CN102184541B - Multi-objective optimized human body motion tracking method - Google Patents

Abstract

The invention discloses a multi-objective optimized human body motion tracking method, which relates to the field of computer vision and is used for tracking human body motion and estimating a posture. The method comprises the following steps of: (1) establishing a human body skeleton model; (2) preprocessing a video image; (3) initiating; (4) constructing similarity functions; (5) optimizing a target function; and (6) selecting an optimal human body motion posture. The method has the advantages that: the method is not limited by a study database and is suitable for the common videos, and video image information can be better used through two similarity functions; meanwhile, compared with the conventional single-objective optimized human body tracking method, the multi-objective optimized human body motion tracking method has the advantages that: the non-dominated neighborhood immune algorithm of a multi-objective optimization algorithm can be prevented from falling into local optimum, and the human body motion tracking accuracy is improved.

Description

The multiple-objection optimization human body motion tracking method

Technical field

The invention belongs to technical field of image processing; Further relate to and realize a kind of method that human motion is followed the tracks of in the computer vision field; Adopt a kind of method of multiple-objection optimization to realize human motion tracking and 3 d pose estimation, can be used for fields such as athletic training and cartoon making.

Background technology

The main task that human motion is followed the tracks of is from video image, to detect human body contour outline, and the articulation point to human body positions again, identifies the human motion attitude on this basis, final reconstruction of three-dimensional human motion attitude.Because video image is the projection of human body contour outline on two dimensional image in the three-dimensional scenic at present; So; Lost a large amount of depth informations, and in the human motion process, human limb takes place often from blocking phenomenon; There is ambiguousness in video image, and this makes and is difficult to from unmarked monocular video, recover the human motion attitude.But, owing in various aspects such as therapeutic treatment, athletic training, cartoon making, intelligent monitor systems potential application and economic worth are arranged all based on the human motion tracking of monocular video, so received a lot of scholars' concern.The method of following the tracks of based on the human motion of video so far, mainly is divided into two big types: follow the tracks of and follow the tracks of based on the human motion of model based on the human motion of study.

First kind, based on the human body motion tracking method of study.This method is at first extracted accurate characteristics of image at the video image and the target video image lane database of training; The characteristics of image in learning training vedio data storehouse and the mapping between the movement capturing data then, direct end user's body characteristics recovers 3 d pose on target video image at last.Urtasun et al. (R.Urtasun and T.Darrell.Local Probabilistic Regression for Activity-Independent Human Pose Inference IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 2008) use balance Gaussian process dynamic model to instruct and in the monocular video sequence, follow the tracks of 3 d human motion, this dynamic model be from the less training exercise data middle school acquistion that comprises various modes to.Sigal et al. (L.Sigal andM.Black.Measure Locally; Reason Globally:Occlusion-sensitive articulated pose estimation.IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 2006.) Bayesian frame of proposition; Comprise sequential importance sampling and annealing particle filter, used multiple motion model during tracking.In order to make 3-d recovery meet the anatomical joints restriction more and to reduce the search volume, from training data, learn motion model, use the Euclidean distance difference of virtual tag to measure as error.The shortcoming of this method is to extract accurate characteristics of image to require a great deal of time, and follows the tracks of video and receive the restriction whether learning database exists, and can't accomplish tracking when not having learning database.

Second kind, based on the human body motion tracking method of model.This method does not need learning database, directly on target video image, extracts image information, sets up similarity function, thereby similarity function is optimized the optimum state of search in the state space of higher-dimension.Thereby obtain human body attitude more accurately.The C.Sminchisescu of institut national de recherche en infomatique et automatique (INRIA) adopts this kind method to realize multiple manikin and tracking (C.Sminchisescu and A.Jepson.GenerativeModeling for Continuous Non-Linearly Embedded Visual Inference.Intemational Conference on Machine Learning (ICML), 2004).Deutscher et al. uses border and silhouette to make up the similarity function of weighting as characteristics of image; Use annealing particle filter framework and realize human motion tracking (J.Deutscher and I.Reid.Articulated body motion capture by stochastic search.International Journal of Computer Vision (IJCV); 61 (2): 185-205,2004.).Because this method only sets up a similarity function, and the monocular calibration method that is used to optimize similarity function is easy to be absorbed in local optimum when the search optimal result, and the human body attitude that causes tracing into is inaccurate, and the time complexity of algorithm is high.

