number coding based clonal selection algorithm, and mainly solves the problem that the image registering accuracy is not high due to the fact that function optimization tends to local optimization in the prior art. The method comprises the steps that (1) a reference image and a floating image are input; (2) an antibody population is initialized in a real number coding method; (3) a normalized mutual information function is constructed and serves as an objective function; (4) the affinity of population antibodies is calculated; (5) selection, clone and variation are carried out on the antibodies; (6) a memory antibody set is formed; (7) the antibodies are memorized; (8) whether a termination condition is reached is determined, if no, the step (4) is returned to, and otherwise, a step (9) is turned to; and (9) an image is registered by utilizing the optimal antibody, and a result is output. The mutual information function is optimized in the real number coding based clonal selection algorithm, the problem that function optimization tends to local optimization in the prior art is solved, and the image registering accuracy is improved.
Description
Technical field
The invention belongs to image registration field, relate to the optimization method of objective function, specifically a kind of method for registering images based on clonal selection algorithm.
Background technology
Image registration techniques, as a very important research topic, has been widely used in computer vision, pattern match, medical image analysis and remote sensing image processing.Due to the novel sensor emerged in an endless stream, people are to the acquisition capability fast lifting of image.Novel sensor has various different characteristic, and thus different types of remote sensing images are also much continuous gets up.In view of sensor is when obtaining image information, such as spectral information, geometric size and time order and function etc. have obvious difference, so utilize single image information to be difficult to meet actual conditions in all fields.In order to the various information of different images better can be obtained, thus obtain more high-resolution remote sensing images, need the advantage making full use of multi-modality images, merge the image utilized under the different image-forming condition of distinct device, the prerequisite of image co-registration is exactly image registration.In brief, its fundamental purpose is exactly will to Same Scene not in the same time, and two width images of different angles or different sensors shooting align.
The method of present stage image registration mainly contains two types: based on the method for registering images of gray scale and the method for registering images of feature based.
Method for registering images based on gray scale does not need to carry out feature extraction to image, but the subregion of use entire image or image estimates the gray consistency between two width images, and conventional method for measuring similarity has: cross-correlation, and phase place is correlated with and mutual information.Although the computation complexity of these class methods is higher, these class methods have been proved to be and have achieved good effect in image registrations.Wherein, mutual information has following characteristics: the first, and it does not need the hypothesis of the image of different modalities being carried out to gray scale correspondence.The second, mutual information has good robustness to noise.3rd, when two width images reach accuracy registration, mutual information obtains maximal value.
The method for registering images of feature based needs first to the feature of image, and comprise a little, line or region are detected, and mate one by one afterwards to the feature detected, then by the spatial transform relation between the feature assessment two width image mated, and then registering images.Conventional characteristic detection method has: method and the consistent extract minutiae of phase place etc. of Harris Corner Detection, Canny detective operators, Iamge Segmentation, need after obtaining unique point to utilize spatial relationship or constant descriptor to carry out one_to_one corresponding to feature.When there being abundant feature, the method for registering images of feature based can try to achieve the conversion parameter close to global optimum easily.
Usually, image registration problem can be converted to the optimization problem of function after determining similarity measures, obtains the image of registration when similarity measurement reaches maximal value.Common function optimization method has genetic algorithm, particle cluster algorithm etc.Genetic algorithm is the evolution algorithm grown up based on the natural selection of Darwin's biological evolution theory and genetic mechanisms, it produces initial population at random, then fitness evaluation is carried out to each individuality of initial population, iteration starts that the high individuality of rear selection fitness carries out intersecting with certain probability, mutation operation, produce new individuality, again new individuality is carried out to the evaluation of fitness, certain iterations is set usually or iteration stopping after objective function reaches a certain threshold value, exports final solution.Its shortcoming is that computation complexity is high, easy Premature Convergence, and is easily absorbed in local optimum.
