CN103377293B - The holographic touch interactive exhibition system of multi-source input, information intelligent optimization process - Google Patents

Abstract

The invention discloses the holographic touch interactive exhibition system of the input of a kind of multi-source, information intelligent optimization process.This system, by multisensor perception ambient condition information, when without audient, starts the generation of cigarette propagated sensation sensor monitoring fire; There being audient to deposit in case, the face in video being identified, recommending suitable information broadcasting, increase the acceptance of client.This display systems introduces interactive mode, and audient can realize human-computer interaction by holographic touch screen, obtains the information wanting to know.Simultaneously, the characteristic sum intelligent information mining algorithm extracted by color co-occurrence matrix, the searching area-of-interest of expression recognition and feature extraction are combined into a step, utilize neighbour's propagation clustering algorithm to carry out cluster to face, according to mood change information broadcasting information.It effectively can push information according to the age of audient, sex under complex background environment, and it fully can be shown and lead audient to understand information more comprehensively, increases the acceptance of audient.

Description

The holographic touch interactive exhibition system of multi-source input, information intelligent optimization process

Technical field

The invention belongs to multimedia information dissemination field, the holographic touch interactive exhibition system of particularly a kind of multi-source input, information intelligent optimization process.

Background technology

Along with the fast development of modern society, multimedia system is just penetrating into the every field of life with extremely strong vitality, as advertisement, game, education, archives, books, amusement, art, financial transaction etc.Advertising display conducts promotion to technology flexibly as one, is subject to the great attention of government bodies, enterprises and institutions and every profession and trade individuality.It not only can spreads out merchandise information, Branding image, the aesthetic requirement meeting consumer, guide consumption guiding, and simultaneously it can also excite the moral consciousness of people, causes thinking, the effect having education and inspire.Current advertising display technology mainly considers the unidirectional propelling movement of advertisement, relies on existence and repeatability to reach the effect of publicizing widely.Particularly, giant-screen LCD information display system (as building elevator LCD ad system) application at present widely.It has powerful video, animation, text ability to play, but these systems are all unidirectional, and audience can only passive viewing.Much there is colony and promote mistake in therefore current display systems, the stage of even offensive detest, does not consider that audient experiences and demand, does not more consider the participation of audient completely.Meanwhile, to consume cost huge, cause the significant wastage of social electric power, space, human resources.

China Patent No.: CN101620715A, discloses a kind of intelligent advertisement dissemination method.It, by inputting picture or the video of scene, scans picture or video, and identifies picture or video, extracts user's characteristic information; Whether the characteristic information comparing gathered user or special user mates with the user's characteristic information prestored; Search the advertisement matched with this user's characteristic information in advertising resource storehouse; Export advertisement.

China Patent No.: CN102129644A, discloses a kind of smart ads system with audient's characteristic perception and statistical function.It is by camera collection image, and the image information that collect is sent into computing equipment, computing equipment is responsible for the image obtained from collecting device to carry out processing, computing according to operation result control electric advertisement screen to realize the storage of ad content, display, broadcasting and switching.

There is following weak point in above-mentioned ad system:

(1) above-mentioned patent mainly utilizes the form of image library or image dictionary to mate the face in video or scene picture, determine uniqueness and the uniqueness of audient, be not suitable for the show surroundings of complexity or audient's change, require high when image library data volume is large to computing machine simultaneously, the requirement of system real time can not be met.

(2) fail the mood of audient to take into account, can not according to the broadcasting of the mood self-adaptative adjustment information of audient.

(3) fail to change the passive state of receiving information of audient, client by person operating, with Media, cannot obtain the information wanting to know.

(4) when there being audient, suitable method is not had to promote customer share; In the situation without audient, fail to economize on resources and improve system works efficiency.

Summary of the invention

The object of the invention is to solve the problem, the holographic touch interactive exhibition system of the input of a kind of multi-source, information intelligent optimization process is provided.This system, by multisensor perception ambient condition information, when without audient, starts the generation of cigarette propagated sensation sensor monitoring fire; There being audient to deposit in case, the face in video being identified, recommending suitable information broadcasting, increase the acceptance of client.This display systems introduces interactive mode, and audient can realize human-computer interaction by holographic touch screen, and inquiry client wants the information known.Simultaneously, the characteristic sum intelligent information mining algorithm extracted by color co-occurrence matrix, the searching area-of-interest of expression recognition and feature extraction are combined into a step, utilize neighbour's propagation clustering algorithm to carry out cluster to face, according to mood change information broadcasting information.

