CN101833670A - Image matching method based on lateral inhibition and chaos quantum particle swarm optimization - Google Patents

Abstract

The invention relates to an image matching method based on lateral inhibition and chaos quantum particle swarm optimization, comprising the following steps of: (1) pre-processing images by using lateral inhibition principles, and extracting image edges; (2) initiating each parameter of particle swarm optimization; (3) calculating the fitness value of each particle; (4) comparing the fitness value of each particle of the cycle, i.e. the kth cycle, and the fitness value of each particle obtained from the former step, wherein the maximal fitness value is the global optimization gbest; simultaneously comparing the fitness value of each particle obtained from the former step respectively with the fitness value of the particle obtained in the (k-1)th cycle, wherein the greater fitness value is the optimization Pbest of the particle self of the kth cycle; (5) carrying out particle swarm optimization according to the fitness values of the particles; (6) skipping to the step 3 when the current cycle time plus 1, repeating the operation till the cycle time is greater than the maximum cycle time N; and (7) ending the operation and outputting the optimized allocation position and the optimized fitness value.

Description

A kind of image matching method based on lateral inhibition and chaos quantum particle group optimization

[technical field]

The present invention a kind ofly optimizes based on lateral inhibition and chaos quantum particle group that (Chaotic mutated Quantum-BehavedPSO Based on Lateral Inhibition, image matching method LICQPSO) belongs to the computer vision field of information processing.

[background technology]

Images match is very important research topic in the computer vision, is widely used in the fields such as image registration, target detection, Target Recognition and image retrieval.Images match is meant by certain matching process discerns same place between two width of cloth or multiple image, its essence is under the condition of primitive similarity, obtains the most similar zone according to certain matching criterior search.Images match a kind of method commonly used is based on the coupling of gray scale.Good coupling precision prescribed height, a speed is fast, and this also is vital measurement factor in computer vision field.Studies show that the speed of images match depends primarily on the search strategy of matching process.A good search strategy does not need each pixel on the traversing graph picture just can find the Optimum Matching position accurately.We can say that the pixel of searching under the situation that guarantees matching precision is few more, the efficient of method is also just high more.

(Lateral Inhibition is that people such as Hartline find when the king crab vision is carried out electric Physiological Experiment and confirm LI), is one of cardinal rule of nervous system information processing in lateral inhibition.To studies show that of king crab vision system, lateral inhibition network major function in the information pre-service has the following aspects:

(1) can projecting edge, enhanced contrast;

(2) Hi-pass filter is compressed to very big input variation range within the dynamic range of network itself, and tangible luminance adaptation effect is arranged;

(3) can carry out match to the trickle discontinuities of image, have tangible cluster effect.

Lateral inhibition plays an important role in neural information processing, can be applicable to fields such as Flame Image Process, pattern-recognition and neural network.

(Particle Swarm Optimization, PSO) algorithm is a kind of evolutionary computing that is equaled the nineteen ninety-five exploitation by J.Kennedy and R.C.Eberhart to particle swarm optimization, derives from the simulation of simplifying social model to one.PSO is the uncertain motion for patterned simulation flock of birds is graceful at first.And, find in colony, the social sharing of information to be provided the advantage of an evolution, and with this basis as development approach by observation to animal social behavior.

PSO is based on colony, the zone that the individuality in the colony is moved to according to the fitness to environment.Regard each individuality as D tie up a particulate (point) that does not have volume in the search volume, with certain speed flight, this speed is dynamically adjusted according to the flying experience of itself and companion's flying experience in the search volume.I particulate is expressed as x _i=(x _I1..., x _Id), the desired positions that it lives through (best adaptive value is arranged) is designated as P _i=(p _I1..., p _Id), be also referred to as P _BestThe call number of the desired positions that lives through at all particulates of colony represents with symbol g, i.e. P _g, be also referred to as g _BestThe speed V of particulate i _i=(v _I1..., v _Id) expression.To each generation, its d dimension (1≤d≤D) change according to following equation:

$v_i^{k+1}=w^{*}{v}_i^k+{\mathrm{<figure-callout\; id="1"\; label="C"\; filenames="HSA00000109981400011.png"\; state="\{\{state\}\}">C</figure-callout>}}_1{\mathrm{rand}}_1\left(\right)(p_i^k-x_i^k)+C_2\mathrm{ran}{d}_2\left(\right)(p_g^k-x_i^k)---\left(1\right)$

$x_i^{k+1}=x_i^k+v_i^{k+1}---\left(2\right)$

Wherein k is an iterations, and w is inertia weight (inertia weight), C ₁And C ₂Be aceleration pulse (acceleration constants), rand ₁() and rand ₂() is two random values that change in [0,1] scope.

