CN103593651B - Based on gait and the coal mine down-hole personnel authentication identifying method of two dimension discriminant analysis - Google Patents

Abstract

nullBased on gait and the coal mine down-hole personnel personal identification method of two dimension discriminant analysis，First carry out the gait image pretreatment of coal mine down-hole personnel，Then utilize two dimension Discrimination Analysis Algorithm that gait energy diagram is carried out Dimensionality Reduction，Extract the identification feature of all training gait energy diagrams，And by this identification feature input template data base，The identification feature of each gait energy diagram is corresponding with the identity information of the coal mine down-hole personnel pre-deposited in template database，Nearest neighbor classifier is finally used to carry out coal mine down-hole personnel identification，Present invention employs the gait energy diagram Dimensionality Reduction method of two dimension discriminant analysis，Internal distribution and the geometry of gait image data set can not only be disclosed，And more identifying information can be obtained，Light change when it shoots for gait image simultaneously、The problems such as the displacement rotating between gait image have stronger fault-tolerant ability，Recognition speed is fast，Recognition effect is stable，Practical.

Description

Based on gait and the coal mine down-hole personnel authentication identifying method of two dimension discriminant analysis

Technical field

The invention belongs to human body biological characteristics identification technical field, be specifically related to gait recognition method, particularly to based on Gait and the coal mine down-hole personnel personal identification method of two dimension discriminant analysis.

Background technology

Safety of coal mines is related to the security of the lives and property of the people, is related to reform and development and social stability overall situation.Though Most of coal mining enterprise of the most a lot of provinces and cities is equipped with video monitoring system and based on fingerprint, gait, face, iris or the palm Personnel in the pit's attendance recorders such as vein.But due to underground coal mine bad environments, video light photograph is uneven so that face, fingerprint etc. Image blurring and lack color contrast information, the reasons such as target is similar to background so that existing based on face, fingerprint and iris Deng identification system be substantially all and be installed at aboveground or well head, simply play work attendance effect on and off duty, it is impossible to really real Time monitoring and differentiate the activity of personnel in the pit and identity.Gait Recognition is a kind of emerging biometrics identification technology, it is intended to from The variation characteristic between individuality is found and is extracted in identical walking behavior, to realize the automatic identification of human body.Due to step State identification has the unique advantage that other biometrics does not has, i.e. knowledge in the case of remote or low video quality Other potentiality, the highest to image resolution requirement, and gait be difficult to hide or camouflage, affected by environment less, and be one non- Contact far distance identity identifying method, thus authentication identifying method based on gait to be particularly suitable for coal mine down-hole personnel real-time Video monitoring.

Gait Recognition is a kind of newer biometrics that increasing researcher is paid close attention in recent years.Actually The gait video figure that the automatic Gait Recognition system of one intelligent video monitoring is mainly packaged by CCTV camera, computer and one As the software of series processing with identification is formed.Wherein, the software algorithm of Gait Recognition is most critical.For existing based on line Property conversion Coupling Metric learning method can run into dimension disaster and nonlinear model cannot be described very well when solving practical problems The problems such as type, Wang Kejun et al. is by introducing kernel method, it is proposed that a kind of core Coupling Metric learning method, and is applied to gait knowledge In not [Wang Kejun, Yan Tao. core Coupling Metric learning method and the application in Gait Recognition thereof. pattern recognition and artificial intelligence, 2013,Vol.26(2):169-175】；L.Lee et al. with the moment characteristics of gait profile each several part come in analysing gait [Lee L, Grimson W E L.Gait Analysis for Recognition and Classification.Proceedings of the IEEE International Conference on Automatic Face and Gesture Recognition, 2002】；The University of Southampton of Britain achieve in terms of Gait Recognition a lot of achievement in research [Zhang Rong, Vogler C, Metaxas D.Human gait recognition.Computer Vision and Pattern Recognition, 2004,27(02):18-20】.Although the most existing a lot of gait recognition methods and technology, but due to the complexity of gait image And unstability so that gait recognition method and technology are not the most effectively applied to actual coal mine down-hole personnel identity and differentiate system In system, prior art existence is higher to the pre-processing requirements of original gait image, operand is relatively big, system is excessively complicated, collection To gait image data be easily subject to the defects such as the environmental effect such as illumination, position.