Technical solution of the present invention is as follows:

The holographic interaction of multi-source information input, Intelligent Recognition process touches a display systems, comprises the multisensor, computing equipment and the holographic interaction that gather environmental information and touches display device,

(1) open multisensor awareness apparatus monitoring surrounding enviroment, whether detection infrared sensor, sound transducer exist signal.When the equal no signal of infrared sensor, sound transducer, open the generation of cigarette sense sensor monitoring of environmental fire news feelings; When infrared sensor or sound transducer have signal, start camera and interactive touch display device;

It can be the movable platform being placed on indoor or outdoors that holographic interaction touches display systems, therefore environment residing for it may be changeable, use video capture device very uneconomical round-the-clockly, therefore when without audient, display platform closes screen harvester, stop the broadcasting of information, open environment monitoring device, prevent the generation of the dangerous situations such as fire.

(2) reading images frame or sequence of image frames from video information.Face datection and location, specifically determine whether there is face in input picture, if there is face, then determines the position and size etc. of one or more face.Facial feature extraction, the face detected extracts the feature relevant to sex, age and mood.Conventional feature comprises with color and the Luminance Distribution information global feature that is representative and the local feature that is representative with the position of face and shape profile information.According to the facial characteristics extracted, estimation is made to age of face in image and sex.

(3) intelligent information mining algorithm and neighbour's propagation clustering algorithm is adopted to carry out INTELLIGENT IDENTIFICATION to human face expression and are classified.The characteristic sum intelligent information mining algorithm extracted by color co-occurrence matrix, is combined into a step by the searching area-of-interest of expression recognition and feature extraction, utilizes neighbour's propagation clustering algorithm to carry out cluster to face.

(4) information customization is thrown in and information recommendation input.Inquiry face whether mate with the face in image library, have no touch perceptual signal and human face expression whether positive? do not mate with image library at face, no touch perceptual signal, and the recommendation carrying out information when human face expression is positive is play.When there being touch perception information, stop playing current information, the touch inquire request or the broadcasting audient that respond audient fast specify information.When face mates with image library, do not interrupting preferential its information interested of broadcasting under the current prerequisite just at information broadcasting; When human face expression is not positive, dynamic conditioning shows information classification.

(5) if in the unclosed situation of current presentation information, continue to show current information, otherwise, return step (1).

The multisensor of described collection environmental information comprises: one or more in camera, infrared sensor, microwave remote sensor, ultrasonic sensor, laser sensor, magnetic switch sensor, sound transducer or cigarette propagated sensation sensor, for obtaining the information such as environment temperature, humidity, illumination, time, and audient's fingerprint in measurement environment, height, body weight, blood pressure, heartbeat, body temperature etc.

Touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, its hardware configuration is divided into following two parts.

One, local area control structure.A computing machine at least can drive one or more touch display device, and each display device can be equipped with one group of sensor senses equipment.Multisensor awareness apparatus can be camera, infrared sensor, microwave remote sensor, ultrasonic sensor, laser sensor, magnetic switch sensor, sound transducer and cigarette sense sensor.

Two, wide area syndeton.Multiple stage computing machine can network consisting system, can transmit in a wired or wireless manner, can form bus-type, star-like, ring-like, the network topology structure such as tree-shaped or mixed type.In addition, by various computing machine, external unit and database link up mutually the information handling system of composition, can realize video display file management, application software is shared, show the functions such as schedule, Email and facsimile service service.

In above-mentioned steps (2), Face datection and location, feature extraction is mainly divided into following steps: first, determines whether there is face in input picture, if there is face, then the one or more face of quick position position in the picture and scope.Secondly, pretreatment image, makes the face location in facial image, size and image normalization, and overcome block, attitude, illumination, the factor such as jewelry impact.Again, the face detected extracts the feature relevant to sex and age.Conventional feature comprises with the Luminance Distribution information global feature that is representative and the local feature that is representative with the position of face and shape profile information.Finally, according to the facial characteristics extracted, the sex of face to be identified is judged and estimates the age.

