CN105069445A - Face recognition method based on improved bacterial foraging algorithm - Google Patents
Face recognition method based on improved bacterial foraging algorithm Download PDFInfo
- Publication number
- CN105069445A CN105069445A CN201510592888.7A CN201510592888A CN105069445A CN 105069445 A CN105069445 A CN 105069445A CN 201510592888 A CN201510592888 A CN 201510592888A CN 105069445 A CN105069445 A CN 105069445A
- Authority
- CN
- China
- Prior art keywords
- face
- bacterium
- fitness
- bacterial
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
With the development of science and technology and the advent of the era of big data, the problem of social information security receives much concern. For example, the public security department identifies suspects from the streets, banks need identity authentication of customers, and the customhouse identifies the identity of exit-entry people. The means of identity authentication emerges in endlessly, such as password, fingerprint, ID card, RFID card and other ways of recognition, wherein face recognition is a popular research area at present. In recent years, many excellent talents have conducted research in the field of face recognition, and have made great achievement. In view of the weaknesses and shortcomings of the methods put forward by predecessors, the traditional bacteria foraging method is improved in the invention, and the improved algorithm is applied to face recognition, namely, classification is performed on an original face database by the improved bacteria foraging method first, and then, a target face is recognized by a comparison method. Experiments show that a method of the invention has very high recognition rate. The method of the invention mainly comprises a face input module, a face image preprocessing module, a face feature extraction module, a bacterial foraging algorithm training module, a target face image input module, and a target face image recognition module.
Description
Technical field
The present invention relates to Intelligent Computation Technology, particularly a kind of use improves the technology that bacterial foraging algorithm carries out recognition of face.This technology, in authentication, has a wide range of applications in the fields such as pattern-recognition.
Background technology
Recognition of face is study hotspot in computer vision and area of pattern recognition and difficult point, is the research topic with broad prospect of application.Along with the development of science and technology and the arriving of large data age, social information's safety problem receives much concern.As suspect recognizes from street in public security department, bank needs the authentication of client, the identity validation that customs went out to enter the GATT to every day.The mode of authentication also emerges in an endless stream, as password, and fingerprint, the recognition method such as ID card and RFID radio-frequency card, wherein recognition of face is research field more popular at present, in recent years, the much outstanding talent in the research carrying out field of face identification, and achieves larger achievement.Recognition of face is a special kind of skill utilizing computing machine and recognize face by related algorithm, its algorithm has a variety of, the wherein useful bionic mechanics method for recognition of face, as the intelligent group such as genetic algorithm, ant group algorithm algorithm, but Genetic algorithm searching speed is slow, and ant group algorithm is because adding pheromones, affect recognition efficiency.For this reason, this patent uses a kind of bacterial foraging algorithm of improvement to carry out recognition of face.
Bacterial foraging algorithm (BacterialForagingAlgorithm, BFA)---to look for food optimized algorithm (BacterialForagingOptimizationAlgorithm also referred to as bacterium, BFO/BFOA)---swallowed the behavior of thing at human body intestinal canal endocytosis based on Ecoli Escherichia coli in 2002 by K.M.Passino, a kind of Novel bionic class algorithm of proposition.This algorithm mainly rely on based on distinctive chemotactic behavior, reproductive behavior, migratory behaviour as operator to upgrade position and optimization, realize the evolution of population.Wherein chemotactic behavior is used to the renewal ensureing bacterium position, searches the behavior of more food; Reproductive behavior ensures that bacterial colonies major part remains on the behavior in optimum way; Migratory behaviour is the behavior that the scope allowing bacterial colonies jump out local optimum has reached global optimum.Bacterial foraging algorithm imitates Escherichia coli in intestines, search the community competition coordination mechanism embodied in cooking cycle, and it has Swarm Intelligence Algorithm parallel search, easily departs from the advantages such as local extremum, is one of focus of biologically inspired computation research field.
And the research of current BFO algorithm is scarcely out of swaddling-clothes, the application of BFO is deep not enough, not deep enough especially in recognition of face, and unsatisfactory in the accuracy rate and efficiency of other Swarm Intelligence Algorithms of recognition of face, as genetic algorithm and ant group algorithm etc.It is insensitive that BFO algorithm has initial value and Selecting parameter, and strong robustness, is simply easy to realize, and the advantage such as parallel processing and global search.The situation of the various operators in bacterial foraging algorithm is as follows:
(1) chemotactic operator
Bacterium assembles to the environment being conducive to self and hides the chemotactic operator that the behavior being unfavorable for its own existence region is called bacterium in the algorithm.First every bacterium
bacterial (i)perform chemotactic operator respectively, here each bacterium is made up of two parts, i.e. the position of bacterium and the fitness of bacterium.Bacterial structure i is expressed as follows:
The position encoded representation of bacterium is:
Wherein
c j represent the
jthe classification center of class is a D n dimensional vector n.