Inertia weight w makes particulate keep the inertia of motion, makes its trend that expanded search space is arranged, and has the ability to explore new zone.Aceleration pulse C ₁And C ₂Representative is pushed each particulate to P _BestAnd g _BestThe weight of the statistics acceleration term of position.In addition, the speed v of particulate _iBy a maximal rate v _MaxLimit.v _MaxThe resolution (or precision) in the zone of decision between current location and desired positions.If current acceleration to particulate causes its speed v in certain dimension _IdThe maximal rate v that surpasses this dimension _Maxd, then the speed of this dimension then is set again.

To formula (1), first is the inertia of the previous behavior of particulate, and second portion is " cognitive (cognition) " part, the thinking of expression particulate itself; Third part is " society (social) " part, the information sharing between the expression particulate with cooperate with each other.

PSO is similar with genetic algorithm, is a kind of optimization tool based on iteration, it is advantageous that to realize simply easily and do not have many parameters to need to adjust.Fields such as function optimization, neural metwork training, fuzzy system control have been widely used at present.

Chaos (chaos) is a kind of non-linear phenomena that nature extensively exists, and has characteristics such as randomness, ergodicity, starting condition susceptibility, has been widely used in random optimization, has superior performance in the local optimal searching field.The chaotic maps Logistic iterative equation that uses is:

x(k+1)＝μx(k)(1-x(k)) (3)

And as x (1) ≠ 0.25,0.5,0.75, μ=4 o'clock, equation enters chaos state fully.Can prove through test of many times, utilize the ergodicity of chaos, can obtain good particulate original state.

Quantum particle group algorithm (Quantum-behaved PSO, QPSO) be Sun in 2004 etc. after having studied the achievement in research of people such as Clerc about the particulate convergence behavior, a kind of new PSO model that proposes from quantum-mechanical angle.In vector subspace, the character that satisfies aggregative state of particulate is different fully, and it can be searched in whole feasible solution space, so the performance of the global search of QPSO is far superior to pure PSO.In vector subspace, the speed of particulate and position can not be determined simultaneously.Describe the state of particulate by crossing function, and obtain the particulate probability density function that a certain point occurs in the space, obtain the position equation of particulate again by Monte Carlo stochastic simulation mode by finding the solution schrodinger equation.In having the particle swarm optimization algorithm of quantum behavior, the main iterative formula of particulate is:

$\mathrm{mbest}=\frac{1}{M}\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{p}_i---\left(4\right)$

p(j，i) _d＝φp(j，i) _pd+(1-φ)p(k) _gd (5)

If rand ()＞0.5

$x_{\mathrm{id}}(t+1)=p_{\mathrm{id}}-\mathrm{\&beta;}|\mathrm{mbest}-x\left(t\right)|\mathrm{In}\frac{1}{u}---\left(6\right)$

Otherwise

$x_{\mathrm{id}}(t+1)=p_{\mathrm{id}}+\mathrm{\&beta;}|\mathrm{mbest}-x\left(t\right)|\mathrm{In}\frac{1}{u}---\left(7\right)$

Wherein, mbest is particulate group P _BestThe centre position, p _IdBeg for and be p _PdAnd p _GdBetween random point.φ, u and rand () are the random numbers of [0,1], and β is the shrinkage expansion coefficient of QPSO.

[summary of the invention]

One, goal of the invention:

The present invention proposes a kind of image matching method, its objective is the efficient and the accuracy rate that improve images match, also can solve in other complicated intelligent optimization problem simultaneously based on lateral inhibition and chaos quantum particle group optimization (LICQPSO).

