CN110136165A - A kind of mutation movement method for tracking target based on the optimization of adaptive whale - Google Patents
A kind of mutation movement method for tracking target based on the optimization of adaptive whale Download PDFInfo
- Publication number
- CN110136165A CN110136165A CN201910411490.7A CN201910411490A CN110136165A CN 110136165 A CN110136165 A CN 110136165A CN 201910411490 A CN201910411490 A CN 201910411490A CN 110136165 A CN110136165 A CN 110136165A
- Authority
- CN
- China
- Prior art keywords
- whale
- image block
- target
- adaptive
- candidate image
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/223—Analysis of motion using block-matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention proposes a kind of mutation movement method for tracking target based on the optimization of adaptive whale, convergence rate slower problems not high to solve existing method operational efficiency.Step of the present invention are as follows: the model parameter of init state parameter and adaptive whale optimization algorithm;Coordinate position is randomly generated according to adaptive whale optimization algorithm, intercepts region identical with target image block size as candidate image block, according to the position for shrinking encirclement mechanism or spiral update mechanism update candidate image block;, as best candidate image block, the exploration step-length for shrinking encirclement mechanism is adjusted using 1/5th principle Nonlinear Dynamics for the maximum candidate image block of the similarity value of target image block;The tracking of next frame image is carried out using best candidate image block as the target image block of present frame and the dbjective state parameter of next frame.The present invention improves the stability of global optimum and jumps out the ability of local optimum, improves operational efficiency, can be good at adapting to motion target tracking problem.
Description
Technical field
The present invention relates to target according to technical field more particularly to it is a kind of based on adaptive whale optimization mutation movement
Method for tracking target can realize the duration tracking of target when target mutation movement occurs between consecutive frame well.
Background technique
In computer vision field, the tracking of moving target is a hot research problem in video, but due to
The complexity of track environment, the factors such as the uncertainty of target movement and video camera imaging, often leads to mesh in adjacent two field pictures
Target displacement is very big, and target following is caused to fail.It is traditional to be assumed based on smooth target for this phenomenon of targeted mutagenesis
Numerous algorithms be easily tracked failure.And video tracking can often be converted into and find most in an image sequence or video
The problem of figure of merit, therefore this phenomenon can be solved with the algorithm for finding optimal solution for targeted mutagenesis.Therefore, with optimization
Method solves the situation of moving target mutation, it is ensured that the robustness of target tracking algorism.
In video frequency object tracking, targeted mutagenesis is solved the problems, such as with global optimization method, first has to solve traditional excellent
Change two obvious problems of algorithm: (1) in optimization process, if the Energy distribution of image is coarse, can be easy to make
Algorithm falls into local optimum.If all carrying out global search to each frame picture, efficiency be will be greatly reduced;(2) tradition optimization is calculated
The adjustment parameter of method is more, and convergence rate is slower.Therefore, it is necessary to which finding one kind not only can solve above two problem, but also can guarantee
The motion target tracking method of its versatility.
Summary of the invention
It is not high for existing motion target tracking method operational efficiency, contain more adjustment parameter, convergence rate compared with
Slow technical problem, the present invention propose a kind of mutation movement method for tracking target based on the optimization of adaptive whale, AWOA are searched
Rope is introduced into tracking, so that the ability of searching optimum of AWOA algorithm is strengthened using 1/5th principles, to solve mutation fortune
Dynamic Target Tracking Problem.
In order to achieve the above object, the technical scheme of the present invention is realized as follows: a kind of optimized based on adaptive whale
Mutation movement method for tracking target, its step are as follows:
Step 1: the model parameter of the state parameter of initialized target image block and adaptive whale optimization algorithm: initial
Change the whale number of whale group and the position of whale, the logarithmic spiral shape that maximum cycle and spiral update mechanism is arranged is normal
Number, current iteration number t=1;
Step 2: candidate image block is searched for using adaptive whale optimization algorithm: 1) in the corresponding coordinate of current frame image
Region identical with target image block size is intercepted on position as candidate image block, according to the first of adaptive whale optimization algorithm
Coordinate position is randomly generated as initial position in beginning model parameter, using initial position as being used as target prey, 2) and utilize five points
One of the adjustment of principle Nonlinear Dynamic shrink the exploration step-length of encirclement mechanism, according to shrinking encirclement mechanism or spiral update mechanism more
The position of new candidate image block;3) using with the maximum candidate image block of the similarity value of target image block as current iteration most
Excellent candidate image block, and the coordinate of best candidate image block is stored in (Xbest,Ybest);4) current iteration number t=t+1, will
The best candidate image block of last iteration is as target prey, return step 2);5) when the number of iterations t reaches largest loop iteration
Number, output are stored in (Xbest,Ybest) best candidate image block;
Step 3: using the best candidate image block of step 2 output as the target image block of present frame and the mesh of next frame
State parameter is marked, return step two carries out the tracking of next frame image;
Step 4: repeating step 2 --- step 3, until reaching last frame image, exports the optimal of each frame image
Candidate image block realizes the tracking of moving target.