In the adjustment process of bacterium position, advance in each bacterium Stochastic choice direction, and whether the environment then evaluating self according to fitness function makes moderate progress.If the fitness of bacterium makes moderate progress, just move on by this direction, until fitness do not improve or reach maximum advance number of times (
n s ); If fitness is improved, just advance in random alternative direction, until every bacterium all complete predetermined chemotactic behavior number of times (
n c ) perform reproductive behavior afterwards.The fitness value of organisms
bacterial (i) .fitnessbe a real number, represent the fitness of organisms, as follows:
Wherein
c j represent the classification center of jth class;
d(
x i , C j ) representative sample to the distance of corresponding classification center, be exactly total the within-cluster variance that calculates of whole denominator in formula and, within-cluster variance is less as can be seen here, and the fitness of bacterium is larger.
The mode of bacterium position adjustment can be expressed as:
Wherein
bacterial (i) .locationbe
ithe current location of individual bacterium,
steprepresent the step-length that bacterium advances at every turn, (i) represents the direction that bacterium rolls at random,
bacterial (j) .locationfor the random site of in current individual field.
(2) operator is bred
Bacterium reach maximum chemotactic behavior number of times (
n c ) after, bacterium will carry out reproductive behavior.The reproductive behavior of bacterium follows the principle of nature " survival of the fittest, the survival of the fittest ".In reproductive behavior, every bacterium according to its execute the environment after chemotactic behavior quality (according to
bacterial (i) .fitnesssize) sort, the organisms of environment difference is dead, and the organisms that environment is good carries out numerous, produces new bacterial flora.Then the new bacterial flora produced again circulate perform chemotactic behavior, reproductive behavior until execute predetermined reproductive behavior number of times (
n re ) perform migratory behaviour afterwards.
(3) operator is migrated
Bacterium place surrounding environment is undergone mutation, and can cause the death of the bacterial population of subregion or force to move on to new region.Show as in bacterium looks for food optimized algorithm and allow some bacteriums necessarily
p ed dead and random site produces new organisms under probability, after whole bacterial colonies has carried out a migratory behaviour, again can carry out chemotactic behavior and reproductive behavior, until execute subscribe migratory behaviour number of times (
n ed ).Population recruitment also exports optimum solution, and algorithm terminates.
The present invention is directed to the weakness of conventional bacteria foraging algorithm, make improvements, and the bacterial foraging algorithm of improvement is applied to recognition of face.Thought based on the recognition of face improving bacterial foraging algorithm is by face being extracted from image after Face datection, form the original storehouse of facial image, pre-service is carried out to the original storehouse of this facial image set up, carry out feature extraction and the feature selecting of face again, as eye feature, the feature extraction of face out, with the ratio of horizontal width between these two with two total horizontal widths, the ratio of left eye horizontal width and two total horizontal widths, left eye vertical height and right eye top are to the right corners of the mouth lower vertical ratio of distances constant, right eye vertical height and the ratio of left eye top to the vertical range of the left side corners of the mouth, the ratio of the horizontal width of face and two total horizontal widths, the ratio of vertical height between the vertical height of face and left eye top to face, select this 6 features altogether.Finally utilize simulating biology intelligent algorithm bacterial foraging algorithm to carry out training study to Sample Storehouse, calculate optimal classification decision data.Target facial image is being carried out pre-service, after the step of feature extraction and feature selecting, is comparing with categorised decision data, judge the classification of target face with Nearest Neighbor Method.
terminological interpretation
Pre-service: namely Image semantic classification, separated by each character image and give identification module identification, this process is called Image semantic classification.Pre-service refers to some preliminary works before pattern-recognition, comprises image purification process, removes the obvious noise (interference) in original image, as far as possible interested target image is separated.As image standardization, image equilibration, denoising and Iamge Segmentation etc.
Feature extraction: exactly physical measurement or conversion are carried out to image model, obtain the digital value that a group reflects its feature.It with object to be identified various physics, the performance of form is all related, and has diversified specific process to obtain the digital value of its reflection feature.
Feature selecting: be after feature extraction, therefrom selects representational or effective composition, and discriminant classification is carried out more fast and accurately.The task of feature selecting is exactly be that to select quantity the feature of D be that one group of optimal characteristics of d (D>d) is come from one group of quantity, determining the separability of object on the one hand, is within the as far as possible short time, find out an optimum stack features on the other hand.
Normalization: make some feature of image have a kind of graphics standard form of invariance under given conversion.Some character of image, the area and perimeter of such as object, just had constant character originally for rotation of coordinate.In the ordinary course of things, the impact of some factor or transfer pair image properties is eliminated by normalized or weakens, thus can be selected as the foundation of measurement image.