Summary of the invention

In order to overcome the shortcoming of above-mentioned prior art, it is an object of the invention to provide based on gait and two dimension discriminant analysis Coal mine down-hole personnel personal identification method, have and extract that the identification characteristic velocity of gait energy diagram is fast, identification accuracy The advantages such as high, recognition effect is stable and practical.

In order to achieve the above object, the technical solution adopted in the present invention is:

Based on gait and the coal mine down-hole personnel personal identification method of two dimension discriminant analysis, specifically include following steps:

The first step, the gait image pretreatment of coal mine down-hole personnel: the gait image photographed by video camera directly inputs To computer, shooting does not has the image of gait as background model, after the gait image photographed and background model are subtracted each other Gait image sequence after elimination background；Then use the rectangle frame that ratio is 120:80 of height and width to carry out frame and live gait wheel Exterior feature, the height of a height of body gait image profile of rectangle, the center in rectangular horizontal direction is body gait image profile barycenter Horizontal coordinate, intercept out by this rectangle frame, and be normalized to 120 × 80 sizes by the scaling of 1:1, then carry out two-value Change processes, and obtains the gait image sequence B of binaryzation_t(x y), determines between continuous 2 minimum altitudes when people walks Time difference as a gait cycle, utilize following formula (1) to be synthesized through average method by the gait image in a cycle One width gait energy diagram,

$G(x,y)=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}{B}_t(x,y)---\left(1\right)$

In formula, N is the length of complete gait cycle sequence, and t is the time, and x, y are two dimensional image plane coordinate；

Second step, utilizes two dimension Discrimination Analysis Algorithm that gait energy diagram is carried out Dimensionality Reduction, extracts all training gaits The identification feature of energy diagram, and by this identification feature input template data base, the identification feature of each gait energy diagram is with in advance The identity information of the coal mine down-hole personnel being stored in template database is corresponding, and feature extraction sequentially includes the following steps:

If K class n width gait energy diagram is { G₁,G₂,...,G_n, class number C₁,C₂,...,C_KRepresent, and set C_iClass There is n_iIndividual width image,

Calculate all gait energy diagrams and C_iThe meansigma methods of class gait energy diagram is respectively

$\left\{\begin{array}{c}\stackrel{\‾}{G}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{G}_i\\ \stackrel{\‾}{{G_C}_i}=\frac{1}{n_i}\underset{j=1}{\overset{n_i}{\mathrm{\Σ}}}{G}_j\end{array}\right.---\left(2\right)$

Calculate class scatter matrix S_BWith Scatter Matrix S in class_WIt is respectively

$\left\{\begin{array}{c}{S}_B=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}P\left(C_i\right)(\stackrel{\‾}{{G_C}_i}-\stackrel{\‾}{G}){(\stackrel{\‾}{{G_C}_i}-\stackrel{\‾}{G})}^T\\ S_W=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{n_i}{\mathrm{\Σ}}}P\left(C_i\right){(G_j-\stackrel{\‾}{{G_C}_i})(G_j-\stackrel{\‾}{{G_C}_i})}^T/n_i\end{array}\right.---\left(3\right)$

In formula, P (C_i) it is pattern C_iPrior probability, take

By S_BAnd S_WSet up objective optimization function

J (A)=A^T(S_B-S_W)A (4)

In formula, A is mapping matrix to be asked, and T is matrix transpose computing,

To the S in formula (4)_B-S_WCarry out Eigenvalues Decomposition, calculate (S_B-S_W) d of a=λ a maximum eigenvalue λ₀,λ₁, L,λ_d-1(λ₀≥λ₁≥L≥λ_d-1) corresponding yojan characteristic vector a₁,a₂,...,a_d, wherein λ is characterized value, d be low dimensional feature to The dimension of amount,

By following formula (5) by [a₁,a₂,...,a_d] be converted to an orthogonal matrix P=[p₁, p₂..., p_d],

$p_k=a_k-\underset{i=1}{\overset{k-1}{\mathrm{\Σ}}}\frac{p_i^T{a}_k}{p_i^T{p}_i}{p}_i---\left(5\right)$