The method of the state parameter of initialized target image block in the step 1 are as follows: read the data letter of first frame image
Breath, determines state parameter [x, y, w, h] of the target image block in first frame image, wherein (x, y) is target image block every
The coordinate value of frame image top left corner pixel point, w are the width of target image block, and h is the height of target image block.
The side of coordinate position is randomly generated in the step 2 according to the original model parameter of adaptive whale optimization algorithm
Method is: the position of parameter needed for initialization whale optimization algorithm and N number of whale, and the position conduct of N number of whale is randomly generated
The position in the upper left corner of candidate image block obtains the initial whale group of algorithm;The Optimized model ginseng of adaptive whale optimization algorithm
Number includes number the size N, maximum cycle T, logarithmic spiral shape constant b of whale group.
The probability for shrinking the position that encirclement mechanism or spiral update mechanism update candidate image block is 0.5.
It is described to shrink the method for surrounding the position of new mechanism candidate image block are as follows: to set current optimal location as target prey
After determining optimal whale Search of Individual, other whale individuals will by updating its own position to approach current optimal location,
Establish model are as follows:
Wherein, t is current iteration number,It is whale individual at a distance from target prey,When for the t times iteration
Current optimal location, X (t) are the current location of whale individual,For the updated position of whale individual, | | it is absolute value
Operator accords with for point multiplication operation;Number vectorWithCalculation formula are as follows:
Wherein,For the random vector in [0,1], step-length vector is exploredWith the number of iterations by 2 → 0 gradually linear decreases,
And explore step-length vectorIt indicates are as follows:
Wherein,WithThe maximum value and minimum value for exploring step-length vector are respectively represented, T is maximum number of iterations, fp
(t) dynamic Tuning function is represented.
The side of the exploration step-length of encirclement mechanism is shunk in the step 2 using the adjustment of 1/5th principle Nonlinear Dynamics
Method are as follows: when the evolution rate of whale group is less than 15%, the value for exploring step-length reduces, when the evolution rate of whale group is greater than 25%,
The value for exploring step-length increases, and when the evolution rate of whale group is between 15% and 25%, the value for exploring step-length is constant.
The exploration step-length is by adjusting the realization of dynamic Tuning function, the dynamic Tuning function are as follows:
Wherein, fp(t) and fp(t-1) dynamic Tuning function value of the current iteration t than last iteration (t-1) is respectively indicated,For the evolution rate of population, n is current iteration whale quantity more increased than similarity value in last iteration;N is whale
Sum, fpIt (1)=1 is the initial value of first time iteration, f0For the constant greater than 1.
WhenWhen, it randomly chooses some whale position and makes it away from target prey, find a more preferably prey
Mathematical model is expressed as follows:
Wherein;Indicate whale individual to selected whaleDistanceFor the current location of whale individual,Indicate the position vector of the whale randomly selected from current search agency.
The method of the position of the new candidate image block of spiral update mechanism are as follows: it is spiral that whale individual finds that prey passes through
Movement indicates to capture target prey are as follows:
Wherein,Current optimal location when for the t times iteration,For the updated position of whale individual,Indicate the distance between whale individual and the target prey of current iteration t, b is for limiting logarithm spiral shell
The constant of shape is revolved, l is the random number of [- 1,1].
The calculation method of similarity value in the step 2 are as follows: the HOG feature of target image block and candidate image block is extracted,
Using them as stochastic variable, the similarity value between target image block and candidate image block is obtained:
Wherein, D () indicates variance, and Cov () indicates covariance, and X represents the HOG feature of target image block, and Y represents candidate image block
HOG feature, the value range of ρ (X, Y) are [- 1,1];Calculate similarity value ρ (X, Y), and by maximum similarity value ρ (X,
Y) corresponding candidate image block is stored in (X as current best candidate image block, and by the coordinate of candidate image blockbest,
Ybest)。
Beneficial effects of the present invention: the motion state search mechanisms of global optimum are taken, in the contraction of whale optimization algorithm
On the basis of the search mechanisms of encirclement mechanism and spiral update mechanism, the searching route of whale is influenced using 1/5th principles,
In search process, the Nonlinear Dynamic variation for exploring step-length improves the stability of global optimum and jumps out the energy of local optimum
Power, with the increase of the number of iterations, the optimizing ability of whale and the ability enhancing for jumping out local optimum move closer to whale and hunt
The operational efficiency of algorithm is improved in object, the search space gradually reduced;Due to whale optimizing ability and jump out the energy of local optimum
Power is strong, can be good at adapting to motion target tracking problem, therefore can solve the problems, such as target movement mutation very well, subsequent right
Identification, understanding and the analysis of tracking target are of great significance.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is flow chart of the invention.