Based on lateral inhibition and chaos quantum particle group optimizing method is a kind of method that side inhibition principle, chaotic maps and quantum particle group algorithm are combined.This method combines the advantage of side inhibition principle, chaotic maps and quantum particle group optimizing method, having made full use of lateral inhibition can projecting edge in the image pre-service, the global convergence of the characteristic of enhanced contrast and high-pass filtering and the ergodicity of chaotic maps and quantum particle group optimizing method.

Two, technical scheme:

A kind of image matching method of the present invention based on lateral inhibition and chaos quantum particle group optimization, its concrete steps are as follows:

Step 1: the utilization side inhibition principle is carried out the image pre-service, extracts the image border;

1.1 image read and be converted to gray level image;

According to the transition matrix between the image of different-format, the image of different-format is converted to gray level image earlier, obtain each gray values of pixel points in the gray level image, be convenient to next step and extract the image border.

1.2 extract the image border of former figure and template to be matched;

Select suitable lateral inhibition model, suppress the open country and the lateral inhibition coefficient distributes, treat matching template and former figure extracts the image border.

The lateral inhibition model that the present invention selects for use is as follows:

$R(m,n)=I_0(m,n)+\underset{i=-M}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=-N}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{ij}}{I}_0(m+i,n+j)---\left(8\right)$

Wherein, I ₀(m n) is former pixel (m, the gray-scale value of n) locating; α _IjFor in the network (i, j) the position pixel is to the inhibition coefficient of competition of center pixel; (m n) is former pixel (m, the gray-scale value of n) locating after lateral inhibition is handled to R; I ₀(m+i is that (M * N is for suppressing wild for m+i, the gray-scale value of n+j) locating for former pixel n+j).

It is as follows to suppress the open country and be 5 * 5 lateral inhibition model:

$R(m,n)={\mathrm{\&alpha;}}_0\&times;I_0(m,n)-{\mathrm{\&alpha;}}_1[\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)-I_0(m,n)]$

${-\mathrm{\&alpha;}}_2[\underset{i=-2}{\overset{2}{\mathrm{\&Sigma;}}}\underset{i=-2}{\overset{2}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)-\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=-1}{\overset{1}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)]---\left(9\right)$

In order to guarantee that the lateral inhibition coefficient is 0 substantially, requirement

α ₀-8α ₁-16α ₂＝0 (10)

Choose α among the present invention ₀=1, α ₁=0.075, α ₂=0.025, then the lateral inhibition coefficient of Gou Chenging is distributed as:

$U=\left[\begin{array}{ccccc}0.025& 0.025& 0.025& 0.025& 0.025\\ 0.025& 0.075& 0.075& 0.075& 0.025\\ 0.025& 0.075& 1& 0.075& 0.025\\ 0.025& 0.075& 0.075& 0.075& 0.025\\ 0.025& 0.025& 0.025& 0.025& 0.025\end{array}\right]---\left(11\right)$

Formula below utilizing transforms the gray-scale value of the gray level image after handling through formula (9), becomes gray-scale value and only be 0 or 255 binary map, and wherein T is a preset threshold, and the selected T of the present invention is 120.

$I(m,n)=\left\{\begin{array}{cc}0& R(m,n)\&le;T\\ 255& R(m,n)>T\end{array}\right.---\left(12\right)$

Step 2: initialization particle swarm optimization (Particle Swarm Optimization, PSO) each parameter

2.1 initialization particulate group

Set size and represent the particulate group of PSO for the array P of M * 2, wherein M is that this particulate group's scale is the particulate sum.Each columns value of array P is horizontal ordinate and the ordinate of correspondence image pixel in image coordinate system respectively.In order to reach the effect that particulate spreads all over the entire image space, adopt following chaotic maps (Logistic) iterative equation:

x(k+1)＝μx(k)(1-x(k)) (13)

Wherein, k is the current iteration number of times, and μ is a controlled variable, and x is the chaos vector.When x (1) ≠ 0.25,0.5,0.75, and μ=4 o'clock, equation (13) enters chaos state fully.Can prove that through test of many times the iterative solution of this chaotic maps iterative equation has ergodicity, this method of utilization can obtain good particulate original state in initialization.