Patent " based on the multi-human body tracking method of the attribute relational graph appearance model " (number of patent application 200910043537.5 of Hunan University's application; Publication number CN101561928); This patent is at first set up attribute relational graph appearance model to present frame human detection zone; Calculate the similarity of following the tracks of the attribute relational graph appearance model of human body with previous frame, confirm the coupling of interframe human body, thereby confirm the human body tracking situation and obtain movement locus according to similarity.The deficiency that this patented claim disclosed method exists is, can only carry out human body tracking to fixing scene, and the similarity of display model is not enough to follow the tracks of accurately human body attitude.

Summary of the invention

The objective of the invention is to overcome deficiency of the prior art; A kind of human body motion tracking method of the multiple-objection optimization based on model is proposed; Employing is set up the human skeleton model based on the method for model, utilizes video image to extract the position and the half-tone information of articulation point; Make up two similarity functions, under the restriction of skeleton length, adopt multi-objective optimization algorithm to obtain the tracking of human motion attitude.

For realizing above-mentioned purpose, the concrete performing step of the present invention comprises as follows:

(1) sets up the human skeleton model.

Human skeleton is divided into 14 parts according to 15 joints; Every part is by a shaft-like bone model tormulation; The straight-line segment that has with 14 in the Virtual Space between the articulation point of three-dimensional coordinate is represented this 14 shaft-like skeleton models; Under the connection of corresponding articulation point, form whole three-dimensional human skeleton model, the D coordinates value of 15 articulation points when the human motion of one group of correspondence of input, the human skeleton model will simulate the 3 D human body attitude of motion.

(2) preprocessed video image.

2a) input human body video image divides acquisition human body silhouette through background subtraction, extracts human body contour outline, human body contour outline is carried out micronization processes form the human skeleton line;

2b) in human skeleton line upper edge skeleton line search for to the end, root, knee, pin node coordinate position, use the particle filter prediction to detect remaining human body body joint point coordinate position;

2c) on human body outline profile picture, use the sobel operator to obtain the gray-scale value of image;

(3) initialization.

3a) to step 2b) the initial time video image articulation point position that obtains carries out manual demarcation, obtains the human skeleton model of initial time;

3b) t-1 follow the tracks of the human body attitude that obtains constantly will be as t initialization human skeleton model (t＞0) constantly.

(4) make up similarity function.

4a) initialized human skeleton model projection is obtained the coordinate position of each articulation point to the two dimensional image space;

4b) set up respectively projection articulation point and detected articulation point apart from similarity function and gray level similarity function;

(5) optimization aim function.

Utilize non-domination neighborhood immune algorithm, adjust the distance similarity function and gray level similarity function are optimized, and obtain k all possible human motion attitude constantly;

(6) select human body optimal movement attitude.

To might the human motion attitude and the human body optimal movement attitude that traces into constantly of t-1 subtract each other, select the human body optimal movement attitude that the minimum attitude of difference traces into as t constantly.

The present invention has the following advantages compared with prior art:

1, because the present invention has used particle filter prediction human joint points to obtain more accurate human joint points picture position, and algorithm of the present invention compared to prior art is simple, time complexity is low.

2, because the present invention has used the human body tracing method based on model, directly video image is followed the tracks of, based on the human body tracing method of study, the present invention does not receive the restriction of learning database, can be suitable for general video tracking compared to prior art.

3, because the present invention has adopted two similarity functions in tracing process; Can the better utilization video image information; The non-domination neighborhood immune algorithm of while multi-target evolution algorithm; Optimize human body tracing method than existing single goal and can avoid being absorbed in local optimum, improved the degree of accuracy that human motion is followed the tracks of.

Description of drawings

Fig. 1 is a process flow diagram of the present invention;

Fig. 2 is human joint points testing process figure of the present invention;

Fig. 3 is the three-dimensional tracking results figure of emulation experiment of the present invention;

Fig. 4 is emulation experiment three-dimensional result of the present invention projection and the Error Graph that detects articulation point.

Embodiment

Below in conjunction with accompanying drawing the present invention is done further description.