Artificial immune system (ArtificialImmunesystem, AIs) be a kind of intelligent method of natural imitation function of immune system, it realizes one and inspires by Immune System, by learning the learning art of the natural defense mechanism of external substance, there is provided noise to restrain oneself, teacherless learning, self-organization, the evolutionary learning mechanism such as memory, combine some advantages of the systems such as sorter, neural network and machine inference, therefore have and the novel potentiality of dealing with problems are provided.Clone's (Clone) immunity is the important theory of Immune System theory.Due to heredity and the gene mutation of immunocyte in propagation, define the diversity of immunocyte, the continuous propagation of these cells defines clone.The vegetative propagation of cell is called clone.
Burnet in 1958 etc. propose famous clonal selective theory, and its central idea is: antibody is natural products, is present in cell surface with the form of acceptor, and antigen can optionally react with it.The reaction of antigen and corresponding antibodies can cause cell clonal to rise in value, this colony has identical antibody specificity, wherein some cell clone is divided into antibody-producting cell, other form immunological memory cell and react with the secondary immunity after participating in, Immune Clone Selection is the dynamic process of organism immune system self-adaption antigenic stimulus, in this course, the biological natures such as the study embodied, memory, antibody diversity just artificial immune system are used for reference.Based on the viewpoint of information processing, can think, the essence of Immune Clone Selection is exactly in a generation is evolved, and near candidate's disaggregation, according to affinity degree size, produces the colony that a variation is separated.Clonal selection algorithm is the competition realizing between individuality by antibody one antigen affinity degree, and accommodative excess competition effectively, to keep the diversity of antibody population.
As a kind of new global optimization search, Immune Clonal Selection Algorithm takes into account global search and Local Search on algorithm realization, and constructs mnemon, single for the memory of genetic algorithm optimum individual is become the colony of a memory optimum solution.In general genetic algorithm, intersection is main operators, and variation is background operator, but clonal selection algorithm is then on the contrary, and experiment proves that Immune Clonal Selection Algorithm performance is better than corresponding genetic algorithm.In addition, the selection mechanism of Clone cells itself has memory function, and therefore can ensure that algorithm converges to optimum solution with probability l, the genetic algorithm of standard then can not.
Immune Clonal Selection Algorithm is applied in image registration by the present invention, and adopts the method for real coding, according to the size of fitness function, constantly updates the scale of population clone, tries to achieve optimum image registration parameter, reach good image registration effect.Data analysis by experiment, the inventive method is better than traditional genetic algorithm.
Summary of the invention
The object of the invention is to the deficiency for above-mentioned prior art, propose a kind of method for registering images based on clonal selection algorithm, improve the accuracy that registration parameter is estimated, realize the correct registration to reference picture and floating image.
Technical scheme steps of the present invention comprises as follows:
(1) resolution identical reference picture imageR and floating image imageS is inputted;
(2) antibody population initialization: produce antibody population at random, A represents antibody collection, and antibody collection A is by interim antibody collection A
rwith memory antibody collection A
mcomposition, N=r+m, r represent interim antibody collection A
rin interim antibody levels, m represents memory antibody collection A
min memory antibody quantity, N is the total quantity of antibody in antibody collection A;
(3) represent reference picture imageR, B with A and represent floating image imageS structure normalized mutual information function MI (A, B) as objective function:
Wherein, H (A) is the edge entropy of image A,
p
ai () is the probability that in image A, i-th grade of gray level occurs, M is the progression of gray level in image A, the edge entropy that H (B) is image B,
p
bj () is the probability that in image B, jth level gray level occurs, N is the progression of gray level in image B,
be the combination entropy between image A and image B, p (i, j) is jth level gray level this pair joint probability that same position occurs simultaneously in this two width image of image A and image B in i-th grade of gray level in image A, image B, and
h (i, j) is the joint histogram of image A and image B, and n is the number of image A and image B overlapping region pixel, and wherein joint histogram is defined as follows:
H (a, b) (0≤a≤K-1,0≤b≤L-1, K and L is the scope of gray-scale value in two width images) represents that in image A, gray level is a, and in image B, gray level is the number that the pixel of b is right;
(4) clonal selection algorithm is utilized to be optimized normalized mutual information function MI (A, B)
4a) calculate the affinity of antibody of population, i.e. the value of normalized mutual information function MI (A, B);
4b) antibody is selected: antagonist affinity, according to descending sort, chooses front k the highest individuality of affinity of antibody as interim antibody collection A
r;
4c) antibody cloning: k the antibody chosen will by independent cloning, the number of times of clone is also indefinite, concrete rule is: affinity is higher, affinity sequence number i is less, and the number of times be cloned is more, on the contrary, affinity is lower, affinity sequence number i is larger, is cloned number of times fewer, the total quantity N that antibody is cloned
cas shown in the formula:
Wherein, β is a multiplication factor, and N is the antibody total quantity in collection of antibodies A, and the span of affinity sequence number i is 1 is an independent variable operator to k, round (), represents to nearest integer and rounds, N
ci-th
the scale that corresponding i-th antibody is cloned.