In above-mentioned steps (3), the feature that color co-occurrence matrix extracts comprises and represents redness in RGB color space model and green component, the chromatic component in hsv color spatial model and gray component.For above-mentioned arbitrary component, can comprise multiple characteristic quantity again, these characteristic quantities can be the contrasts of color, energy, entropy, one or more in consistance or correlativity etc.

In above-mentioned steps (3), intelligent information mining algorithm, is made up of the disaggregation of intelligent information mining algorithm multiple candidate solution, and each candidate solution all has and represented by area-of-interest component and characteristic component, and its detailed process is as follows:

1) random initializtion disaggregation;

2) the evaluation function value of each candidate solution is calculated;

3) evaluation function value is greater than the candidate solution of average ratings functional value as input, calls reverse cloud model, obtain the numerical characteristic of Multidimensional Cloud Model;

4) self-adaptative adjustment numerical characteristic;

5) using revised numerical characteristic as input, call forward cloud model, generate new candidate solution, form disaggregation of new generation;

6) if meet end condition, elite's candidate solution and evaluation function value thereof is exported; Otherwise turn to step (2).

In above-mentioned steps (3), neighbour's propagation clustering algorithm using all sample datas as initial representative sample, and each sample data is as potential cluster centre, constantly updated Attraction Degree and the degree of membership of each point by iterative process, until automatically produce several class centers.Concrete steps are as follows:

(a) initial similarity matrix;

B () calculates the attraction angle value between sample point;

C () calculates the ownership angle value between sample point;

D () upgrades Attraction Degree and ownership angle value;

If e () iterations exceedes the maximal value of setting or stops when cluster centre does not change in several times iteration calculating, determine class center and all kinds of sample points; Otherwise return (b).

In above-mentioned steps (4), when information customization is thrown in and information recommends to throw in, follow several principles below: one, respond touch signal at any time; Two, do not interrupting under the current prerequisite just at information broadcasting, preferential answering image library matched signal, carrying out information customization according to match information and show; Three, carry out information broadcasting according to human facial expression information.When human face expression is positive, continue the broadcasting of current information, and recommend the recommendation of next round kind attributes information to play; When human face expression is not positive, dynamic conditioning shows information classification.

The beneficial effect that the present invention reaches:

The present invention is the holographic touch interactive exhibition system of the input of a kind of multi-source, information intelligent optimization process, touch display device by many sensings input equipment, computing equipment and holographic interaction to form, modular construction, is convenient to client and carries out textural association and software adjustment as required.The feature of its existing billboard, can reach again the effect of film advertisement, and audient can also be made to participate in carrying out effective interaction, is desirable integrated communication media system.It effectively can push information according to the age of audient, sex under complex background environment, according to the mood of audient and can also touch perception information self-adaptative adjustment information.It fully can be shown and lead audient to understand information more comprehensively, increases the acceptance of audient.

Accompanying drawing explanation

Fig. 1 overall construction drawing;

Fig. 2 principle of work process flow diagram;

The process flow diagram of Fig. 3 intelligent information mining algorithm;

Fig. 4 candidate solution coding structure figure;

Fig. 5 expresses one's feelings the process flow diagram of INTELLIGENT IDENTIFICATION;

The local of Fig. 6 many groups display systems and wide area annexation figure.

Embodiment

A preferred embodiment of the present invention is specifically introduced below in conjunction with accompanying drawing 1-6.

Referring to Fig. 1, touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, issue various information for enterprises and institutions or individual customer, its general structure is mainly divided into three large modules:

(1.1) acquisition module of Multiple Source Sensor signal, be made up of, as camera, infrared sensor, microwave remote sensor, ultrasonic sensor, laser sensor, magnetic switch sensor, sound transducer and cigarette sense sensor one or more sensors below.Be mainly used in the situation of audient in monitoring of environmental, and the generation of the mishap such as fire.

(1.2) computing equipment module can be computing machine or embedded computer, carries out intellectual analysis and process for the information obtained multisensor;

(1.3) holographic touch display device, touch-screen is combined with LCD/LED display screen by nanometer touch membrane, adopts tempered glass protection, support LAN (Local Area Network) or internet Long-distance Control time suitable, can accurately touch-control Query Information, interactive with audient.