Swarm Intelligence Algorithm: be the heuristic search algorithm of based on each population behavior, known target being carried out to optimizing, the individuality distributed in colony is co-operating, so more can adapt to the operative scenario under current environment; Do not have centralized control, such system has more robustness, can not affect solving of whole problem because of the fault of one or several individualities.Can not by directly communicating between individuality but being cooperated by indirect communication, such system has better expandability.The communication overhead of the system increased due to increase individual in system is very little here.In system, the ability of each individuality is very simple, and the execution time of each like this individuality is shorter, and realizes also fairly simple, has simplicity.But for each individuality, its definition itself is relative, and its size and function are determined, so the realization of the optimizing mode of its intelligence is embodied by the general characteristic of whole intelligent group according to solved problem.
Summary of the invention
The object of this invention is to provide a kind of bacterial foraging algorithm of improvement, and utilize the method train face database and learn, thus realize identifying target facial image.In the present invention, to the improvement that the bacterium method of looking for food is carried out, be mainly reflected in following some:
(1)
the improvement of chemotactic operator
The step-length of traditional bacterial foraging algorithm is larger on search efficiency impact, and step-length is excessive more favourable to global search, but can reduce search precision; Step-length is too little can improve local optimum ability, but search efficiency can reduce, and increases the time of optimizing.At this, according to the fitness value dynamic conditioning step-length of bacterium, see following formula:
Wherein
step c (i, j)represent the step-length after i-th bacterium jth time chemotactic;
wthe stride length shrinks coefficient that to be (0,0.95) be selects according to fitness;
step init(
i) represent the initial step length of i-th bacterium;
(2)
migrate the improvement of operator
Migrate operator mainly in order to prevent optimized algorithm to be absorbed in local optimum, to reach global optimum.And migrate operator by migrate probability (
p ed ) control organisms extinction and position migration, that algorithm jumps out local optimum, finally reach the globally optimal solution of algorithm, in traditional bacterial foraging algorithm, migrate probability is a fixed value, the organisms reached near optimum solution is performed and migrates operator, the organisms that fitness is high can be lost.This application dynamic adjustments migrate probability (
p ed ), be shown below:
Wherein
be the jth of i-th bacterium time to migrate migrate probability,
the jth of i-th bacterium fitness value when time to migrate,
minimum fitness value,
maximum fitness value, P
edinitial fixing migrate probability.Three behaviors in bacterial foraging algorithm, chemotactic behavior is the direction determining whether bacterium can be found favorable environment existence and advance, and is core.Important impact is had on convergence is good and bad.Reproductive behavior ensures that bacterium is in favourable environment always to survive, and is conducive to reaching global optimum; Migratory behaviour is conducive to bacterium and jumps out local optimum and reach global optimum, uses in algorithm
b-gd={location, fitness}record globally optimal solution in whole bacterial colonies, represent the optimal location that bacterium occurs and fitness.
In order to reach the object of recognition of face, this patent adopts following technical scheme:
(1) according to actual needs, the face collection needing to carry out retrieving, original face database { F is set up
k;
(2) to the face database { F set up
kcarry out pre-service: the size of image and gray scale are normalized, and then histogram equalization, in utilization local, normalized illumination compensation method eliminates the even impact of uneven illumination, also have face front plan rotate adjustment and the face degree of depth rotate adjustment, finally carry out 3x3 gaussian filtering carry out denoising obtain process after face database { T (F
k);
(3) to pretreated face database { T (F
k) utilize specific method W to carry out feature extraction to the face of face database, extract the position (ELoc of left and right eyes
l, ELoc
r) and size (Width
l, Width
r, Length
l, Length
r) and the position (MLoc) of face and size (Width
m, Length
m), carry out feature selecting again---the ratio of horizontal width and two total horizontal widths between two, the ratio of left eye horizontal width and two total horizontal widths, left eye vertical height and right eye top are to the right corners of the mouth lower vertical ratio of distances constant, right eye vertical height and the ratio of left eye top to the vertical range of the left side corners of the mouth, the ratio of the horizontal width of face and two total horizontal widths, the ratio of vertical height between the vertical height of face and left eye top to face, select this 6 features altogether, obtain the feature database { T of face sample
w(F
k);
(4) the face sample characteristic storehouse { T drawn is utilized
w(F
k) carry out classification based training, calculate optimum face classification identification data, the bacterial foraging algorithm that recycling is improved carries out recognition of face, and concrete sub-step is as follows:
1. initialization correlation parameter.Comprise distance compute type (Euclidean distance), face classification number (
centerNum), bacterial colonies scale (N), chemotactic behavior correlation parameter, reproductive behavior correlation parameter, migratory behaviour correlation parameter;