Fig. 2 is the range accuracy comparison schematic diagram of the present invention with other algorithms, wherein (a) is HUMAN7 video, (b) is
FACE1 video (c) is DEER video, (d) is BOY video.
Fig. 3 is the Duplication comparison schematic diagram of the present invention with other algorithms, wherein (a) is DEER video, (b) is FACE1
Video (c) is BOY video, (d) is HUMAN7 video.
Fig. 4 is the tracking effect schematic diagram of the present invention with other algorithms, wherein (a) is BOY video, (b) is regarded for DEER
Frequently, (c) it is FACE1 video, (d) is HUMAN7 video.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other under that premise of not paying creative labor
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, a kind of mutation movement method for tracking target based on the optimization of adaptive whale, main thought are:
(1) adaptive whale optimization algorithm is introduced into tracking field, Target Tracking Problem is converted into optimization problem.(2), in whale
On the basis of the search mechanisms that the contraction of fish optimization algorithm is surrounded and spiral updates, searching for whale is influenced using 1/5th principles
Rope path enhances the optimizing ability of whale with the ability for jumping out local optimum.(3) as whale moves closer to prey, by
The search space reduced is walked, the operational efficiency of algorithm is improved.Specific step is as follows:
Step 1: the Optimized model parameter of the state parameter of initialized target image block and adaptive whale optimization algorithm.
The data information for reading first frame image, determines state parameter [x, y, w, h] of the target in first frame image,
In, (x, y) is the coordinate value of image block top left corner pixel point, and w is target width, and h is object height;Whale optimization is initialized to calculate
The position of parameter needed for method and N number of whale obtains the initial whale group of algorithm;The Optimized model parameter of AWOA algorithm includes
Number the size N, maximum number of iterations T, logarithmic spiral shape constant b, Evolution of Population rate P=0.2 of whale group.
Step 2: candidate image block is searched for using adaptive whale optimization algorithm, realizes target following.
Using the implementation method of adaptive whale optimization algorithm search candidate image block are as follows: 1) in the corresponding of current frame image
Region identical with target image block size is intercepted on coordinate position as candidate image block, according to adaptive whale optimization algorithm
Original model parameter coordinate position is randomly generated as initial position, using initial position as current optimal location is used as, work as t
When=1, using current optimal location as target prey (that is: best search agent), 2) utilize 1/5th principle Nonlinear Dynamics
The exploration step-length of encirclement mechanism is shunk in state adjustment, according to the position for shrinking encirclement mechanism or spiral update mechanism update candidate image block
It sets;3) using with the maximum candidate image block of the similarity value of target image block as the best candidate image block of current iteration, and
The coordinate of best candidate image block is stored in (Xbest,Ybest);4) current iteration number t=t+1 will change last time as t > 1
The best candidate image block in generation is as target prey, return step 2);5) when the number of iterations t reaches largest loop the number of iterations,
Output is stored in (Xbest,Ybest) best candidate image block.
It is exactly the fitness function utilized between target image block and candidate image block using solution tracking problem is optimized,
The maximum similarity value of candidate image block and target image block is found in each frame image.When a new frame image arrives,
According to the information of AWOA algorithm and the best candidate image block found in previous frame image, searched on a new frame image
Dbjective state parameter of the best candidate image block as next frame image.Mainly include following content:
A. adaptive value is calculated.According to the upper left corner for the candidate image block being randomly generated in the initial model parameter of AWOA algorithm
Position, upper interception image block identical with target image block size is as candidate image block at various locations.Extract image block
HOG feature obtain the similarity between target image block and candidate image block using them as stochastic variable:
Wherein, D () indicates the variance of image block, and Cov () indicates the covariance of two image blocks, and X represents target figure
As the HOG feature of block, Y represents the HOG feature of candidate image block, and the value range of similarity ρ (X, Y) is [- 1,1];By phase
The maximum value of closing property is that the candidate image block position of the maximum absolute value of ρ (X, Y) is stored in as current best candidate image block
(Xbest,Ybest)。
B. the position of candidate image block is updated by more newly arriving for whale position, whale optimization algorithm is by simulating seat head
Whale is scanned in predation using contraction encirclement mechanism and spiral update mechanism, the position of Lai Gengxin candidate image block.?