Referring to Fig. 2, touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, its principle of work flow process comprises the steps:

(2.1) open multisensor awareness apparatus monitoring surrounding enviroment, whether detection infrared sensor, sound transducer exist signal.When the equal no signal of infrared sensor, sound transducer, open cigarette sense sensor monitoring of environmental fire news feelings; When infrared sensor or sound transducer have signal, start camera and interactive touch presentation device;

(2.2) reading images frame or sequence of image frames from video information.

(2.3) Face datection and location, specifically determines whether there is face in input picture, if there is face, then determines the position and size etc. of one or more face.Facial feature extraction, the face detected extracts the feature relevant to sex, age and mood.Conventional feature comprises with color and the Luminance Distribution information global feature that is representative and the local feature that is representative with the position of face and shape profile information.

(2.4) according to the facial characteristics extracted, estimation is made to age of face in image and sex, identify the expression of face and classify.

(2.5) inquire about face whether mate with the face in image library, have no touch perceptual signal and human face expression whether positive? do not mate with image library at face, no touch perceptual signal, and the recommendation carrying out information when human face expression is positive is play.When there being touch perception information, stop playing current information, the touch inquire request or the broadcasting audient that respond audient fast customize information.When face mates with image library, preferential its information interested of broadcasting; When human face expression is not positive, dynamic conditioning shows information classification.

(2.6) if in the unclosed situation of current presentation information, continue to show current information, otherwise, return step (2.1).

Referring to Fig. 3, touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, the workflow diagram of the intelligent information mining algorithm that it adopts comprises the steps:

(3.1) random initializtion disaggregation X (0).Without loss of generality, consider that one containing N number of candidate solution, code length is candidate's disaggregation (hereinafter referred to as the disaggregation) X={x of L ₁, x ₂..., x _n∈ S ⁿ, wherein S is candidate solution space encoder, S ⁿfor N rank disaggregation space encoder.Disaggregation X can use vector matrix representation below:

$X=\left[\begin{array}{c}{x}_1\\ x_2\\ .\\ .\\ .\\ x_N\end{array}\right]=\left[\begin{array}{cccc}{x}_{11}& x_{12}& \mathrm{...}& x_{1L}\\ x_{21}& x_{22}& \mathrm{...}& x_{2L}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ x_{N1}& x_{N2}& \mathrm{...}& x_{NL}\end{array}\right]=\[X_1,X_2,\mathrm{...},X_L\]$

Wherein x _i=(x _i1, x _i2..., x _iL) ∈ S represents i-th candidate solution in disaggregation X; x _ijrepresent that the jth of i-th candidate solution separates constituent element; I=1,2 ..., N, j=1,2 ..., L.

(3.2) the evaluation function value f (x of each candidate solution is calculated _i(t)), disaggregation average ratings functional value with the best evaluation function value of disaggregation be determine elite's candidate solution x _b(t), wherein $\begin{array}{ccc}\stackrel{\‾}{f}\left(X\right(t\left)\right)={\Σ}_{i=1}^{N}f\left(x_i\right(t\left)\right)/N,& \hat{f}\left(X\right(t\left)\right)=\max f\left(x_i\right(t\left)\right),& \left\{x_b\left(t\right)\right|f\left(x_b\right(t\left)\right)=\hat{f}\left(X\right(t\left)\right),b\∈\[1,N\],t\≥0\}.\end{array}$

(3.3) evaluation function value is greater than the candidate solution of average ratings functional value as input, calls reverse cloud model, obtain the numerical characteristic of Multidimensional Cloud Model.Specifically can be divided into following steps again:

(3.3.1) from disaggregation X (t), select evaluation function value to be greater than the candidate solution of disaggregation average ratings functional value, namely meet candidate solution x _i(t), i=1,2 ... k, wherein 1≤k≤N.