2. the initialization of bacterial colonies.For the input feature vector of each facial image, it random is assigned as a class, as initial classification, and calculates the center of each classification, as position encoded (Bacterial (i) .location []) of bacterium, and record optimum bacterium;
3. start to enter successively and migrate, breeding and chemotactic behavior.First bacterium starts to overturn Xi (j+1 at random, k, l), move forward with the dynamic step length improved, and calculate fitness (Bacterial (i) .fitness), if fitness makes moderate progress, just upgrade organisms position encoded (Bacterial (i) .location []);
4. for each facial image (Xi), according to the classification center proper vector (C of bacterium
j) value, according to arest neighbors rule, determine face classification, according to the classification results (category) of each bacterium, recalculate classification center proper vector (C
j), upgrade germy fitness (Bacterial (i) .fitness), and record current optimum solution (B_gd);
If 5. carried out chemotactic behavior (
n c ), start to carry out reproductive behavior, current bacterium pressed ranking fitness, by input Reproductive Parameters (
r re ) breed, low for alternative fitness high for fitness, keep high fitness;
If 6. carried out reproductive behavior (
n re ), just by improve dynamically migrate probability (
p ed ) carry out migratory behaviour.To jump out local extremum scope, upgrade the fitness of each bacterium according to method above;
If 7. reach chemotactic (Nc), breeding (Nre) and the number of times of migratory behaviour (Ned), then terminate algorithm, the face characteristic data of output global optimum
b_gd tW (F), otherwise repeat above-mentioned steps.
(5) to target facial image X
icarry out above-mentioned same data acquisition, pre-service and feature extraction, finally obtain the feature T of target facial image
w_xi;
(6) with the grouped data of the recognition of face drawn by bacterial foraging algorithm
b_gd tW (F)carry out Nearest Neighbor Method comparison, find the minimum distance of target face characteristic and grouped data Zhong Lei center, determine the recognition result of target facial image with this.
Advantage major embodiment of the present invention is as follows:
(1) for recognition of face on the basis of Face datection, because the optimizing ability of bacterial foraging algorithm is comparatively strong, there is something special to be not easy to be absorbed in local, and the accuracy rate of recognition of face is higher;
(2) for recognition of face on Face datection basis, because bacterial foraging algorithm also has speed of convergence faster, obtain sooner restraining result relative to genetic algorithm and ant group algorithm;
(3), when database update or supplementary new data, the bacterial foraging algorithm of improvement can be utilized to realize automatic classification.
Accompanying drawing explanation
Fig. 1 is schematic diagram of the present invention;
Fig. 2 is one of original image of the embodiment of the present invention;
Fig. 3 is one of feature extraction result of the embodiment of the present invention;
The recognition of face process flow diagram of Fig. 4 position embodiment of the present invention.
Embodiment
The present embodiment be Dell LATITUDED630 notebook computer on realize, the processor of this machine is Intel (R) Core (TM) 2DuoCPUT72502.00GHz, the operating system used is WindowsXP, and the simulation software of use is matlab2012.As can be seen from Figure 1, the present invention is divided into following several module generally:
(1) face database image imports module: according to actual needs, the face collection needing to carry out retrieving, sets up original face database { F
k;
(2) face database image pre-processing module: to the face database { F set up
kcarry out pre-service: the size of image and gray scale are normalized, and then histogram equalization, in utilization local, normalized illumination compensation method eliminates the even impact of uneven illumination, utilize eyes position carry out the adjustment of face front plan rotation and utilize face to carry out the adjustment of face degree of depth rotation to the distance at face edge in addition, finally carry out 3x3 gaussian filtering and carry out denoising and obtain the face database { T (F after processing
k);
(3) face database characteristic extracting module: to pretreated face database { T (F
k) use gradation of image integral projection figure, gradation of image variance projection figure, second-order differential and the Canny outline map specific method W such as carrying out feature extraction that combines to carry out feature extraction to the face of face database, extract the position (ELoc of left and right eyes
l, ELoc
r) and size (Width
l, Width
r, Length
l, Length
r) and the position (MLoc) of face and size (Width
m, Length
m), carry out feature selecting again---the ratio of horizontal width and two total horizontal widths between two, the ratio of left eye horizontal width and two total horizontal widths, left eye vertical height and right eye top are to the right corners of the mouth lower vertical ratio of distances constant, right eye vertical height and the ratio of left eye top to the vertical range of the left side corners of the mouth, the ratio of the horizontal width of face and two total horizontal widths, the ratio of vertical height between the vertical height of face and left eye top to face, select this 6 features altogether, obtain the feature database { T of face sample
w(F
k);
(4) improve bacterial foraging algorithm training module: utilize the face database feature drawn, use bacterial foraging algorithm to carry out training and learning, calculate optimum face classification identification data, carry out recognition of face with this;
1. initialization correlation parameter.Comprise distance compute type (Euclidean distance), face classification number (
centerNum), bacterial colonies scale (N), chemotactic behavior correlation parameter, reproductive behavior correlation parameter, migratory behaviour correlation parameter;