4d) antibody variation: mutation operation is done to the antibody after clone.Random number between producing one 0 to 1, if this number is less than the mutation probability preset, then makes a variation to corresponding parameter:
x
new＝x+δ*λ
Wherein, x
newfor the parameter after variable x variation, δ=0.1* (x
max-x
min) be the region of search factor that narrows, x
maxfor the maximal value of variable x, x
minfor the minimum value of variable x, λ is a random number between 0 to 1, the step-length made a variation for regulating x;
4e) form memory antibody collection: recalculate by fitness value of antibody each in antibody population after above-mentioned mutation operation i.e. the size of affinity, if by the affinity of antibody in population after above-mentioned mutation operation higher than A
rin the affinity of antibody before corresponding variation, just replace original A with the antibody after mutation operation
rin, form memory antibody collection A
m;
4f) antibody memory: simulate naturally withering away of the B cell of 5% in artificial immune system bioselection, namely from antibody collection A, delete the antibody of affinity minimum 5%, reinitialize the new antibodies of generation 5%, with the antibody of 5% of new generation, substitute the antibody that 5% affinity deleted in antibody collection A is minimum.
(5) judge whether to meet evolution conditions: if iterations does not reach predefined evolutionary generation gene, then jump to 4a) continue to perform, otherwise, perform (6);
(6) when iterations reaches predefined evolutionary generation gene, the individuality that affinity is the highest is optimum antibody, the conversion parameter representated by optimum antibody is utilized to carry out spatial alternation to floating image imageS, calculate and output transform after floating image and reference picture imageR between association relationship, carry out registration with reference to the floating image after image imageR and conversion, finally export the image after registration.
The present invention regards image registration problem as a function optimization problem, wherein, normalized mutual information function is as objective function, the clonal selection algorithm based on real coding is utilized to be optimized objective function, solution when reaching maximum with association relationship carries out spatial alternation, obtain registering images, compared with prior art tool has the following advantages:
The first, image registration is regarded as a function optimization problem by the present invention, utilizes the clonal selection algorithm based on real coding to be optimized objective function mutual information, is easily absorbed in the shortcoming of local optimum when overcoming traditional genetic algorithm registering images.
The second, because antagonist of the present invention have employed the method for real coding, compare with traditional binary coding method and save a large amount of Code And Decode steps, therefore operational efficiency is higher, and result is also more accurate.
3rd, the main operators that mutation operator operates as it by clonal selection algorithm, expands search volume in certain algebraically.Clone cells itself has memory function in addition, and make algorithm convergence with probability 1 in globally optimal solution, the genetic algorithm of standard then can not.Therefore, carried out the optimizing of registration parameter by clonal selection algorithm, the effect obtained is more accurate.
Accompanying drawing explanation
Fig. 1 is realization flow figure of the present invention;
Fig. 2 is that the present invention and the registration result of existing method to the Yellow River estuary image first width cut-away view picture that RadarSat-2 satellite obtains 2008 and 2009 contrast, wherein, Fig. 2 (a) is reference picture, Fig. 2 (b) is floating image, Fig. 2 (c) is the image after registration of the present invention, Fig. 2 (d) is DE algorithm registering images, and Fig. 2 (e) is PSO algorithm registering images, and Fig. 2 (f) is GA algorithm registering images;
Fig. 3 is that the present invention and existing method are to the evolution curve of the Average Mutual value of the Yellow River estuary image first width cut-away view picture that RadarSat-2 satellite obtains 2008 and 2009;
Fig. 4 is that the present invention and the registration result of existing method to the Yellow River estuary image second width cut-away view picture that RadarSat-2 satellite obtains 2008 and 2009 contrast, wherein, Fig. 4 (a) is reference picture, Fig. 4 (b) is floating image, Fig. 4 (c) is the image after registration of the present invention, Fig. 4 (d) is DE algorithm registering images, and Fig. 4 (e) is PSO algorithm registering images, and Fig. 4 (f) is GA algorithm registering images;
Fig. 5 is that the present invention and existing method are to the evolution curve of the Average Mutual value of the Yellow River estuary image second width cut-away view picture that RadarSat-2 satellite obtains 2008 and 2009.