The initialization of particulate group P is as follows:

P(m+1，i)＝4P(m，i)(1-P(m，i)) (14)

I=1 wherein, 2; M=1 ..., M; P (1, i) ≠ 0.25,0.5,0.75

When m=1, P (1,1), P (1,2) get the random number of [1, a], [1, b] respectively, and a and b are respectively the length of image to be matched and wide, and unit is a pixel.Iterating by formula (14) obtains whole values of array P, is the original state of particulate group P.

2.2 establish the iterations of the maximum of LICQPSO is that cycle index is N, and current cycle index is k.

2.3 with chaotic maps iterative equation initialization shrinkage expansion factor beta

The common obtaining value method of β is the linear decrease method, adopts the β after improving among the present invention, adopts chaotic maps (Logistic) iterative equation to shine upon the two-dimensional array that obtains with the iterations variation at the graininess of each dimension, and k is current cycle index.

β _i(k+1)＝4β _i(k)(1-β _i(k))， (15)

I=1 wherein, 2; β _i(1) ≠ 0.25,0.5,0.75

Step 3: the fitness value that calculates each particulate

Image coordinate is that (m, (m n), next can be calculated the fitness value of each particulate by following formula (16) to have become I after the lateral inhibition of the gray-scale value of n) locating process step 1 is handled in the step 1.

The present invention adopts following fitness function, calculates the fitness value of each particulate

$f=\underset{i=0}{\overset{K-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{W-1}{\mathrm{\&Sigma;}}}I(m+i,n+j)---\left(16\right)$

Wherein, K * W is the size of template to be matched, and f is a fitness value, and (m+i is that bianry image after formula (12) is handled is at pixel (m+i, gray-scale value n+j) n+j) to I.

Step 4: relatively this time circulation is the fitness value of k each particulate of circulation time, i.e. the fitness value of each particulate that obtains by previous step, and wherein maximum adaptation degree value is global optimum g _BestThe fitness value of this particulate of obtaining with the k-1 time circulation time respectively of the fitness value of each particulate that previous step is obtained compares simultaneously, and bigger fitness value is the optimal value P of this particulate self of the k time circulation time _Best

Step 5: the fitness value according to particulate carries out the PSO optimizing

According to following formula calculate mbest and p (m, i) _d:

$\mathrm{mbest}\left(i\right)=\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}P(j,i),i=\mathrm{1,2}---\left(17\right)$

P(m，i) _d＝φP(m，i) _pd+(1-φ)P(k，i) _gd，m＝1，…，M (18)

Wherein φ is the random number of [0,1], and m is a j particulate.P (m, i) _PdBe m the desired positions that particulate lives through, and P (k, i) _GdThe desired positions of particulate group when being k for iterations.

If rand ()＞0.5

$P(m,i)=P{(m,i)}_d-\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(19\right)$

Otherwise

$P(m,i)=P{(m,i)}_d+\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(20\right)$

Wherein rand (), u are the random number of [0,1].Current particulate is carried out iterative computation, and the particulate P of generation a new generation (m, i).

The fitness function that is provided by formula (16) calculates the fitness value of new particulate then, and compares with the pairing former generation particulate fitness value of working as respectively.Note all and calculate back fitness values better particulate that becomes at experience PSO.

Step 6: current cycle time adds 1, and promptly k=k+1 jumps to step 3, rerun, and (be k＞N) until cycle index greater than maximum cycle N.

Step 7: finish operation, and output Optimum Matching position and optimal-adaptive degree value.

Three, advantage and effect:

The present invention proposes a kind of method of optimizing images match based on lateral inhibition and chaos quantum particle group, wherein side inhibition principle energy projecting edge, enhanced contrast, feasible coupling is more accurate, this method combines the ergodicity of chaotic maps and the global convergence and the high efficiency of quantum particle group algorithm simultaneously, and a kind of effective way of images match is provided.It also can be widely used in field and other the complicated intelligent optimization problem that robot, Aeronautics and Astronautics, commercial production etc. relate to Image Information Processing.