2. the initialization of bacterial colonies.For the input feature vector of each facial image, it random is assigned as a class, is initial classification, and calculates each classification center, as bacterium position encoded (
bacterial (i) .location []), and record optimum bacterium;
3. start to enter successively and migrate, breeding and chemotactic behavior.First bacterium starts random upset
x i (j+1, k, l), move forward with the dynamic step length improved, and calculate fitness (
bacterial (i) .fitness), if fitness makes moderate progress, just upgrade organisms position encoded (
bacterial (i) .location []);
4. for each facial image (
x i ), according to the classification center proper vector of bacterium (
c j ) value, according to arest neighbors rule, determine face classification, according to the classification results (category) of each bacterium, recalculate classification center proper vector (
c j ), the germy fitness of renewal (
bacterial (i) .fitness), and record current optimum solution (
b_gd);
If 5. carried out chemotactic behavior (
n c ), start to carry out reproductive behavior, current bacterium pressed ranking fitness, by input breeding number of times (
r re ) breed, low for alternative fitness high for fitness, keep high fitness;
If 6. carried out reproductive behavior (
n re ), just by improve dynamically migrate probability (
p ed ) carry out migratory behaviour.To jump out local extremum scope, upgrade the fitness of each bacterium according to method above;
If 7. reach chemotactic (
n c ), breeding (
n re ) and migratory behaviour (
n ed ) number of times, then terminate algorithm, export global optimum face feature vector
b_gd tW (F), otherwise repeat above-mentioned steps.
(5) target face load module: to target facial image X
icarry out the data acquisition that above-mentioned steps (1), (2), (3) are same, pre-service and feature extraction, finally obtain the feature T of target facial image
w_xi;
(6) target face recognition module: with the grouped data of the recognition of face drawn by bacterial foraging algorithm
b_gd tW (F)carry out Nearest Neighbor Method comparison, find the minimum distance of target face characteristic and grouped data Zhong Lei center, determine the recognition result of target facial image with this.
Describe the present invention below in conjunction with accompanying drawing and embodiment:
(1) as required, set up need retrieval face original image storehouse, often open facial image (MxN), as shown in Figure 2 be exactly set up need retrieve face database in a face (80x80);
(2) carry out pre-service to the face database set up, as gray-scale value normalized, 3x3 gaussian filtering carries out denoising and utilizes the image processing methods such as the normalized illumination compensation disposal route in local;
(3) gradation of image integral projection figure is used to face database, gradation of image variance projection figure, second-order differential and canny outline map combine and carry out feature extraction, select face characteristic, draw the correlated characteristic of face, the positional information of the right and left eyes of face and the positional information of face as what extract from Fig. 3, the ratio of horizontal width between two and two total horizontal widths is drawn with this, the ratio of left eye horizontal width and two total horizontal widths, left eye vertical height and right eye top are to the right corners of the mouth lower vertical ratio of distances constant, right eye vertical height and the ratio of left eye top to the vertical range of the left side corners of the mouth, the ratio of the horizontal width of face and two total horizontal widths, the ratio of vertical height between the vertical height of face and left eye top to face, extract 6 character numerical values altogether,
(4) utilize the face database feature drawn, carry out training and learning with the bacterial foraging algorithm improved, calculate optimum face classification identification data, carry out recognition of face with this:
1. initialization correlation parameter.Comprise distance compute type (Euclidean distance), face classification number (
centerNum), bacterial colonies scale (N), chemotactic behavior correlation parameter, reproductive behavior correlation parameter, migratory behaviour correlation parameter.It is below the detailed description of each step for bacterial foraging algorithm training process in Fig. 4;
2. the initialization of bacterial colonies.For the input feature vector often opening facial image, it random is assigned as a class, as initial classification, and calculates each classification center, as position encoded (Bacterial (i) .location []) of bacterium, and record optimum bacterium;
3. start to enter successively and migrate, breeding and chemotactic behavior.First bacterium starts to overturn X at random
i(j+1, k, l), move forward with the dynamic step length improved, and calculate fitness (Bacterial (i) .fitness), if fitness makes moderate progress, just upgrade organisms position encoded (Bacterial (i) .location []);
4. for each facial image (X
i), according to the classification center proper vector (C of bacterium
j) value, according to arest neighbors rule, determine face classification, according to the classification results (category) of each bacterium, recalculate classification center proper vector (Cj), upgrade germy fitness (Bacterial (i) .fitness), and record current optimum solution (B_gd);
If 5. carried out chemotactic behavior (
n c ), start to carry out reproductive behavior, current bacterium pressed ranking fitness, by input breeding number of times (
r re ) breed, low for alternative fitness high for fitness, keep high fitness;
If 6. carried out reproductive behavior (
n re ), just by improve dynamically migrate probability (
p ed ) carry out migratory behaviour.To jump out local extremum scope, upgrade the fitness of each bacterium according to method above;
If 7. reach chemotactic (Nc), breeding (Nre) and the number of times of migratory behaviour (Ned), then terminate algorithm, the face feature vector of output global optimum
b_gd tW (F), otherwise repeat above-mentioned steps.