Whale is as search agent in AWOA algorithm, and the maximum whale of similarity magnitude is as prey, with the increase of the number of iterations, whale
Fish is gradually close towards prey, gradually finds best candidate solution.It specifically includes:
1) update of search agent position.The update of whale position by its whale self-position and is worked as in search space
The influence of preceding optimal whale position, the update for each search agent position: a. search agent is by shrinking encirclement mechanism
Target prey is surrounded, by spiral update mechanism or shrinking encirclement mechanism makes search agent move closer to target prey, Lai Jinhang
The update of search agent position;B. after search agent gradually surrounds target prey, shrink pack is randomly choosed with 50% probability
It encloses mechanism or spiral update mechanism updates the position of whale individual;C. in the heuristic process of prey, a whale is randomly choosed
The update that position is carried out as prey, enhances the global optimizing ability of algorithm;D. under the collective effect of step a ﹑ b and c, own
Search agent form new location sets.E. the set according to the new position of search agent calculates what all new positions were formed
The similarity value of image block;The location sets for comparing all search agents are formed by the phase of candidate image block with target image block
Like angle value, the maximum candidate image block of similarity value as current iteration best candidate image block and as next iteration
Target prey.
1. shrinking encirclement mechanism
Humpback may search for and surround prey to carry out location updating.It is assumed that current optimal location is target prey
(or having gained on the chase) after determining optimal location (i.e. optimal whale Search of Individual), other whale individuals will be by updating it certainly
Body position can establish the model to approach current optimal location are as follows:
In formula, t is current iteration number,It is whale individual at a distance from target prey,When for the t times iteration
Current optimal location,For whale individual current location,For the updated position of whale individual, | | it is absolute value
Operator accords with for point multiplication operation.WithIt is number vector, their calculation formula is as follows:
In formula,With the number of iterations by 2 → 0 gradually linear decreases,For the random vector in [0,1].It may be expressed as:
In formula,WithRespectively represent maximum value and minimum value.T is current iteration number, and T is maximum number of iterations, fp
(t) dynamic Tuning function is represented.
Since in Swarm Intelligent Algorithm, according to 1/5th principles, the Optimal Evolution rate of population is about 20%.This
Invention uses 1/5th on the basis of standard whale optimization algorithm (Whale Optimization Algorithm, WOA)
This Nonlinear Dynamic feedback strategy of principle replaces the linear mechanism of characterization exploration step change in standard whale optimization algorithm,
That is: when the evolution rate of population is less than 15%, reduce the value for exploring step-length, when the evolution rate of population is greater than 25%, make to visit
The value of Suo Buchang increases, and when the evolution rate of population is between 15% and 25%, then the value for exploring step-length is constant.
(1) when the evolution rate p of population is less than 15%, desired value is not achieved in the evolution of whale, and search space is big, causes to receive
Accuracy decline is held back, scarce capacity is explored.Therefore, in order to improve exploitation performance, dynamic Tuning function f is usedp(t) it is explored to reduce
Step-length vectorValue.
(2) when the evolution rate p of population is greater than 25%, the evolution of whale has been more than desired value.Therefore, it is explored to improve
Performance uses dynamic Tuning function fp(t) step-length vector is explored to increaseValue.
(3) when the evolution rate p of population is between 15%-25%, then the evolution of whale has approximately reached most preferably.This meaning
Taste explore and exploitation performance obtained effective balance.It therefore, there is no need to change and explore step-length vectorValue.
It can be described as:
In formula,For the evolution rate of population, n is to compare in current iteration with last time iteration, and similarity value increases
Whale quantity;N is whale sum, fpIt (1)=1 is the initial value of first time iteration, initial dynamic adjustment degree f0For greater than 1
Constant.
2. spiral update mechanism
Humpback discovery prey is bagged the game by screw movement, and model may be expressed as:
In formula,Indicate the distance between whale and prey, b is for limiting logarithmic spiral shape
Constant, l be [- 1,1] random number.
3. encirclement mechanism and spiral update mechanism, two kinds of positions are shunk in selection during candidate image block location updating
The probability of update is identical, is 0.5.
4. in order to enhance the global exploring ability of algorithm, whenWhen, it randomly chooses some whale position and makes it away from
Target prey, to find a more preferably prey, mathematical model is expressed as follows:
In formula,Indicate whale individual to selected whaleDistance,For the current location of whale individual,Indicate the position vector of the whale randomly selected from current search agency.
5. the new location sets formed by all whale individuals calculate image block and target figure that these new positions are formed
As the similarity value of block;Wherein the maximum image block of similarity value as current iteration optimum image block and with (X beforebest,
Ybest) formed image block compare, retain preferably one in (Xbest,Ybest)。
2), when current iteration is completed, the optimal image block of current iteration is saved, and judge whether to reach greatest iteration
Number exports tracking target of the present frame optimal solution as present frame, not up to greatest iteration when the maximum number of iterations is reached
When number, into next iteration, and continue step 1).