[description of drawings]

Fig. 1 LICQPSO process flow diagram

Fig. 2 LICQPSO iteration curve

Number in the figure and symbol description are as follows:

Fitness Value---particulate fitness value

K---current cycle time

N---maximum cycle

F (i)---the fitness value of the k time circulation time particulate i

The optimal-adaptive degree value of f (Pi)---particulate i

F (Pg)---the k time circulation time global optimum fitness value

Yes---(being) satisfies condition

No---(denying) do not satisfy condition

[embodiment]

As depicted in figs. 1 and 2, a kind of image matching method based on lateral inhibition and chaos quantum particle group optimization, its specific implementation step is as follows:

Step 1: side inhibition principle is extracted the image border to the image pre-service.

1.1 image read and be converted to gray level image;

According to the transition matrix between the image of different-format, the image of different-format is converted to gray level image earlier, obtain each gray values of pixel points in the gray level image, be convenient to next step and extract the image border.

1.2 extract the image border of former figure and template to be matched;

Select suitable lateral inhibition model, suppress the open country and the lateral inhibition coefficient distributes, treat matching template and former figure extracts the image border.

The lateral inhibition model that the present invention selects for use is as follows:

$R(m,n)=I_0(m,n)+\underset{i=-M}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=-N}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{ij}}{I}_0(m+i,n+j)---\left(8\right)$

Wherein, I ₀(m n) is former pixel (m, the gray-scale value of n) locating; α _IjFor in the network (i, j) the position pixel is to the inhibition coefficient of competition of center pixel; (m n) is former pixel (m, the gray-scale value of n) locating after lateral inhibition is handled to R; I ₀(m+i is that (M * N is for suppressing wild for m+i, the gray-scale value of n+j) locating for former pixel n+j).

It is as follows to suppress the open country and be 5 * 5 lateral inhibition model:

$R(m,n)={\mathrm{\&alpha;}}_0\&times;I_0(m,n)-{\mathrm{\&alpha;}}_1[\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)-I_0(m,n)]$

${-\mathrm{\&alpha;}}_2[\underset{i=-2}{\overset{2}{\mathrm{\&Sigma;}}}\underset{i=-2}{\overset{2}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)-\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=-1}{\overset{1}{\mathrm{\&Sigma;}}}{I}_0(m+i,n+j)]---\left(9\right)$

In order to guarantee that the lateral inhibition coefficient is 0 substantially, requirement

α ₀-8α ₁-16α ₂＝0 (10)

Choose α among the present invention ₀=1, α ₁=0.075, α ₂=0.025, then the lateral inhibition coefficient of Gou Chenging is distributed as:

$U=\left[\begin{array}{ccccc}0.025& 0.025& 0.025& 0.025& 0.025\\ 0.025& 0.075& 0.075& 0.075& 0.025\\ 0.025& 0.075& 1& 0.075& 0.025\\ 0.025& 0.075& 0.075& 0.075& 0.025\\ 0.025& 0.025& 0.025& 0.025& 0.025\end{array}\right]---\left(11\right)$

Formula below utilizing transforms the gray-scale value of the gray level image after handling through formula (9), becomes gray-scale value and only be 0 or 255 binary map, and wherein T is a preset threshold, and the selected T of the present invention is 120.

$I(m,n)=\left\{\begin{array}{cc}0& R(m,n)\&le;T\\ 255& R(m,n)>T\end{array}\right.---\left(12\right)$

Step 2: initialization particulate swarm parameter

2.1 initialization particulate group

Set size and represent the particulate group of PSO for the array P of M * 2, wherein M is 150.Each columns value of array P is horizontal ordinate and the ordinate of correspondence image pixel in image coordinate system respectively.In order to reach the effect that particulate spreads all over the entire image space, adopt following chaotic maps (Logistic) iterative equation:

x(k+1)＝μx(k)(1-x(k)) (13)

Wherein, k is the current iteration number of times, and μ is a controlled variable, and x is the chaos vector.When x (1) ≠ 0.25,0.5,0.75, and μ=4 o'clock, equation (13) enters chaos state fully.Can prove that through test of many times the iterative solution of this chaotic maps iterative equation has ergodicity, this method of utilization can obtain good particulate original state in initialization.

The initialization of particulate group P is as follows:

P(m+1，i)＝4P(m，i)(1-P(m，i)) (14)

I=1 wherein, 2; M=1 ..., M; P (1, i) ≠ 0.25,0.5,0.75

When m=1, P (1,1), P (1,2) get the random number of [1, a], [1, b] respectively, and a and b are respectively the length of image to be matched and wide, and unit is a pixel.Iterating by formula (14) obtains whole values of array P, is the original state of particulate group P.