(5) target facial image X is inputted
i, and pre-service same before carrying out and feature extraction operation, extract the feature T of target facial image
w_xi;
(6) with the feature T of the target facial image extracted
w_xiwith the grouped data of face database
b_gd tW (F)carry out Nearest Neighbor Method comparison, find the minimum distance of target face characteristic and grouped data Zhong Lei center, determine the classification results of target facial image with this.
In embodiments of the present invention, the image in face database is 200 personal images.When carrying out 400 front face images in employing FERET face database, carry out the face recognition experiment of bacterial foraging algorithm, Fig. 4 is the whole process flow diagram of recognition of face.The foregoing is only an embodiment in the present invention, be not limited to the present invention.
Claims (4)
1., based on the face identification method improving bacterial foraging algorithm, it is characterized in that:
Traditional bacterium method of looking for food is improved, then the bacterial foraging algorithm after improving is utilized to carry out face classification identification, that is: the raw data of face database is first obtained, and pre-service is carried out to face database raw data, again feature extraction and feature selecting operation are carried out to face database, then with the bacterial foraging algorithm improved, test training is carried out to the face database proposing face characteristic, draw optimum face classification data, then face to be identified is carried out pre-service, after carrying out feature extraction, identifying operation is carried out with the face of having classified before, finally draw the net result of recognition of face.
2. a kind of face identification method based on improving bacterial foraging algorithm according to claim 1, is characterized in that the improvement showing as chemotactic operator:
The step-length of traditional bacterial foraging algorithm is larger on search efficiency impact, and step-length is excessive more favourable to global search, but can reduce search precision; Step-length is too little can improve local optimum ability, but search efficiency can reduce, and increases the time of optimizing, at this, according to the fitness value dynamic conditioning step-length of bacterium, sees following formula:
Wherein
step c (i, j)represent the step-length after i-th bacterium jth time chemotactic;
wthe stride length shrinks coefficient that to be (0,0.95) be selects according to fitness;
step init(
i) represent the initial step length of i-th bacterium.
3. a kind of face identification method based on improving bacterial foraging algorithm according to claim 1, is characterized in that the improvement of migrating operator:
Migrate operator mainly in order to prevent optimized algorithm to be absorbed in local optimum, to reach global optimum, and migrate operator by migrate probability (
p ed ) control organisms extinction and position migration, that algorithm jumps out local optimum, finally reach the globally optimal solution of algorithm, in traditional bacterial foraging algorithm, migrate probability is a fixed value, the organisms reached near optimum solution is performed and migrates operator, the organisms that fitness is high can be lost, apply in this patent dynamic adjustments migrate probability (
p ed ), be shown below:
Wherein
be the jth of i-th bacterium time to migrate migrate probability,
the jth of i-th bacterium fitness value when time to migrate,
minimum fitness value,
maximum fitness value, P
edbe initial fixing migrate probability, the three behaviors in bacterial foraging algorithm, chemotactic behavior is the direction determining whether bacterium can be found favorable environment and survive and advance, and is core, has important impact to convergence is good and bad; Reproductive behavior ensures that bacterium is in favourable environment always to survive, and is conducive to reaching global optimum; Migratory behaviour is conducive to bacterium and jumps out local optimum and reach global optimum, uses in algorithm
b-gd={location, fitness}record globally optimal solution in whole bacterial colonies, represent the optimal location that bacterium occurs and fitness.