Step 3: using the best candidate image block of output as the target image block of current frame image and the target of next frame
State parameter carries out the tracking of next frame image.
Step 4: step 2 --- step 3 is repeated, realizes the tracking of moving target.
Implementation steps of the invention are as follows: the Optimized model parameter of initialized target state parameter and AWOA algorithm, use
AWOA algorithm, which updates candidate image block position, makes the HOG feature and target image block of candidate image block using similarity metric function
HOG feature compare, the best candidate image block of present frame is found, using this image block as the tracking mesh of next frame image
Mark;It repeats above operation, realizes the tracking to moving target.When target mutates movement between consecutive frame, the present invention
It can prevent the loss of target from realizing effective duration tracking, improve the adaptability of the tracking under complex environment.This
Hardware environment of the invention for implementation are as follows: Intel (R) Core (TM) i5CPU 2.3G computer, 8GB memory, the software of operation
Environment is: Matlab R2017a and Windows10.Video sequence can be on the http://visualtracking.net of website
It obtains.
The present invention evaluates its validity using qualitative and quantitative two ways.Qualitative evaluation mode uses the method for the present invention
Optimize (WOA) track side Fa ﹑ based on the related of context-aware frame to based on population (PSO) tracking, based on whale
Filtering (CACF) track side's method ﹑ is based on simulated annealing (SA) track side's method ﹑ to be compared based on core correlation filtering (KCF) tracking
Compared with their tracking effect figures in partial frame are as shown in figure 4, it is followed successively by FACE1 ﹑ BOY ﹑ DEER ﹑ HUMAN7 from top to bottom.
Wherein, in first and second video, the interframe movement displacement of target in the horizontal direction respectively reaches 117 and 89 pixels
Big displacement movement, the interframe movement displacement of target in vertical direction reaches 85 and 21 pictures in third and the 4th video
Element, these mutation movements and the interframe big displacement that generates make it is classical track algorithm --- CACF tracker shows to be not suitable with,
And be equally although that the PSO tracker of optimization algorithm and WOA tracker sometimes can adapt to the movements of these big displacements,
But in tracking accuracy, it will be apparent that be worse than method proposed by the present invention, the tracking effect that method proposed by the present invention has obtained.
Quantitative assessment mode is evaluated using center error rate and target Duplication, and range accuracy DP refers to the threshold value according to setting
The frame number of target and the ratio of totalframes can be successfully tracked in image sequence.Wherein, threshold value is by tracking result and true
What errors of centration value as a result determined, it is 0.5 that threshold value, which is arranged, in the present invention.Errors of centration is the centre bit by calculating tracking target
The Euclidean distance between actual position is set, generally its value is smaller illustrates that tracking result is more excellent.Target Duplication OP refer to
Track result and real goal region area and operation and and operation ratio, value is bigger, illustrates that tracking result is better.Fig. 2 exhibition
Show that the range accuracy DP value comparison result of the present invention with CACF, PSO, WOA, Fig. 3 show the target overlapping of corresponding track algorithm
The comparison result of rate OP value.4 result of complex chart indicates that it is prominent that method for tracking target provided by the invention can well solve target
The motion problems of change obtain preferably tracking performance.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of mutation movement method for tracking target based on the optimization of adaptive whale, which is characterized in that its step are as follows:
Step 1: the model parameter of the state parameter of initialized target image block and adaptive whale optimization algorithm: initialization whale
The logarithmic spiral shape constant of maximum cycle and spiral update mechanism is arranged in the whale number of the shoal of fish and the position of whale,
Current iteration number t=1;
Step 2: candidate image block is searched for using adaptive whale optimization algorithm: 1) in the corresponding coordinate position of current frame image
It is upper to intercept region identical with target image block size as candidate image block, according to the introductory die of adaptive whale optimization algorithm
Coordinate position is randomly generated as initial position in shape parameter, using initial position as being used as target prey, 2) and utilize 1/5th
The exploration step-length of encirclement mechanism is shunk in the adjustment of principle Nonlinear Dynamic, is updated and is waited according to contraction encirclement mechanism or spiral update mechanism
Select the position of image block;3) using with the maximum candidate image block of the similarity value of target image block as the optimal time of current iteration
Image block is selected, and the coordinate of best candidate image block is stored in (Xbest,Ybest);4) current iteration number t=t+1, by last time
The best candidate image block of iteration is as target prey, return step 2);5) when the number of iterations t reaches largest loop iteration time
Number, output are stored in (Xbest,Ybest) best candidate image block;
Step 3: using the best candidate image block of step 2 output as the target image block of present frame and the target-like of next frame
State parameter, return step two carry out the tracking of next frame image;
Step 4: repeating step 2 --- step 3, until reaching last frame image, exports the best candidate of each frame image
Image block realizes the tracking of moving target.