2.2 establish the iterations of the maximum of LICQPSO is that cycle index is N=80, and current cycle index is k.

2.3 with chaotic maps iterative equation initialization shrinkage expansion factor beta

The common obtaining value method of β is the linear decrease method, adopts the β after improving among the present invention, adopts chaotic maps (Logistic) iterative equation to obtain the two-dimensional array that iterations changes at the graininess mapping of each dimension, and k is current cycle index.

β _i(k+1)＝4β _i(k)(1-β _i(k))， (15)

I=1 wherein, 2; β _i(1) ≠ 0.25,0.5,0.75

Step 3: the fitness value that calculates each particulate

Image coordinate is that (m, (m n), next can be calculated the fitness value of each particulate by following formula (16) to have become I after the lateral inhibition of the gray-scale value of n) locating process step 1 is handled in the step 1.

The present invention adopts following fitness function, calculates the fitness value of each particulate

$f=\underset{i=0}{\overset{K-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{W-1}{\mathrm{\&Sigma;}}}I(m+i,n+j)---\left(16\right)$

Wherein, K * W is the size of template to be matched, and f is a fitness value, and (m+i is that bianry image after formula (12) is handled is at pixel (m+i, gray-scale value n+j) n+j) to I.

Step 4: relatively this time circulation is the fitness value of K each particulate of circulation time, i.e. the fitness value of each particulate that obtains by previous step, and wherein maximum adaptation degree value is global optimum g _BestThe fitness value of this particulate of obtaining with the K-1 time circulation time respectively of the fitness value of each particulate that previous step is obtained compares simultaneously, and bigger fitness value is the optimal value P of this particulate self of the K time circulation time _Best

Step 5: the fitness value according to particulate carries out optimizing

According to following formula calculate mbest and p (m, i) _d:

$\mathrm{mbest}\left(i\right)=\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}P(j,i),i=\mathrm{1,2}---\left(17\right)$

P(m，i) _d＝φP(m，i) _pd+(1-φ)P(k，i) _gd，m＝1，…，M (18)

Wherein φ is the random number of [0,1], and m is a j particulate.P (m, i) _PdBe m the desired positions that particulate lives through, and P (k, i) _GdThe desired positions of particulate group when being k for iteration time following number.

If rand ()＞0.5

$P(m,i)=P{(m,i)}_d-\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(19\right)$

Otherwise

$P(m,i)=P{(m,i)}_d+\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(20\right)$

Wherein rand (), u are the random number of [0,1].Current particulate is carried out iterative computation, and the particulate P of generation a new generation (m, i).

The fitness function that is provided by formula (16) calculates the fitness value of new particulate then, and compares with the pairing former generation particulate fitness value of working as respectively.Note all and calculate back fitness values better particulate that becomes at experience PSO.

Step 6: current cycle time adds 1, and promptly k=k+1 jumps to step 3, rerun, and (be k＞N) until cycle index greater than maximum cycle N.

Step 7: finish operation, and output Optimum Matching position and optimal-adaptive degree value.

Claims (1)

1. image matching method of optimizing based on lateral inhibition and chaos quantum particle group, it is characterized in that: its concrete steps are as follows:

Step 1: the utilization side inhibition principle is carried out the image pre-service, extracts the image border;

1.1 reading images also is converted to gray level image;

According to the transition matrix between the image of different-format, the image of different-format is converted to gray level image earlier, obtain each gray values of pixel points in the gray level image, to extract the image border;

1.2 extract the image border of former figure and template to be matched;

Select suitable lateral inhibition model, suppress the open country and the lateral inhibition coefficient distributes, treat matching template and former figure extracts the image border;

The lateral inhibition model is as follows:

$R(m,n)=I_0(m,n)+\underset{i=-M}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=-N}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{\mathrm{ij}}{I}_0(m+i,n+j)---\left(8\right)$