Among Fig. 1, concrete implementation procedure of the present invention is following:

Step 1 is set up the human skeleton model.

According to anatomical knowledge, though human skeleton receives age and health affected and constantly changes, the composition of skeleton is constant, and human body roughly comprises: shin bone, femur, hipbone, trunk, radius, humerus, clavicle, neck, head.The present invention is expressed as human body the skeleton pattern of following shaft-like bone to form by 15 articulation points and 14 in this case.The straight-line segment that has with 14 in the Virtual Space between the articulation point of three-dimensional coordinate is represented this 14 shaft-like skeleton models,

The coordinate representation of each articulation point is v _i=[x _i, y _i, z _i] ^T, the whole human body skeleton representation is V=[v ₁, v ₂..., v _n] ^T, the bone length of adjacent two articulation points is expressed as || v _p-v _q||=l _{P, q}Can obtain the restrictive condition of human skeleton model thus || L _iV||=l _i, i=1,2 ..., m.L wherein _iBe 3 * 3n matrix, l _iBe the length of i root bone, m is total bone number.

Under the connection of corresponding articulation point, form whole three-dimensional human skeleton model, the D coordinates value of 15 articulation points when the human motion of one group of correspondence of input, the human skeleton model will simulate the 3 D human body attitude of motion.

Step 2, the preprocessed video image.

With reference to Fig. 2, the concrete realization of this step is following:

2a) input human body video image obtains the human body silhouette through background subtraction, extracts human body contour outline, human body contour outline is carried out micronization processes form the human skeleton line.

Adopt least square intermediate value LMedS method to obtain background image, human motion image and background image are done pixel difference, obtain the background subtraction image; Adopt morphological method to remove the noise of cutting apart in the background subtraction image, obtain human body silhouette clearly; Adopt border following algorithm to obtain human body silhouette outline; Refinement human body silhouette obtains the skeleton line of human body silhouette.

2b) in human skeleton line upper edge skeleton line search for to the end, root, knee, pin node coordinate position, use the particle filter prediction to detect the coordinate position of remaining human body articulation point.

Use the concentric circles template to search for along skeleton line, with the human body silhouettes point center of circle the most for a long time that falls into annulus as head node; Choosing human body silhouette center of gravity position is root node, with the arithmetic mean of everyone the side shadow point x coordinate figure x coordinate as root node, with the arithmetic mean of the y coordinate figure y coordinate as root node; With the three-dimensional human skeleton model is benchmark projection on video image with the root node, obtains trunk central point, clavicle joint point and left and right sides buttocks articulation point.

Use particle filter to detect hand, elbow, shoulder joint node location: to generate with the rectangle frame of articulation point according to the articulation point position, articulation point priori features note is made the not displacement characteristic of pixel in the rectangle frame as the center.Adopt second-order autoregressive model to upgrade the articulation point position probing articulation point x that t-1 obtains constantly _tThe time, upgrade the articulation point position, obtain a sampling particle, according to the rectangle frame size, obtaining with the sampling particle is the template image at center, the not displacement characteristic of calculating sampling particle i

Weights W eigh _t:

${\mathrm{Weight}}_i=-\exp (1-\frac{{\mathrm{phi}}_p^i\·{\mathrm{phi}}_t^T}{\mathrm{norm}\left({\mathrm{phi}}_p^i\right)\×\mathrm{norm}\left({\mathrm{phi}}_i\right)})$

Wherein,

Be not displacement characteristic of i sampling particle, phi _tBe the articulation point priori features of the corresponding articulation point of sampling particle, norm is 2 norms;

Articulation point x _iThe position calculation formula following:

$x_t=\underset{i=1}{\overset{N_s}{\mathrm{\Σ}}}{\mathrm{phrticle}}_i\×{\mathrm{Weight}}_i$

Wherein, particle _iThe picture position of representing i sampling particle.

Use lower limb length detection knee and pin articulation point position: the length of calculating left thigh and right thigh according to the result of former frame.According to the kneed position of the prediction of result of front cross frame, if the predicted position of left knee is the center of circle with left stern point then on the left side or the right of right knee; With left thigh length is radius, from silhouette left side or the right side begin to draw circle to shank, get first with the human body silhouette picture position when crossing; As left patella point, simultaneously, be the center of circle with right stern point; With right thigh length is radius; From the silhouette right side or the left side begin to draw circle to shank, get first with the human body silhouette picture position when crossing, as right patella point.Use same detection method detection pin articulation point with knee joint point.