2. the mutation movement method for tracking target according to claim 1 based on the optimization of adaptive whale, which is characterized in that
The method of the state parameter of initialized target image block in the step 1 are as follows: read the data information of first frame image, determine
State parameter [x, y, w, h] of the target image block in first frame image, wherein (x, y) is target image block in every frame image
The coordinate value of top left corner pixel point, w are the width of target image block, and h is the height of target image block.
3. the mutation movement method for tracking target according to claim 1 based on the optimization of adaptive whale, which is characterized in that
It is according to the method that coordinate position is randomly generated in the original model parameter of adaptive whale optimization algorithm in the step 2: initial
The position of parameter needed for changing whale optimization algorithm and N number of whale, is randomly generated the position of N number of whale as candidate image block
The upper left corner position, obtain the initial whale group of algorithm;The Optimized model parameter of adaptive whale optimization algorithm includes whale
Number the size N, maximum cycle T, logarithmic spiral shape constant b of group.
4. the mutation movement method for tracking target according to claim 1 based on the optimization of adaptive whale, which is characterized in that
The probability for shrinking the position that encirclement mechanism or spiral update mechanism update candidate image block is 0.5.
5. the mutation movement method for tracking target according to claim 1 or 4 based on the optimization of adaptive whale, feature exist
In described to shrink the method for surrounding the position of new mechanism candidate image block are as follows: setting current optimal location is target prey true
After fixed optimal whale Search of Individual, other whale individuals will approach current optimal location by updating its own position, establish
Model are as follows:
Wherein, t is current iteration number,It is whale individual at a distance from target prey,It is current when for the t times iteration
Optimal location, X (t) are the current location of whale individual,For the updated position of whale individual, | | it is signed magnitude arithmetic(al)
Symbol accords with for point multiplication operation;Number vectorWithCalculation formula are as follows:
Wherein,For the random vector in [0,1], step-length vector is exploredWith the number of iterations by 2 → 0 gradually linear decreases, and spy
Rope walks long vectorIt indicates are as follows:
Wherein,WithThe maximum value and minimum value for exploring step-length vector are respectively represented, T is maximum number of iterations, fp(t) generation
Table dynamic Tuning function.
6. the mutation movement method for tracking target according to claim 1 or 5 based on the optimization of adaptive whale, feature exist
In the method for utilizing 1/5th principle Nonlinear Dynamics to adjust the exploration step-length of contraction encirclement mechanism in the step 2 are as follows:
When the evolution rate of whale group is less than 15%, the value for exploring step-length reduces, and when the evolution rate of whale group is greater than 25%, explores step
Long value increases, and when the evolution rate of whale group is between 15% and 25%, the value for exploring step-length is constant.
7. the mutation movement method for tracking target according to claim 6 based on the optimization of adaptive whale, which is characterized in that
The exploration step-length is by adjusting the realization of dynamic Tuning function, the dynamic Tuning function are as follows:
Wherein, fp(t) and fp(t-1) dynamic Tuning function value of the current iteration t than last iteration (t-1) is respectively indicated,For the evolution rate of population, n is current iteration whale quantity more increased than similarity value in last iteration;N is whale
Sum, fpIt (1)=1 is the initial value of first time iteration, f0For the constant greater than 1.
8. the mutation movement method for tracking target according to claim 5 based on the optimization of adaptive whale, which is characterized in that
WhenWhen, it randomly chooses some whale position and makes it away from target prey, find the mathematical model of a more preferably prey
It is expressed as follows:
Wherein;Indicate whale individual to selected whaleDistanceFor the current location of whale individual,Table
Show the position vector of the whale randomly selected from current search agency.
9. the mutation movement method for tracking target according to claim 1 or 4 based on the optimization of adaptive whale, feature exist
In the method for the position of the new candidate image block of spiral update mechanism are as follows: whale individual discovery prey passes through screw movement
It is indicated to capture target prey are as follows:
Wherein,Current optimal location when for the t times iteration,For the updated position of whale individual,Indicate the distance between whale individual and the target prey of current iteration t, b is for limiting logarithm spiral shell
The constant of shape is revolved, l is the random number of [- 1,1].