By following formula (6) by all of gait energy diagram { G₁,G₂,...,G_nIt is respectively mapped to lower-dimensional subspace,

G_i→Y_i=P^TG_i (6)

By { Y₁,Y₂,...,Y_nBuild the feature templates of coal mine down-hole personnel identification, formula (6) appoint to be tested Anticipate a width gait energy diagram G_newIt is mapped as low-dimensional characteristic image, i.e. G_new→Y_new=P^TG_new；

3rd step, uses nearest neighbor classifier to carry out coal mine down-hole personnel identification: to utilize euclidean distance metric conduct Distance between any two matrix A, B, i.e. d (A, B)=| | A-B | |,

All training gait energy diagram { G are obtained by second step₁,G₂,...,G_nLow-dimensional mapping matrix be respectively { Y₁, Y₂,...,Y_n, for the gait energy diagram G that any one is to be tested_newLow-dimensional mapping matrix Y_new, calculate Y respectively_newWith {Y₁,Y₂,...,Y_nDistance between }, ifThen will test gait energy diagram G_newIt is judged to kth Class, thereby determines that the identity of coal mine down-hole personnel to be identified.

Present invention employs the gait energy diagram Dimensionality Reduction method of two dimension discriminant analysis, gait image can not only be disclosed The internal distribution of data set and geometry, and more identifying information can be obtained, it shoots for gait image simultaneously Time light change, the problem such as displacement rotating between gait image there is stronger fault-tolerant ability；And due at gait figure As directly processing each gait two dimensional image matrix data during data dimension yojan, so recognition speed is fast, recognition effect Stable, practical.

Accompanying drawing explanation

Accompanying drawing is the FB(flow block) of the present invention.

Detailed description of the invention

The present invention is described in detail below in conjunction with the accompanying drawings.

Referring to the drawings, based on gait and the coal mine down-hole personnel personal identification method of two dimension discriminant analysis, including gait figure As pretreatment, train and identify three processes, specifically including following steps:

The first step, the gait image pretreatment of coal mine down-hole personnel: the gait image photographed by video camera directly inputs To computer, shooting does not has the image of gait as background model.Gait image pretreatment includes the location of gait image, gait The extraction in gait region, center, denoising, gait image normalization and the gait image enhancing etc. of image；Then, will photograph Body gait image and the background model without human body are eliminated after subtracting each other the gait image sequence after background.Use one high and The rectangle frame that ratio is 120:80 of width carrys out frame and lives body gait profile.Intercept out by this rectangle frame, and press the scaling of 1:1 It is normalized to 120 × 80 sizes；Carry out binary conversion treatment again, obtain the gait image sequence B of binaryzation_t(x,y)；Determine one The time difference between continuous 2 minimum altitudes when people walks is as a gait cycle.The gait image in one cycle is entered Row averagely synthesizes gait energy diagram, and computing formula is as follows

$G(x,y)=\frac{1}{N}\underset{t=1}{\overset{N}{\mathrm{\Σ}}}{B}_t(x,y)---\left(1\right)$

In formula, N is the length of complete gait cycle sequence, and t is the time, and x, y are two dimensional image plane coordinate.

Gait energy diagram contains many gait information such as phase place, frequency, profile, can reflect gait motion feature, simultaneously Again data volume is reduced to original 1/tens.Again all of gait energy diagram is divided into training set and test set.? Before being identified task, need pretreated training sample to be stored in data base.In the process, it is generally required to check and be stored in The quality of the training sample of data base, it is ensured that the feature of training sample high for quality is stored in data base, simultaneously by device for underground man car The personal information that member is corresponding, as name, No. ID etc. be stored in identification system.