2c) obtain gray-scale value.

On the basis of human body silhouette, adopt morphological method to use the sobel operator, extract the level of human body and the gray-scale value of vertical direction respectively.

Step 3: initialization.

3a) human body attitude of the initial time of the video image of emulation experiment employing of the present invention is that health is stood; Level is stretched out both hands; Integrating step 2b) hand dipping is carried out in initial time video image articulation point position, can obtain the human skeleton model parameter of initial time, and note is V ₀

3b) t initialization human skeleton model constantly equals t-1 and follows the tracks of the best human attitude (t＞0) that obtains constantly.Note, note is made V ₀

Step 4: make up similarity function.

4a) the present invention has used weak perspective projection model, and preset root node depth of field parameter D makes three-dimensional (3 D) manikin can project to the plane of delineation, is complementary with the corresponding point that detect on the plane.Initialized human skeleton model projection is obtained the coordinate position of each articulation point to the two dimensional image space; Kinematic parameter to be estimated is 15 key point V=[v on the manikin ₁, v ₂..., v _n] ^TThree-dimensional coordinate determine v wherein by these 45 parameters _i=[x _i, y _i, z _i] ^TThe coordinate that projects to behind the plane of delineation does

[u _i，w _i，1] ^T＝P _iV i＝1，2，…，n

Here P _i=T _iD, T _iBe the rotation matrix of i articulation point to root node.

4b) set up respectively projection articulation point and detected articulation point apart from similarity function and gray level similarity function.15 articulation points in the detected image are designated as z _i=[x ' _i, y ' _i] ^T, the tripleplane of i articulation point and the distance of detection are designated as

Total distance of 15 articulation points can be expressed as

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|\right|v_i^{\mathrm{proj}}-z_i\left|\right|}_2=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|\right|P_{i}V-z_i\left|\right|}_2$

Because detected key point is obtained by the preprocessed video image, human body tracking just can be expressed as asking a nonlinear parameter optimization problem, confirms thus apart from similarity function:

$\arg \underset{V}{\min }{f}_1\left(V\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|\right|P_{i}V-z_i\left|\right|}_2$

To the articulation point in the video image; T should be the same with t-1 gray-scale value constantly constantly; Note is made

according to step 2c) in the gray-scale value of the articulation point position that obtains, gray level similarity function can be expressed as:

$\underset{V}{\arg \min }{f}_2\left(V\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{S}_t{V}_t^{\mathrm{proj}}-S_{t-1}{V}_{t-1}^{\mathrm{proj}}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{S}_t{P}_i{V}_t-S_{t-1}{P}_i{V}_{t-1}$

Wherein

is the t gray-scale value of articulation point position constantly.

Step 5: optimization aim function.

By step 4b) in two similarity functions and the restriction of the skeleton in the step 1 that obtain, the human body tracking problem can be described as following multi-objective optimization question:

$\left\{\begin{array}{c}\underset{V}{\arg \min }F\left(V\right)={[f_1\left(V\right),f_2\left(V\right)]}^T\\ f_1\left(V\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|\right|P_{i}V-z_i\left|\right|}_2\\ f_2\left(V\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{S}_t{P}_i{V}_t-S_{t-1}{P}_i{V}_{t-1}\\ s.t.\left|\right|L_{i}V\left|\right|=l_i,i=\mathrm{1,2},...,m\end{array}\right.$

Non-domination neighborhood immune algorithm is a kind of algorithm of effective solution multi-objective problem; It has simulated the various antibody symbiosis in the immune response, minority antibody activation phenomena; Individuality is sorted according to crowding distance; Select the relatively independent non-domination individuality of minority to clone, recombinate, make a variation, strengthened search, avoid being absorbed in local optimum optimum solution.