10. the mutation movement method for tracking target according to claim 1 based on the optimization of adaptive whale, feature exist
In the calculation method of similarity value in the step 2 are as follows: the HOG feature for extracting target image block and candidate image block, by it
Be used as stochastic variable, obtain the similarity value between target image block and candidate image block:Its
In, D () indicates variance, and Cov () indicates covariance, and X represents the HOG feature of target image block, and Y represents candidate image block
HOG feature, the value range of ρ (X, Y) are [- 1,1];Calculate similarity value ρ (X, Y), and by maximum similarity value ρ (X,
Y) corresponding candidate image block is stored in (X as current best candidate image block, and by the coordinate of candidate image blockbest,
Ybest)。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910411490.7A CN110136165A (en) | 2019-05-17 | 2019-05-17 | A kind of mutation movement method for tracking target based on the optimization of adaptive whale |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910411490.7A CN110136165A (en) | 2019-05-17 | 2019-05-17 | A kind of mutation movement method for tracking target based on the optimization of adaptive whale |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110136165A true CN110136165A (en) | 2019-08-16 |
Family
ID=67574879
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910411490.7A Pending CN110136165A (en) | 2019-05-17 | 2019-05-17 | A kind of mutation movement method for tracking target based on the optimization of adaptive whale |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110136165A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111598928A (en) * | 2020-05-22 | 2020-08-28 | 郑州轻工业大学 | Abrupt change moving target tracking method based on semantic evaluation and region suggestion |
CN111696645A (en) * | 2020-06-07 | 2020-09-22 | 长春理工大学 | Hand exoskeleton rehabilitation training device and method based on surface electromyographic signals |
CN112163808A (en) * | 2020-09-23 | 2021-01-01 | 贵州工程应用技术学院 | Method for solving logistics center addressing problem by self-adaptive whale algorithm based on opponent learning |
CN113043278A (en) * | 2021-04-13 | 2021-06-29 | 南京工业职业技术大学 | Mechanical arm track planning method based on improved whale searching method |
CN117609768A (en) * | 2024-01-23 | 2024-02-27 | 昆明理工大学 | Vertical water pump unit fault diagnosis method based on improved whale algorithm |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1684543A (en) * | 2004-04-16 | 2005-10-19 | 华为技术有限公司 | Control method for activating time deviation |
CN108010069A (en) * | 2017-12-01 | 2018-05-08 | 湖北工业大学 | Optimize the rapid image matching method of algorithm and grey correlation analysis based on whale |
CN108305272A (en) * | 2018-02-27 | 2018-07-20 | 郑州轻工业学院 | A kind of mutation movement method for tracking target based on ALO search |
CN108389211A (en) * | 2018-03-16 | 2018-08-10 | 西安电子科技大学 | Based on the image partition method for improving whale Optimization of Fuzzy cluster |
CN108460548A (en) * | 2018-05-17 | 2018-08-28 | 西安建筑科技大学 | A kind of more Metal Open production-operational-plan methods based on improvement grey wolf algorithm |
CN108896306A (en) * | 2018-03-26 | 2018-11-27 | 四川大学 | Method for Bearing Fault Diagnosis based on adaptive atom dictionary OMP |
CN109345005A (en) * | 2018-09-12 | 2019-02-15 | 中国电力科学研究院有限公司 | A kind of integrated energy system multidimensional optimization method based on improvement whale algorithm |
CN109461166A (en) * | 2018-10-26 | 2019-03-12 | 郑州轻工业学院 | A kind of fast-moving target tracking based on KCF mixing MFO |
CN109658436A (en) * | 2018-12-04 | 2019-04-19 | 郑州轻工业学院 | A kind of fast-moving target tracking of mixing GOA and TLBO search |
-
2019
- 2019-05-17 CN CN201910411490.7A patent/CN110136165A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1684543A (en) * | 2004-04-16 | 2005-10-19 | 华为技术有限公司 | Control method for activating time deviation |
CN108010069A (en) * | 2017-12-01 | 2018-05-08 | 湖北工业大学 | Optimize the rapid image matching method of algorithm and grey correlation analysis based on whale |
CN108305272A (en) * | 2018-02-27 | 2018-07-20 | 郑州轻工业学院 | A kind of mutation movement method for tracking target based on ALO search |
CN108389211A (en) * | 2018-03-16 | 2018-08-10 | 西安电子科技大学 | Based on the image partition method for improving whale Optimization of Fuzzy cluster |
CN108896306A (en) * | 2018-03-26 | 2018-11-27 | 四川大学 | Method for Bearing Fault Diagnosis based on adaptive atom dictionary OMP |
CN108460548A (en) * | 2018-05-17 | 2018-08-28 | 西安建筑科技大学 | A kind of more Metal Open production-operational-plan methods based on improvement grey wolf algorithm |