Embodiment
With reference to Fig. 1, performing step of the present invention is as follows:
Step 1, input reference picture and floating image;
Step 2, antibody population initialization: based on the antigen determined, random some antibody population of generation, A represents antibody collection, and antibody collection A is by interim antibody collection A
rwith memory antibody collection A
mcomposition, N=r+m, r represent interim antibody collection A
rin interim antibody levels, m represents memory antibody collection A
min memory antibody quantity, N is the total quantity of antibody in antibody collection A;
Step 3, represents reference picture imageR, B with A and represents floating image imageS structure normalized mutual information function MI (A, B) as objective function:
Wherein, H (A) is the edge entropy of image A,
p
ai () is the probability that in image A, i-th grade of gray level occurs, M is the progression of gray level in image A, the edge entropy that H (B) is image B,
p
bj () is the probability that in image B, jth level gray level occurs, N is the progression of gray level in image B,
be the combination entropy between image A and image B, p (i, j) is jth level gray level this pair joint probability that simultaneously occurs of same position in the picture in i-th grade of gray level in image A, image B, and
h (i, j) is the joint histogram of image A and image B, and n is the number of image A and image B overlapping region pixel, and wherein joint histogram is defined as follows:
H (i, j) (0≤i≤K-1,0≤j≤L-1, K and L is the scope of gray-scale value in two width images) represents that in image A, gray level is i, and in image B, gray level is the number that the pixel of j is right;
Step 4, utilizes clonal selection algorithm to be optimized normalized mutual information function MI (A, B), specifically comprises:
4a) calculate the affinity of antibody of initial population, i.e. the value of normalized mutual information function MI (A, B);
4b) antibody is selected: antagonist affinity, according to descending sort, chooses front k the highest individuality of affinity of antibody as interim antibody collection A
r;
4c) antibody cloning: according to the size of affinity, clones k the individuality that affinity is the highest chosen;
4d) antibody variation: mutation operation is done to the antibody after clone.Random number between producing one 0 to 1, if this number is less than mutation probability, then to x
oldmake a variation:
x
new＝x+δ*λ
Wherein, x
newfor the parameter after variable x variation, δ=0.1* (x
max-x
min) be the region of search factor that narrows, x
maxfor the maximal value of variable x, x
minfor the minimum value of variable x, λ is a random number between 0 to 1, the step-length made a variation for regulating x;
4e) form memory antibody collection: recalculate by fitness value of antibody each in antibody population after above-mentioned mutation operation i.e. the size of affinity, if by the affinity of antibody after above-mentioned mutation operation higher than A
rin the affinity of antibody before corresponding variation, replace original A with regard to the antibody after mutation operation
rin the antibody that do not make a variation, form memory antibody collection A
m;
4f) antibody memory: simulate naturally withering away of the B cell of 5% in artificial immune system bioselection, namely delete from antibody collection A affinity minimum 5% antibody, reinitialize the new antibodies of generation 5%, with the antibody of 5% of new generation, substitute the antibody that in antibody collection A, part affinity is minimum.
Step 5, judges whether to meet evolution conditions: if iterations does not reach predefined evolutionary generation gene, then jump to 4a) continue to perform, otherwise, perform step 6;
Step 6, when iterations reaches predefined evolutionary generation gene, the individuality that affinity is the highest is optimum antibody, the conversion parameter representated by optimum antibody is utilized to carry out spatial alternation to floating image imageS, calculate and output transform after floating image and reference picture imageR between association relationship, carry out registration with reference to the floating image after image imageR and conversion, finally export the image after registration.