Among the present invention; During non-domination neighborhood immune algorithm optimization aim function, at t constantly with the initialized attitude of human body as initial population, all individualities are sorted according to crowding distance; Select the relatively independent non-domination individuality of minority to clone, recombinate, make a variation; The search globally optimal solution, under the evolutionary generation of setting, more new population finally obtains one group of optimum solution.

Step 6: select human body optimal movement attitude.

The used video rate of emulation of the present invention is per second 64 frames, front and back frame video differ very little.Separate obtaining constantly each of t-1 in the step 5, it is poor that the corresponding articulation point of the human body optimal movement attitude that traces into its all articulation points and t-1 is done, and separating that the error that obtains is minimum will be t moment optimal movement attitude.

Effect of the present invention can obtain checking through following emulation experiment:

The used data of emulation experiment of the present invention are that the human motion video is to shoot the video certainly, and the video image size is 320 * 240.

Emulation content: human body articulation point position from input video sequence; Adopt the method for multiple-objection optimization that the human motion in the sequence of video images is followed the tracks of; Three-dimensional tracking results is as shown in Figure 3; Wherein first classify input as video image; Secondary series is the subpoint (circle) and detected human skeleton line (line) of human motion attitude, and the 3rd row are the 3 D human body optimal movement attitudes that trace into.The tracking error of experiment is represented as shown in Figure 4 by three-dimensional result projection and the error that detects articulation point.

Analysis of simulation result: from Fig. 3, can know; The human motion attitude estimated result of multi-model tracking of the present invention is basic identical with real human motion attitude; Effectively solve the ambiguity problem that human motion is followed the tracks of, improved the accuracy and the stability of following the tracks of.Can know that from Fig. 4 Multipurpose Optimal Method average error of the present invention is in 1 pixel, the visible projection error of multi-model process tracking results of the present invention that uses is less.

Emulation experiment of the present invention compiles completion on Matlab, execution environment is the HP workstation under the Windows framework.

The present invention uses the method for multiple-objection optimization to carry out human motion and follows the tracks of, and it is more accurate that the articulation point detection algorithm detects effect, and the working time that consumes is less; Employing as objective function, has made full use of the information of video image apart from likelihood function and gray scale likelihood function; Non-domination neighborhood immune algorithm is used to optimize two objective functions, avoids being absorbed in local optimum, has improved the degree of accuracy of following the tracks of, and has saved the time.Simulation result shows that this tracking has obtained three-dimensional pose recovery accurately, has reduced the human motion ambiguousness, has reduced time complexity.

Claims (5)

1. a multiple-objection optimization human body motion tracking method comprises the steps:

(1) sets up the human skeleton model;

(2) pre-service human body video image

2a) input human body video image divides acquisition human body silhouette through background subtraction, extracts human body contour outline, human body contour outline is carried out micronization processes form the human skeleton line;

2b) in human skeleton line upper edge skeleton line search for to the end, root, knee, pin node coordinate position, use the particle filter prediction to detect remaining human body body joint point coordinate position;

2c) on human body outline profile picture, use the sobel operator to obtain the gray-scale value of image;

(3) initialization

3a) to step 2b) the initial time human body video image articulation point position that obtains carries out manual demarcation, obtains the human skeleton model of initial time;

3b) t-1 follow the tracks of the best human attitude that obtains constantly will be as t initialization human skeleton model t＞0 constantly;

(4) make up similarity function

4a) initialized human skeleton model projection is obtained the coordinate position of each articulation point to the two dimensional image space;

4b) set up respectively the projection articulation point with detect articulation point apart from similarity function and gray level similarity function:

Set up apart from similarity function according to following formula:

Wherein,

Expression f ₁(V) in V is minimized, V is three-dimensional body joint point coordinate matrix, and n is the human joint points number, n=15, ‖ ‖ ₂Represent 2 norms, P _iBe projective parameter, z _iBe the body joint point coordinate in the detected image;

Set up gray level similarity function according to following formula:

Wherein,

Expression f ₂(V) in V is minimized, V is three-dimensional body joint point coordinate matrix, and n is the human joint points number, n=15,

Be the t gray-scale value of articulation point position constantly, S _tBe t sobel operator constantly,

Be t three-dimensional body joint point coordinate matrix V of the moment _tProject to the coordinates matrix behind the plane of delineation,

Be the t-1 gray-scale value of articulation point position constantly, S _T-1Be t-1 sobel operator constantly,