Second step, utilizes two dimension Discrimination Analysis Algorithm that gait energy diagram carries out Dimensionality Reduction, extracts mapping matrix and low Dimensional feature masterplate, is stored in masterplate data base.The identification feature of each gait energy diagram with pre-deposit in template database The personal information of coal mine down-hole personnel is corresponding.Identify that feature is extracted according to the following steps:

If K class n width gait energy diagram is { G₁,G₂,...,G_n, class number C₁,C₂,...,C_KRepresent, and set C_iClass There is n_iWidth energy diagram, then

Calculate all gait energy diagrams and C_iThe meansigma methods of class gait energy diagram is respectively

$\left\{\begin{array}{c}\stackrel{\‾}{G}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{G}_i\\ \stackrel{\‾}{{G_C}_i}=\frac{1}{n_i}\underset{j=1}{\overset{n_i}{\mathrm{\Σ}}}{G}_j\end{array}\right.---\left(2\right)$

Calculate class scatter matrix S_BWith Scatter Matrix S in class_WIt is respectively

$\left\{\begin{array}{c}{S}_B=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}P\left(C_i\right)(\stackrel{\‾}{{G_C}_i}-\stackrel{\‾}{G}){(\stackrel{\‾}{{G_C}_i}-\stackrel{\‾}{G})}^T\\ S_W=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{n_i}{\mathrm{\Σ}}}P\left(C_i\right){(G_j-\stackrel{\‾}{{G_C}_i})(G_j-\stackrel{\‾}{{G_C}_i})}^T/n_i\end{array}\right.---\left(3\right)$

In formula, P (C_i) it is pattern C_iPrior probability, take

By S_BAnd S_WSetting up objective optimization function is

J (A)=A^T(S_B-S_W)A (4)

In formula, A is mapping matrix to be tried to achieve, and T is matrix transpose computing.

Calculate (S_B-S_W) d of a=λ a maximum eigenvalue λ₀,λ₁,L,λ_d-1(λ₀≥λ₁≥L≥λ_d-1) characteristic of correspondence Vector is a₁,a₂,...,a_d, wherein λ is characterized value, and d is the dimension of low dimensional feature vector.By [a₁,a₂,...,a_d] be converted to one Individual orthogonal matrix P=[p₁,p₂,...,p_d], conversion formula is as follows,

$p_k=a_k-\underset{i=1}{\overset{k-1}{\mathrm{\Σ}}}\frac{p_i^T{a}_k}{p_i^T{p}_i}{p}_i---\left(5\right)$

By P=[p₁,p₂,...,p_d] by all gait energy diagram { G in training set₁,G₂,...,G_nBe respectively mapped to low N-dimensional subspace n, obtains { Y₁,Y₂,...,Y_n, i.e. G_i→Y_i=P^TG_i.By { Y₁,Y₂,...,Y_nBuild the knowledge of coal mine down-hole personnel identity Another characteristic template.The low-dimensional feature masterplate of each gait training gait image collection and the coal mine down-hole personnel being previously stored in Personally identifiable information such as name, the correspondence such as No. ID.

By any one width gait energy diagram G of concentration to be tested_newIt is mapped as low-dimensional characteristic image, i.e.

G_new→Y_new=P^TG_new (6)

3rd step: use nearest neighbor classifier to carry out coal mine down-hole personnel identification: to utilize euclidean distance metric conduct Distance between any two matrix A, B, i.e. d (A, B)=| | A-B | |.The low-dimensional of the gait energy diagram of personnel to be identified is special Levy Y_newWith the gait feature template { Y in data base₁,Y₂,...,Y_nCarry out " one-to-many " search coupling, find out Y_newWith {Y₁,Y₂,...,Y_nThe label that distance between } is minimum, it follows that the conclusion of " whom this person is ".

Claims (1)

1. based on gait and the coal mine down-hole personnel personal identification method of two dimension discriminant analysis, it is characterised in that include following step Rapid:

The first step, the gait image pretreatment of coal mine down-hole personnel: the gait image photographed by video camera is directly inputted to meter Calculation machine, shooting does not has the image of gait as background model, is disappeared after the gait image photographed and background model being subtracted each other Except the gait image sequence after background；Then use the rectangle frame that ratio is 120:80 of height and width to carry out frame and live gait profile, The height of a height of body gait image profile of rectangle, the center in rectangular horizontal direction is the water of body gait image profile barycenter Flat coordinate, intercepts out by this rectangle frame, and is normalized to 120 × 80 sizes by the scaling of 1:1, then carries out at binaryzation Reason, obtains the gait image sequence B of binaryzation_t(x, y), determine between continuous 2 minimum altitudes when people walks time Between differ from as a gait cycle, utilize following formula (1) that through average method, the gait image in one cycle is synthesized a width Gait energy diagram,