Effect of the present invention can be further illustrated by following emulation:
1. simulated conditions
This example, under Intel (R) Core (TM) 2DuoCPU2.33GHzWindows7 system, on Matlab2014a operation platform, completes the present invention and existing methodical emulation experiment.
2. emulation experiment content
Choose respectively and carry out Experimental comparison by RadarSat-2 satellite in the Yellow River estuary image two width cut-away view picture (size is 400*400) of 2008 and acquisition in 2009, its real registration parameter is all unknown.In order to verify performance of the present invention, method of the present invention and existing several optimized algorithm being compared analysis, is DE (differential evolution algorithm), PSO (particle cluster algorithm) and GA (genetic algorithm) respectively.In experiment, optimum configurations is as follows: in four kinds of algorithms, Population Size popsize is 350, and iterations is 30, in CSA algorithm of the present invention, and β=0.16, mutation probability P
min=0.1, DE algorithm, mutagenic factor FD is set to FD=0.7, and crossover probability CR is set to CR=0.5, in PSO algorithm, weights omega is set to from 0.9 to 0.4 linear decrease, and arranges c1=c2=2.0, in GA algorithm, mutation probability M is set to 0.05, and crossover probability is set to 0.8.First the result of registration is evaluated by the size of similarity measurement normalized mutual information, close to 1, its value, between 0 to 1, more represents that registration effect is better.Next is the result quality judging registration according to Visual Observations Observations.Following CSA represents the method for registering images based on Immune Clone Selection of the present invention, and DE represents the method for registering images based on differential evolution algorithm, and PSO represents the method for registering images based on particle cluster algorithm, and GA represents the method for registering images based on genetic algorithm.
(c), (d), (e) and (f) of Fig. 2 is CSA, DE, PSO and GA experimental result picture to first group of data respectively, can obviously find out, image can be carried out registration by the solution of trying to achieve of CSA more accurately, in CSA operation result figure, the complete seamless coincidence of black part of top.And PSO registration effect is the poorest, obvious dislocation can be seen.Figure 3 shows that four kinds of optimized algorithms are tested first group of data respectively, often kind of algorithm runs 20 times, and the mutual information mean value getting 20 every generations is drawn as above-mentioned curve.Can obviously be found out by Fig. 3, CSA can obtain higher fitness value compared to other three kinds of optimized algorithms, thus obtains more accurate registration result.
As shown in Figure 3, for the evolution curve of first group of HUANGHE ESTUARY image, four kinds of algorithm registrations, CSA algorithm starts just to show high association relationship in the first generation, and its association relationship is also maximum when final convergence.Although the convergence of DE algorithm early, association relationship is but very little, and therefore registration result is also poor.The effect of GA and PSO is between CSA and DE, and both are better relative to effect DE, but still not as good as CSA, as can be seen from curve, CSA shows good stability and convergence for asking during mutual information.
Have recorded four kinds of algorithms carry out the maximum, minimum of the fitness value of iteration optimization and mean value situation of change to image in Table 1 in detail.Can find out: the maximum adaptation angle value of CSA (clonal selection algorithm) and average fitness value are all higher than other three kinds of algorithms, next is GA (genetic algorithm), then be PSO (particle cluster algorithm), compare, for this width figure, the effect of DE (differential evolution algorithm) is relatively not ideal, mainly because remote sensing images have noise and differential evolution algorithm for caused by noise-sensitive.From above-mentioned chart, we obviously can find out that CSA (clonal selection algorithm) has very high adaptability and validity.