4. a kind of face identification method based on improving bacterial foraging algorithm according to claim 1, its method and technology scheme is as follows:
(1) according to actual needs, the face collection needing to carry out retrieving, original face database { F is set up
k;
(2) to the face database { F set up
kcarry out pre-service: the size of image and gray scale are normalized, and then histogram equalization, in utilization local, normalized illumination compensation method eliminates the even impact of uneven illumination, also have face front plan rotate adjustment and the face degree of depth rotate adjustment, finally carry out 3x3 gaussian filtering carry out denoising obtain process after face database { T (F
k);
(3) to pretreated face database { T (F
k) utilize specific method W to carry out feature extraction to the face of face database, extract the position (ELoc of left and right eyes
l, ELoc
r) and size (Width
l, Width
r, Length
l, Length
r) and the position (MLoc) of face and size (Width
m, Length
m), carry out feature selecting again---the ratio of horizontal width and two total horizontal widths between two, the ratio of left eye horizontal width and two total horizontal widths, left eye vertical height and right eye top are to the right corners of the mouth lower vertical ratio of distances constant, right eye vertical height and the ratio of left eye top to the vertical range of the left side corners of the mouth, the ratio of the horizontal width of face and two total horizontal widths, the ratio of vertical height between the vertical height of face and left eye top to face, select this 6 features altogether, obtain the feature database { T of face sample
w(F
k);
(4) the face sample characteristic storehouse { T drawn is utilized
w(F
k) carry out classification based training, calculate optimum face classification identification data, the bacterial foraging algorithm that recycling is improved carries out recognition of face, and concrete sub-step is as follows:
1. initialization correlation parameter:
Comprise distance compute type (Euclidean distance), face classification number (
centerNum), bacterial colonies scale (N), chemotactic behavior correlation parameter, reproductive behavior correlation parameter, migratory behaviour correlation parameter;
2. the initialization of bacterial colonies:
For the input feature vector of each facial image, it random is assigned as a class, as initial classification, and calculates the center of each classification, as position encoded (Bacterial (i) .location []) of bacterium, and record optimum bacterium;
3. start to enter successively and migrate, breeding and chemotactic behavior:
First bacterium starts to overturn Xi (j+1 at random, k, l), move forward with the dynamic step length improved, and calculate fitness (Bacterial (i) .fitness), if fitness makes moderate progress, just upgrade organisms position encoded (Bacterial (i) .location []);
4. for each facial image (Xi), according to the classification center proper vector (C of bacterium
j) value, according to arest neighbors rule, determine face classification, according to the classification results (category) of each bacterium, recalculate classification center proper vector (C
j), upgrade germy fitness (Bacterial (i) .fitness), and record current optimum solution (B_gd);
If 5. carried out chemotactic behavior (
n c ), start to carry out reproductive behavior, current bacterium pressed ranking fitness, by input Reproductive Parameters (
r re ) breed, low for alternative fitness high for fitness, keep high fitness;
If 6. carried out reproductive behavior (
n re ), just by improve dynamically migrate probability (
p ed ) carry out migratory behaviour, to jump out local extremum scope, upgrade the fitness of each bacterium according to method above;
If 7. reach chemotactic (Nc), breeding (Nre) and the number of times of migratory behaviour (Ned), then terminate algorithm, the face characteristic data of output global optimum
b_gd tW (F), otherwise repeat above-mentioned steps;
(5) to target facial image X
icarry out above-mentioned same data acquisition, pre-service and feature extraction, finally obtain the feature T of target facial image
w_xi;
(6) with the grouped data of the recognition of face drawn by bacterial foraging algorithm
b_gd tW (F)carry out Nearest Neighbor Method comparison, find the minimum distance of target face characteristic and grouped data Zhong Lei center, determine the recognition result of target facial image with this.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510592888.7A CN105069445B (en) | 2015-07-09 | 2015-09-17 | A kind of face identification method based on improvement bacterial foraging algorithm |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201520492397 | 2015-07-09 | ||
| CN2015204923970 | 2015-07-09 | ||
| CN201510592888.7A CN105069445B (en) | 2015-07-09 | 2015-09-17 | A kind of face identification method based on improvement bacterial foraging algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105069445A true CN105069445A (en) | 2015-11-18 |
| CN105069445B CN105069445B (en) | 2019-07-02 |
Family
ID=54498807
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510592888.7A Active CN105069445B (en) | 2015-07-09 | 2015-09-17 | A kind of face identification method based on improvement bacterial foraging algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105069445B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112100632A (en) * | 2020-09-03 | 2020-12-18 | 北卡科技有限公司 | Image steganography method based on bacterial foraging optimization edge detection and XOR coding |