$G(x,y)=\frac{1}{N}\underset{t=1}{\overset{N}{\Σ}}{B}_t(x,y)---\left(1\right)$

In formula, N is the length of complete gait cycle sequence, and t is the time, and x, y are two dimensional image plane coordinate；

Second step, utilizes two dimension Discrimination Analysis Algorithm that gait energy diagram is carried out Dimensionality Reduction, extracts all training gait energy The identification feature of figure, and by this identification feature input template data base, the identification feature of each gait energy diagram with pre-deposit The identity information of the coal mine down-hole personnel in template database is corresponding, and feature extraction sequentially includes the following steps:

If K class n width gait energy diagram is { G₁,G₂,...,G_n, class number C₁,C₂,...,C_KRepresent, and set C_iApoplexy due to endogenous wind has n_i Width image,

Calculate all gait energy diagrams and C_iThe meansigma methods of class gait energy diagram is respectively

$\left\{\begin{array}{c}\stackrel{\‾}{G}=\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}{G}_i\\ \stackrel{\‾}{G_{C_i}}=\frac{1}{n_i}\underset{j=1}{\overset{n_i}{\Σ}}{G}_j\end{array}\right.---\left(2\right)$

Calculate class scatter matrix S_BWith Scatter Matrix S in class_WIt is respectively

$\left\{\begin{array}{c}{S}_B=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}P\left(C_i\right)\left(\stackrel{\‾}{G_{C_i}}-\stackrel{\‾}{G}\right){\left(\stackrel{\‾}{G_{C_i}}-\stackrel{\‾}{G}\right)}^T\\ S_W=\underset{i=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{n_i}{\mathrm{\Σ}}}P\left(C_i\right)\left(G_j-\stackrel{\‾}{G_{C_i}}\right){\left(G_j-\stackrel{\‾}{G_{C_i}}\right)}^T/n_i\end{array}\right.---\left(3\right)$

In formula, P (C_i) it is C_iThe prior probability of class, takes

By S_BAnd S_WSet up objective optimization function

J (A)=A^T(S_B-S_W)A (4)

In formula, A is mapping matrix to be asked, and T is matrix transpose computing,

To the S in formula (4)_B-S_WCarry out Eigenvalues Decomposition, calculate (S_B-S_W) d of a=λ a maximum eigenvalue λ₀,λ₁,L,λ_d-1 (λ₀≥λ₁≥L≥λ_d-1) corresponding yojan characteristic vector a₁,a₂,...,a_d, wherein λ is characterized value, and d is low dimensional feature vector Dimension,

By following formula (5) by [a₁,a₂,...,a_d] be converted to an orthogonal matrix P=[p₁,p₂,...,p_d],

$p_k=a_k-\underset{i=1}{\overset{k-1}{\Σ}}\frac{p_i^T{a}_k}{p_i^T{p}_i}{p}_i---\left(5\right)$

By following formula (6) by all of gait energy diagram { G₁,G₂,...,G_nIt is respectively mapped to lower-dimensional subspace,

G_i→Y_i=P^TG_i (6)

By { Y₁,Y₂,...,Y_nBuild the feature templates of coal mine down-hole personnel identification, by formula (6) by be tested any one Width gait energy diagram G_newIt is mapped as low-dimensional characteristic image, i.e. G_new→Y_new=P^TG_new；

3rd step, uses nearest neighbor classifier to carry out coal mine down-hole personnel identification: to utilize euclidean distance metric as arbitrarily Distance between two matrix A, B, i.e. d (A, B)=| | A-B | |,

All training gait energy diagram { G are obtained by second step₁,G₂,...,G_nLow-dimensional mapping matrix be respectively { Y₁,Y₂,..., Y_n, for the gait energy diagram G that any one is to be tested_newLow-dimensional mapping matrix Y_new, calculate Y respectively_newWith { Y₁, Y₂,...,Y_nDistance between }, ifThen will test gait energy diagram G_newIt is judged to kth class, Thereby determine that the identity of coal mine down-hole personnel to be identified.