Table 1
| Algorithm | MI(best) | MI(worst) | MI(average) |
| CSA | 0.6314 | 0.4636 | 0.5770 |
| DE | 0.4241 | 0.1967 | 0.3703 |
| PSO | 0.5602 | 0.3316 | 0.4926 |
| GA | 0.6093 | 0.3546 | 0.5374 |
(c), (d), (e) and (f) of Fig. 4 is CSA, DE, PSO and GA experimental result picture to second group of data respectively, can find out, image can be carried out registration by solution that CSA tries to achieve more accurately, in CSA operation result figure, the complete seamless coincidence of upper right corner black line part, and in figure, dark trapezoidal region is also aimed at accurately.DE, PSO and GA all have various obvious out of true performance, PSO and GA upper right corner black line part is not all aimed at completely.Figure 5 shows that four kinds of optimized algorithms are tested the 3rd group of data respectively, often kind of algorithm runs 20 times, and the mutual information mean value getting 20 every generations is drawn as above-mentioned curve.As seen from Figure 5, CSA can obtain higher fitness value compared to other three kinds of optimized algorithms, thus obtains more accurate registration result.
As shown in Figure 5, for the evolution curve of second group of data HUANGHE ESTUARY image, four kinds of algorithm registrations, four kinds of algorithm the convergence speed are all very fast, and all very stable convergence, but the final association relationship of CSA and fitness value are a little more than its excess-three kind algorithm, and all algorithms all reach optimal value about 15 generations.Can not only Fast Convergent from above-mentioned figure: CSA, and the precision of image registration can also be improved while ensureing speed of convergence.
Have recorded the situation of change of image being carried out to the maximum, minimum of the fitness value of iteration optimization and mean value of four kinds of algorithms in table 2.Can find out: the maximum adaptation angle value of CSA (clonal selection algorithm) and average fitness value are all higher than other three kinds of algorithms, next is GA (genetic algorithm), then be PSO (particle cluster algorithm) and DE (differential evolution algorithm), compare, obviously can find out that CSA (clonal selection algorithm) has higher adaptability and validity.
Table 2
| Algorithm | MI(best) | MI(worst) | MI(average) |
| CSA | 0.6178 | 0.4404 | 0.5936 |
| DE | 0.4172 | 0.2164 | 0.3870 |
| PSO | 0.5828 | 0.4537 | 0.5524 |
| GA | 0.5874 | 0.4216 | 0.5560 |
In a word, the present invention regards image registration problem the optimization problem of function as, using normalized association relationship as objective function, the clonal selection algorithm based on immunity is adopted to be optimized objective function, the shortcoming that classic method is easily absorbed in local optimum can be overcome, and from the convergence curve of registration result figure and mutual information function, the present invention is better than existing four kinds of control methodss, and this also demonstrates validity of the present invention and robustness.
Claims (6)
1. based on a method for registering images for real coding clonal selection algorithm, it is characterized in that, described method comprises the following steps:
(1) input resolution identical reference picture imageR and floating image imageS, and judge whether imageR and imageS is gray level image, if not, be then converted into gray level image;
(2) antibody population initialization: produce antibody population at random, A represents antibody collection, and antibody collection A is by interim antibody collection A
rwith memory antibody collection A
mcomposition, N=r+m, r represent interim antibody collection A
rin interim antibody levels, m represents memory antibody collection A
min memory antibody quantity, N is the total quantity of antibody in antibody collection A;
(3) represent reference picture imageR, B with A and represent floating image imageS structure normalized mutual information function MI (A, B) as objective function:
Wherein, H (A) is the edge entropy of image A,
p
ai () is the probability that in image A, i-th grade of gray level occurs, M is the progression of gray level in image A, the edge entropy that H (B) is image B,
p
bj () is the probability that in image B, jth level gray level occurs, N is the progression of gray level in image B,
be the combination entropy between image A and image B, p (i, j) is jth level gray level this pair joint probability that same position occurs simultaneously in this two width image of image A and image B in i-th grade of gray level in image A, image B, and
h (i, j) is the joint histogram of image A and image B, and n is the number of image A and image B overlapping region pixel;
(4) utilize clonal selection algorithm to be optimized normalized mutual information function MI (A, B), specifically comprise:
4a) calculate the affinity of antibody of initial population, i.e. the value of normalized mutual information function MI (A, B);
4b) antibody is selected: antagonist affinity, according to descending sort, chooses front k the highest individuality of affinity of antibody as interim antibody collection A
r;
4c) antibody cloning: according to the size of affinity, clones k the individuality that affinity is the highest chosen;
4d) antibody variation: mutation operation is done to the antibody after clone;
4e) form memory antibody collection: recalculate by fitness value of antibody each in antibody population after above-mentioned mutation operation i.e. the size of affinity, if by the affinity of antibody after above-mentioned mutation operation higher than A
rin the affinity of antibody before corresponding variation, just replace original A with the antibody after mutation operation
rin the antibody that do not make a variation, form memory antibody collection A
m;
4f) antibody memory: delete from antibody collection A affinity minimum 5% antibody, the new antibodies of the generation 5% that reinitializes, with new produce 5% antibody, substitute the minimum antibody of affinity of in antibody collection A deleted 5%;
(5) judge whether to meet evolution conditions: if iterations does not reach predefined evolutionary generation gene, then jump to 4a) continue to perform, otherwise, perform (6);
(6) when iterations reaches predefined evolutionary generation gene, the individuality that affinity is the highest is optimum antibody, the conversion parameter representated by optimum antibody is utilized to carry out spatial alternation to floating image imageS, obtain the floating image after converting, calculate and output transform after floating image and reference picture imageR between association relationship, carry out registration with reference to the floating image after image imageR and conversion, finally export the image after registration.