Be t-1 three-dimensional body joint point coordinate matrix V of the moment _T-1Project to the coordinates matrix behind the plane of delineation, P _iBe projective parameter, V _tBe t human body three-dimensional body joint point coordinate matrix constantly, V _T-1Be t-1 human body three-dimensional body joint point coordinate matrix constantly;

(5) optimization aim function:

Wherein,

Among the expression F (V) V is minimized, V is three-dimensional body joint point coordinate matrix, and n is the human joint points number, n=15, ‖ ‖ ₂Represent 2 norms, ‖ ‖ ^TThe expression transposition,

Be the t gray-scale value of articulation point position constantly, S _tBe t sobel operator constantly,

Be t three-dimensional body joint point coordinate matrix V of the moment _tProject to the coordinates matrix behind the plane of delineation,

Be the t-1 gray-scale value of articulation point position constantly, S _tBe t-1 sobel operator constantly,

Be t-1 three-dimensional body joint point coordinate matrix V of the moment _T-1Project to the coordinates matrix behind the plane of delineation, P _iBe projective parameter, V _tBe t human body three-dimensional body joint point coordinate matrix constantly, V _T-1Be t-1 human body three-dimensional body joint point coordinate matrix constantly, l _iBe the length of i root bone, m is total bone number;

Utilize non-domination neighborhood immune algorithm, adjust the distance similarity function and gray level similarity function are optimized, and obtain tAll possible human motion attitude of the moment;

(6) select human body optimal movement attitude

To might the human motion attitude and the human body optimal movement attitude that traces into constantly of t-1 do poorly, select the human body optimal movement attitude that the minimum attitude of difference traces into as t constantly.

2. multiple-objection optimization human body motion tracking method according to claim 1; It is characterized in that: said step (1) human skeleton model makes up by following process; Human skeleton is divided into 14 parts according to 15 joints; Every part is by a shaft-like bone model tormulation, and the straight-line segment that has with 14 in the Virtual Space between the articulation point of three-dimensional coordinate is represented this 14 shaft-like skeleton models, at the connection of the corresponding articulation point whole three-dimensional human skeleton model of composition down; When D coordinates value corresponding to 15 articulation points of human motion of one group of input, the human skeleton model will simulate the 3 D human body attitude of motion.

3. multiple-objection optimization human body motion tracking method according to claim 1; It is characterized in that: said step 2a) step of background difference is following; Input human body video image; Adopt least square intermediate value LMedS method to obtain background image, human body video image and background image are done pixel difference, obtain the background subtraction image; Adopt morphological method to remove the noise of cutting apart in the background subtraction image, obtain human body silhouette clearly.

4. multiple-objection optimization human body motion tracking method according to claim 1; It is characterized in that: the step of particle filter prediction said step 2b) is following; When detecting hand, elbow, shoulder joint node location; Generating with the articulation point according to the articulation point position is the rectangle frame at center, articulation point priori features note is made the not displacement characteristic of pixel in the rectangle frame; Adopt second-order autoregressive model to upgrade the articulation point position that t-1 obtains constantly, then when detecting articulation point x _tThe time, promptly upgrade articulation point x _tThe position, obtain a sampling particle, according to rectangle frame size, obtaining with the sampling particle is the template image at center, the not displacement characteristic of calculating sampling particle i

Weight

Wherein,

Be not displacement characteristic of i sampling particle, phi _tBe the articulation point priori features of the corresponding articulation point of sampling particle, norm is 2 norms, () ^TThe expression transposition;

Articulation point x _tThe position calculation formula following:

Wherein, particle _iThe picture position of representing i sampling particle, N _SBe the sampling population.

5. multiple-objection optimization human body motion tracking method according to claim 1; It is characterized in that: the optimized step of said step (5) non-domination neighborhood immune algorithm is; At t constantly, during non-domination neighborhood immune algorithm optimization aim function, with the initialized attitude V of human body ₀As initial population, all individualities are sorted according to crowding distance, select the relatively independent non-domination individuality of minority to clone, recombinate, make a variation, the optimum solution of the search overall situation, under the evolutionary generation of setting, more new population finally obtains one group of optimum solution.

Images