(3.3.2) the candidate solution evaluation function f (x will selected _i(t)) be mapped to the degree of certainty of water dust $\tilde{f}\left(x_i\right(t\left)\right)\∈(0,1),$ Namely

$\tilde{f}\left(x_i\right(t\left)\right)=1-\frac{f\left(x_i\right(t\left)\right)+\hat{f}\left(X\right(t\left)\right)}{\stackrel{\‾}{f}\left(X\right(t\left)\right)}-\mathrm{\ϵ}$

Wherein ε is minimum constant, to guarantee prevent from that logarithm occurs when calling backward cloud generator to overflow.

(3.3.3) will i=1,2 ... k, as input, calls multidimensional backward cloud generator, calculates the cloud numerical characteristic [Ex (t) of disaggregation; En (t); He (t)].Specifically,

$\left[\begin{array}{c}Ex\left(t\right)\\ En\left(t\right)\\ He\left(t\right)\end{array}\right]=\left[\begin{array}{cccc}{\mathrm{Ex}}_1\left(t\right)& {\mathrm{Ex}}_2\left(t\right)& ...& {\mathrm{Ex}}_L\left(t\right)\\ {\mathrm{En}}_1\left(t\right)& {\mathrm{En}}_2\left(t\right)& ...& {\mathrm{En}}_L\left(t\right)\\ {\mathrm{He}}_1\left(t\right)& {\mathrm{He}}_2\left(t\right)& ...& {\mathrm{He}}_L\left(t\right)\end{array}\right].$

Its computation process is as follows:

${\mathrm{Ex}}_j\left(t\right)=\frac{1}{k}\underset{i=1}{\overset{k}{\Σ}}{x}_{ij}\left(t\right)$

${\mathrm{En}}_j\left(t\right)=\sqrt{\frac{1}{k}\underset{i=1}{\overset{k}{\Σ}}{\left(x_{ij}\right(t)-{\mathrm{Ex}}_j(t\left)\right)}^2}$

To every a pair calculate

${\mathrm{En}}_{ij}^{\′}\left(t\right)=\sqrt{\frac{-{\left(x_{ij}\right(t)-{\mathrm{Ex}}_j(t\left)\right)}^2}{2\ln \tilde{f}\left(x_i\right(t\left)\right)}},$

${\mathrm{He}}_j\left(t\right)=\sqrt{\frac{1}{k}\underset{i=1}{\overset{k}{\Σ}}{\left({\mathrm{En}}_{ij}^{\′}\right(t)-\stackrel{\‾}{E}{n}_j^{\′}(t\left)\right)}^2}.$

Wherein $\stackrel{\‾}{E}{n}_j^{\′}\left(t\right)=\frac{1}{k}\underset{i=1}{\overset{k}{\Σ}}\stackrel{\‾}{E}{n}_{ij}^{\′}\left(t\right).$

(3.4) using numerical characteristic as input, call forward cloud model, generate temporary solution group X _temp1and determine its elite's candidate solution x (t+1) _b(t+1);

For a known multidimensional numerical characteristic [Ex (t); En (t); He (t)], t>=0, according to forward cloud model, produces the new candidate solution x with this problem uncertain knowledge to be optimized _i(t+1), i=1,2 ... N.Its process is as follows:

(3.4.1) generating an expectation value is En (t), and standard deviation is the multidimensional normal random number of He (t)

En _i′(t)＝NormRand(En(t),He(t)).

(3.4.2) generating an expectation value is Ex (t), and standard deviation is En _ithe multidimensional normal random number of ' (t)

y _i(t)＝NormRand(Ex(t),En _i′(t)).

(3.4.3) formation disaggregation X is produced _temp1(t+1) a candidate solution x _i, and x (t+1) _i(t+1)=y _i(t);

(3.4.4) x is calculated _i(t+1) evaluation function value f (x _i(t+1));

(3.4.5) repeat (3.4.1) to (3.4.4), till generating N number of candidate solution.

(3.5) if occur, continuous span is for elite, namely meets candidate's disaggregation that random generation is new

(3.6) from disaggregation X (t) and middle selection evaluation function value is greater than the candidate solution of disaggregation average ratings functional value, calls reverse cloud model, sets up new disaggregation characteristic parameter

(3.7) by revised numerical characteristic as input, call forward cloud model, generate new temporary solution group X _temp2(t+1);

(3.8) at temporary solution group X _temp1and X (t+1) _temp2(t+1) evaluation function value is selected to be not less than in candidate solution, form disaggregation X of new generation (t+1);

(3.9) if meet end condition, elite's candidate solution and the conduct of evaluation function value thereof is exported; Otherwise turn to step (3.2).