2. the method for registering images based on clonal selection algorithm according to claim 1, wherein, the random generation antibody population described in step (2) comprises the steps:
(2a) determine the scope of the affined transformation anglec of rotation, namely according to the size of input picture, determine the scope of affined transformation displacement in the x and y direction;
(2b) adopt real number coding method, in step (2a) determined scope, produce antibody population at random.
3. the method for registering images based on clonal selection algorithm according to claim 1, wherein, described in step (3), joint histogram is defined as follows:
H (a, b) represents same position in this two width image of image A and image B, and in image A, gray level is a, in image B, gray level is the number that the pixel of b is right, wherein, and a≤0≤K-1,0≤b≤L-1, K and L is the scope of the gray level of pixel in image A and image B respectively.
4. the method for registering images based on clonal selection algorithm according to claim 1, wherein step 4c) according to the size of affinity, k the specific rules that the individuality that affinity is the highest is cloned selected is comprised: affinity is higher, affinity sequence number i is less, and the number of times be cloned is more, on the contrary, affinity is lower, affinity sequence number i is larger, is cloned number of times fewer, the total quantity N that antibody is cloned
cas shown in the formula:
Wherein, β is a multiplication factor, and N is the antibody total quantity in collection of antibodies A, and the span of affinity sequence number i is 1 is an independent variable operator to k, round (), represents to nearest integer and rounds, N
ci-th
the scale that corresponding i-th antibody is cloned.
5. the method for registering images based on clonal selection algorithm according to claim 1, wherein step 4d) mutation operation is done to the antibody after clone, comprise the steps:
Random number between producing one 0 to 1, if this number is less than the mutation probability preset, general mutation probability between 0.01 to 0.1, then makes a variation to the antibody x after clone:
x
new＝x+δ*λ
Wherein, x
newfor the parameter after variable x variation, δ=0.1* (x
max-x
min) be the region of search factor that narrows, x
maxfor the maximal value of variable x, x
minfor the minimum value of variable x, λ is a random number within the scope of 0 to 1 open interval, for the step-length regulating the antibody x after clone to make a variation.