| CN113743266A (en) * | 2021-08-24 | 2021-12-03 | 三峡大学 | Face recognition method based on artificial slime |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101968853A (en) * | 2010-10-15 | 2011-02-09 | 吉林大学 | Improved immune algorithm based expression recognition method for optimizing support vector machine parameters |
| CN102103691A (en) * | 2011-03-14 | 2011-06-22 | 南京邮电大学 | Identification method for analyzing face based on principal component |
| CN103258186A (en) * | 2013-03-17 | 2013-08-21 | 北京工业大学 | Integrated face recognition method based on image segmentation |
-
2015
- 2015-09-17 CN CN201510592888.7A patent/CN105069445B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101968853A (en) * | 2010-10-15 | 2011-02-09 | 吉林大学 | Improved immune algorithm based expression recognition method for optimizing support vector machine parameters |
| CN102103691A (en) * | 2011-03-14 | 2011-06-22 | 南京邮电大学 | Identification method for analyzing face based on principal component |
| CN103258186A (en) * | 2013-03-17 | 2013-08-21 | 北京工业大学 | Integrated face recognition method based on image segmentation |
Non-Patent Citations (2)
| Title |
|---|
| 关良华 等: ""改进的细菌觅食算法在图像分割中的应用"", 《湖南工业大学学报》 * |
| 张璇 等: ""基于细菌觅食优化算法的多阈值图像分割"", 《计算机光盘软件与应用》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112100632A (en) * | 2020-09-03 | 2020-12-18 | 北卡科技有限公司 | Image steganography method based on bacterial foraging optimization edge detection and XOR coding |
| CN112100632B (en) * | 2020-09-03 | 2023-05-23 | 北卡科技有限公司 | Image steganography method based on bacterial foraging optimization edge detection and XOR (exclusive or) coding |
| CN113743266A (en) * | 2021-08-24 | 2021-12-03 | 三峡大学 | Face recognition method based on artificial slime |
| CN113743266B (en) * | 2021-08-24 | 2023-11-24 | 三峡大学 | Human face recognition method based on artificial myxobacteria |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105069445B (en) | 2019-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yi et al. | Age estimation by multi-scale convolutional network | |
| CN111898736B (en) | Efficient pedestrian re-identification method based on attribute perception | |
| CN110263697A (en) | Pedestrian based on unsupervised learning recognition methods, device and medium again | |
| CN105574510A (en) | Gait identification method and device | |
| Parizi et al. | Automatic discovery and optimization of parts for image classification | |
| CN104008395B (en) | A kind of bad video intelligent detection method based on face retrieval | |
| Huang et al. | DeepDiff: Learning deep difference features on human body parts for person re-identification | |
| CN104616316B (en) | Personage's Activity recognition method based on threshold matrix and Fusion Features vision word | |
| CN105574063A (en) | Image retrieval method based on visual saliency | |
| CN112784763A (en) | Expression recognition method and system based on local and overall feature adaptive fusion | |
| CN103854016A (en) | Human body behavior classification and identification method and system based on directional common occurrence characteristics | |
| CN113920472A (en) | Unsupervised target re-identification method and system based on attention mechanism | |
| CN109903339A (en) | A kind of video group personage's position finding and detection method based on multidimensional fusion feature | |
| KR20210151773A (en) | Target re-recognition method and apparatus, terminal and storage medium | |
| CN112883931A (en) | Real-time true and false motion judgment method based on long and short term memory network | |
| Zhou et al. | Fast minutiae extractor using neural network | |
| Zhang et al. | Contour detection via stacking random forest learning | |
| Zhi et al. | Gray image segmentation based on fuzzy c-means and artificial bee colony optimization | |
| CN112200260B (en) | Figure attribute identification method based on discarding loss function | |
| CN105069445A (en) | Face recognition method based on improved bacterial foraging algorithm | |
| CN103207993B (en) | Differentiation random neighbor based on core embeds the face identification method analyzed | |
| CN111898400A (en) | Fingerprint activity detection method based on multi-modal feature fusion | |
| CN105354585A (en) | Improved cat swarm algorithm based target extraction and classification method | |
| CN105893967A (en) | Human body behavior detection method and system based on time sequence preserving space-time characteristics | |
| Li | Global face pose detection based on an improved PSO-SVM method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201030 Address after: 233000 Xianghe Home Floor No.9, Xinhuai Road, Dongsheng Street, Longzihu District, Bengbu City, Anhui Province Patentee after: Bengbu Qibang Science and Technology Information Consulting Co.,Ltd. Address before: 412007 No.88 Taishan Road, Tianyuan District, Zhuzhou, Hunan Patentee before: HUNAN University OF TECHNOLOGY |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230511 Address after: Room 507-1, Building 1, Guangdong Pharmaceutical College, No. 280 Waihuan East Road, Xiaoguwei Street, Panyu District, Guangzhou City, Guangdong Province, 510000 Patentee after: Yilan Technology (Guangzhou) Co.,Ltd. Address before: 233000 No.9, 1st floor, Xianghe Jiayuan, Xinhuai Road, Dongsheng Street, Longzihu District, Bengbu City, Anhui Province Patentee before: Bengbu Qibang Science and Technology Information Consulting Co.,Ltd. |