CN109345005A (en) * | 2018-09-12 | 2019-02-15 | 中国电力科学研究院有限公司 | A kind of integrated energy system multidimensional optimization method based on improvement whale algorithm |
CN109461166A (en) * | 2018-10-26 | 2019-03-12 | 郑州轻工业学院 | A kind of fast-moving target tracking based on KCF mixing MFO |
CN109658436A (en) * | 2018-12-04 | 2019-04-19 | 郑州轻工业学院 | A kind of fast-moving target tracking of mixing GOA and TLBO search |
Non-Patent Citations (5)
Title |
---|
CHRISTIAN LGEL 等: "A Computational Efficient Covariance Matrix Updata and a (1+1)-CMA for Evolution Strategies", 《GECOO 2006》 * |
SEYEDALI MIRJALILI 等: "The Whale Optimization Algorithm", 《ADVANCES IN ENGINEERING SOFTWARE》 * |
STEFANKERN 等: "Learning Probability Distributions in Continuous Evolutionary Algorithms – A Comparative Review", 《RESERACHGATE》 * |
魏政磊 等: "控制参数值非线性调整策略的灰狼优化算法", 《空军工程大学学报(自然科学版)》 * |
龙文 等: "收敛因子非线性变化的鲸鱼优化算法", 《兰州理工大学学报》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111598928A (en) * | 2020-05-22 | 2020-08-28 | 郑州轻工业大学 | Abrupt change moving target tracking method based on semantic evaluation and region suggestion |
CN111598928B (en) * | 2020-05-22 | 2023-03-10 | 郑州轻工业大学 | Abrupt motion target tracking method based on semantic evaluation and region suggestion |
CN111696645A (en) * | 2020-06-07 | 2020-09-22 | 长春理工大学 | Hand exoskeleton rehabilitation training device and method based on surface electromyographic signals |
CN112163808A (en) * | 2020-09-23 | 2021-01-01 | 贵州工程应用技术学院 | Method for solving logistics center addressing problem by self-adaptive whale algorithm based on opponent learning |
CN112163808B (en) * | 2020-09-23 | 2022-04-01 | 贵州工程应用技术学院 | Method for solving logistics center addressing problem by self-adaptive whale algorithm based on opponent learning |
CN113043278A (en) * | 2021-04-13 | 2021-06-29 | 南京工业职业技术大学 | Mechanical arm track planning method based on improved whale searching method |
CN113043278B (en) * | 2021-04-13 | 2023-02-03 | 南京工业职业技术大学 | Mechanical arm track planning method based on improved whale searching method |
CN117609768A (en) * | 2024-01-23 | 2024-02-27 | 昆明理工大学 | Vertical water pump unit fault diagnosis method based on improved whale algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110136165A (en) | A kind of mutation movement method for tracking target based on the optimization of adaptive whale | |
Li et al. | Block-wisely supervised neural architecture search with knowledge distillation | |
CN108305272B (en) | Abrupt change moving target tracking method based on ALO search | |
CN107341820B (en) | A kind of fusion Cuckoo search and the mutation movement method for tracking target of KCF | |
Wang et al. | Drpose3d: Depth ranking in 3d human pose estimation | |
CN105678338B (en) | Target tracking method based on local feature learning | |
Erdem et al. | Fragments based tracking with adaptive cue integration | |
CN107247961A (en) | A kind of trajectory predictions method of application blurring trajectorie sequence | |
Li et al. | Online multi-expert learning for visual tracking | |
Zhang et al. | Distractor-aware fast tracking via dynamic convolutions and mot philosophy | |
CN110909591B (en) | Self-adaptive non-maximum suppression processing method for pedestrian image detection by using coding vector | |
Rosenbaum et al. | Learning models for visual 3d localization with implicit mapping | |
Zhao et al. | Generative tracking of 3D human motion by hierarchical annealed genetic algorithm | |
CN106952304A (en) | A kind of depth image computational methods of utilization video sequence interframe correlation | |
Kamyab et al. | Using a self-adaptive neighborhood scheme with crowding replacement memory in genetic algorithm for multimodal optimization | |
CN109461166A (en) | A kind of fast-moving target tracking based on KCF mixing MFO | |
Pan et al. | VA-RED $^ 2$: Video Adaptive Redundancy Reduction | |
Zhang et al. | Sequential 3D Human Pose Estimation Using Adaptive Point Cloud Sampling Strategy. | |
Wang et al. | Achieving cooperation through deep multiagent reinforcement learning in sequential prisoner's dilemmas | |
Luo et al. | Object detection in video with spatial-temporal context aggregation | |
CN114973071A (en) | Unsupervised video target segmentation method and system based on long-term and short-term time sequence characteristics | |
KR20200084467A (en) | Device and method for tracking object in image based on deep learning using rotatable elliptical model | |
CN113763420A (en) | Target tracking method, system, storage medium and terminal equipment | |
CN109658436B (en) | Fast moving target tracking method for mixed GOA and TLBO search | |
Li et al. | Future video frame prediction based on generative motion-assistant discriminative network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190816 |