6. the method for registering images based on clonal selection algorithm according to claim 1, wherein, k is equal with the quantity r of interim antibody collection, namely gets k=r.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510229255.XA CN105184764B (en) | 2015-05-07 | 2015-05-07 | A kind of method for registering images based on real coding clonal selection algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510229255.XA CN105184764B (en) | 2015-05-07 | 2015-05-07 | A kind of method for registering images based on real coding clonal selection algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105184764A true CN105184764A (en) | 2015-12-23 |
| CN105184764B CN105184764B (en) | 2018-06-15 |
Family
ID=54906818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510229255.XA Active CN105184764B (en) | 2015-05-07 | 2015-05-07 | A kind of method for registering images based on real coding clonal selection algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105184764B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107300562A (en) * | 2017-05-19 | 2017-10-27 | 温州大学 | A kind of X-ray lossless detection method of measuring relay finished product contact spacing |
| CN111797903A (en) * | 2020-06-12 | 2020-10-20 | 武汉大学 | Multi-mode remote sensing image registration method based on data-driven particle swarm optimization |
| CN116152316A (en) * | 2023-04-17 | 2023-05-23 | 山东省工业技术研究院 | Image registration method based on self-adaptive parameter particle swarm algorithm |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103091676A (en) * | 2013-01-22 | 2013-05-08 | 中国矿业大学 | Mining area surface subsidence synthetic aperture radar interferometry monitoring and calculating method |
| US8731334B2 (en) * | 2008-08-04 | 2014-05-20 | Siemens Aktiengesellschaft | Multilevel thresholding for mutual information based registration and image registration using a GPU |
-
2015
- 2015-05-07 CN CN201510229255.XA patent/CN105184764B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8731334B2 (en) * | 2008-08-04 | 2014-05-20 | Siemens Aktiengesellschaft | Multilevel thresholding for mutual information based registration and image registration using a GPU |
| CN103091676A (en) * | 2013-01-22 | 2013-05-08 | 中国矿业大学 | Mining area surface subsidence synthetic aperture radar interferometry monitoring and calculating method |
Non-Patent Citations (1)
| Title |
|---|
| 李爽: "基于克隆选择算法的PET-CT医学图像融合的实现", 《万方硕士学位论文数据库》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107300562A (en) * | 2017-05-19 | 2017-10-27 | 温州大学 | A kind of X-ray lossless detection method of measuring relay finished product contact spacing |
| CN111797903A (en) * | 2020-06-12 | 2020-10-20 | 武汉大学 | Multi-mode remote sensing image registration method based on data-driven particle swarm optimization |
| CN116152316A (en) * | 2023-04-17 | 2023-05-23 | 山东省工业技术研究院 | Image registration method based on self-adaptive parameter particle swarm algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105184764B (en) | 2018-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223324B (en) | Target tracking method of twin matching network based on robust feature representation | |
| CN109583425B (en) | Remote sensing image ship integrated recognition method based on deep learning | |
| CN106780485B (en) | SAR image change detection method based on super-pixel segmentation and feature learning | |
| CN110263705A (en) | Towards two phase of remote sensing technology field high-resolution remote sensing image change detecting method | |
| CN101980250B (en) | Method for identifying target based on dimension reduction local feature descriptor and hidden conditional random field | |
| CN103295239B (en) | A kind of autoegistration method of the laser point cloud data based on datum plane image | |
| CN106875395A (en) | Super-pixel level SAR image change detection based on deep neural network | |
| CN102750696B (en) | Affine invariant feature and coastline constraint-based automatic coastal zone remote-sensing image registration method | |
| CN105809693A (en) | SAR image registration method based on deep neural networks | |
| CN109766936A (en) | Image change detection method based on information transmitting and attention mechanism | |
| CN112949416B (en) | Supervised hyperspectral multiscale graph volume integral classification method | |
| CN111766625B (en) | Seismic background noise dispersion curve extraction method based on deep learning | |
| CN110287770B (en) | Water individual target matching identification method based on convolutional neural network | |
| CN114022729A (en) | Heterogeneous image matching positioning method and system based on twin network and supervised training | |
| CN104680151B (en) | A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account | |
| CN110334656A (en) | Multi-source Remote Sensing Images Clean water withdraw method and device based on information source probability weight | |
| CN105184764A (en) | Image registering method in real number coding based clonal selection algorithm | |
| CN109766838A (en) | A kind of gait cycle detecting method based on convolutional neural networks | |
| CN115099500A (en) | Water level prediction method based on weight correction and DRSN-LSTM model | |
| CN105095913A (en) | Remote sensing image classification method and system based on neighbor regular joint sparse representation | |
| CN111832404B (en) | Small sample remote sensing ground feature classification method and system based on feature generation network | |
| CN115359366A (en) | Remote sensing image target detection method based on parameter optimization | |
| CN114913434B (en) | High-resolution remote sensing image change detection method based on global relation reasoning | |
| CN114283285A (en) | Cross consistency self-training remote sensing image semantic segmentation network training method and device | |
| CN117313516A (en) | Fermentation product prediction method based on space-time diagram embedding |
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 |