Wherein, I ₀(m n) is former pixel (m, the gray-scale value of n) locating; α _IjFor in the network (i, j) the position pixel is to the inhibition coefficient of competition of center pixel; (m n) is former pixel (m, the gray-scale value of n) locating after lateral inhibition is handled to R; I ₀(m+i is that (M * N is for suppressing wild for m+i, the gray-scale value of n+j) locating for former pixel n+j);

Step 2: each parameter of initialization particle swarm optimization

2.1 initialization particulate group

Set size and represent the particulate group of PSO for the array P of M * 2, wherein M is that this particulate group's scale is the particulate sum; Each columns value of array P is horizontal ordinate and the ordinate of correspondence image pixel in image coordinate system respectively; In order to reach the effect that particulate spreads all over the entire image space, adopt following chaotic maps iterative equation:

x(k+1)＝μx(k)(1-x(k)) (13)

Wherein, k is the current iteration number of times, and μ is a controlled variable, and x is the chaos vector; When x (1) ≠ 0.25,0.5,0.75, and μ=4 o'clock, equation (13) enters chaos state fully;

The initialization of particulate group P is as follows:

P(m+1，i)＝4P(m，i)(1-P(m，i)) (14)

I=1 wherein, 2; M=1 ..., M; P (1, i) ≠ 0.25,0.5,0.75

2.2 establish the iterations of the maximum of LICQPSO is that cycle index is N, and current cycle index is k;

2.3 with chaotic maps iterative equation initialization shrinkage expansion factor beta

Adopt the chaotic maps iterative equation to shine upon the two-dimensional array that obtains with the iterations variation at the graininess of each dimension, k is current cycle index;

β _i(k+1)＝4β _i(k)(1-β _i(k))， (15)

I=1 wherein, 2; β _i(1) ≠ 0.25,0.5,0.75

Step 3: the fitness value that calculates each particulate

Image coordinate is that (m, the gray-scale value of n) locating have become I after handling through step 1 lateral inhibition, and (m n), next can be calculated the fitness value of each particulate by following formula (16) in the step 1;

The present invention adopts following fitness function, calculates the fitness value of each particulate

$f=\underset{i=0}{\overset{K-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{W-1}{\mathrm{\&Sigma;}}}I(m+i,n+j)---\left(16\right)$

Wherein, K * W is the size of template to be matched, and f is a fitness value, I (m+i, n+j) for the bianry image after handling at pixel (m+i, gray-scale value n+j);

Step 4: relatively this time circulation is the fitness value of k each particulate of circulation time, i.e. the fitness value of each particulate that obtains by previous step, and wherein maximum adaptation degree value is global optimum g _BestThe fitness value of this particulate of obtaining with the k-1 time circulation time respectively of the fitness value of each particulate that previous step is obtained compares simultaneously, and bigger fitness value is the optimal value P of this particulate self of the k time circulation time _Best

Step 5: the fitness value according to particulate carries out the PSO optimizing

According to following formula calculate mbest and p (m, i) _d:

$\mathrm{mbest}\left(i\right)=\frac{1}{M}\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}P(j,i),i=\mathrm{1,2}---\left(17\right)$

P(m，i) _d＝φP(m，i) _pd+(1-φ)P(k，i) _gd，m＝1，…，M (18)

Wherein φ is the random number of [0,1], and m is a j particulate; P (m, i) _PdBe m the desired positions that particulate lives through, and P (k, i) _GdThe desired positions of particulate group when being k for iterations;

If rand ()＞0.5

$P(m,i)=P{(m,i)}_d-\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(19\right)$

Otherwise

$P(m,i)=P{(m,i)}_d+\mathrm{\&beta;}|\mathrm{mbest}\left(i\right)-P(m,i)|\mathrm{In}\frac{1}{u}---\left(20\right)$

Wherein rand (), u are the random number of [0,1]; Current particulate is carried out iterative computation, and the particulate P of generation a new generation (m, i);

The fitness function that is provided by formula (16) calculates the fitness value of new particulate then, and compares with the pairing former generation particulate fitness value of working as respectively; Note all and calculate back fitness values better particulate that becomes at experience PSO;

Step 6: current cycle time adds 1, and promptly k=k+1 jumps to step 3, rerun, and (be k＞N) until cycle index greater than maximum cycle N;

Step 7: finish operation, and output Optimum Matching position and optimal-adaptive degree value.

Images