Referring to Fig. 4, display systems is touched by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, adopt computing intelligence the selection of area-of-interest in video or image and feature extraction organically to be combined, its candidate solution coding structure comprises as lower part:

(4.1) area-of-interest C ₁, C ₂..., C _i... C _m, wherein C _irepresent the selection of i-th area-of-interest, m represents the number of area-of-interest.

Specifically for area-of-interest C _i, its coding can be subdivided into a again _i, x _ci, y _ci, h _ci, w _ci, wherein a _irepresent corresponding C _iwhether be activated, x _ci, y _cirepresent area-of-interest C respectively _ithe horizontal ordinate of central point and ordinate, h _ci, w _cirepresent area-of-interest C respectively _iheight and width.

(4.2) proper vector of area-of-interest is extracted wherein represent the redness in RGB color space model and green; represent the colourity in hsv color spatial model; represent the gray component of area-of-interest.

For above-mentioned arbitrary component, multiple characteristic quantity can be comprised again.First with for example, these characteristic quantities can be the contrasts of color, energy, entropy, consistance, correlativity etc.

Referring to Fig. 5, touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, the process flow diagram of its expression INTELLIGENT IDENTIFICATION is mainly divided into as lower module:

(5.1) intelligent information mining algorithm module.The disaggregation of the intelligent information mining algorithm be made up of multiple candidate solution, each candidate solution all has and is represented, specifically see Fig. 4 by area-of-interest component and characteristic component.The disaggregation renewal process of intelligent information mining algorithm is see Fig. 3.

(5.2) color co-occurrence matrix parameter selecting module.Color co-occurrence matrix parameter comprises pel spacing d and direction θ, if consider, one 8 is communicated with neighborhood, gets intelligent information mining algorithm carries out feature construction by the parameter of adjustment color co-occurrence matrix, and concrete steps are

(5.2.1) color levels of image is defined as L level, color co-occurrence matrix M (i, j) is L × L matrix, the i-th row jth column element m wherein _ijrepresent gray scale to be i and gray scale be two of j at a distance of the pixel for (Δ x, Δ y) to the number of times occurred.To the arbitrary region R in image, definition H is the right set of pixel wherein with particular space contact, and the color co-occurrence matrix of the area-of-interest activated in so each candidate solution can be expressed as

M(i,j)＝#{[(x ₁,y ₁),(x ₂,y ₂)∈H|f(x ₁,y ₁)＝i&f(x ₂,y ₂)＝j]}

$\hat{M}(i,j)=\frac{M(i,j)}{\#\left(H\right)}$

Wherein, i ∈ [0, L-1], j ∈ [0, L-1], x ₂=x ₁+ dcos θ, y ₂=y ₁+ dsinq, # (H) represent the number to the contributive element of M (i, j) in set H.

(5.2.2) usually too much can cause dimension disaster for region R pixel, usually choose multiple characteristic quantities and represent.With the green in RGB color space model for example,

(a) contrast: ${\mathrm{\γ}}_{G1}=\underset{i=0}{\overset{L-1}{\Σ}}\underset{j=0}{\overset{L-1}{\Σ}}{(i-j)}^2\hat{M}(i,j)$

(b) energy: ${\mathrm{\γ}}_{G2}=\underset{i=0}{\overset{L-1}{\Σ}}\underset{j=0}{\overset{L-1}{\Σ}}{\[\hat{M}(i,j)\]}^2$

(c) entropy: ${\mathrm{\γ}}_{G3}=-\underset{i=0}{\overset{L-1}{\Σ}}\underset{j=0}{\overset{L-1}{\Σ}}\hat{M}(i,j)log\hat{M}(i,j),$ Wherein when $\hat{M}(i,j)=0$ Time, $log\hat{M}(i,j)=0$

(d) consistance: ${\mathrm{\γ}}_{G4}=-\underset{i=0}{\overset{L-1}{\Σ}}\underset{i=0}{\overset{L-1}{\Σ}}\hat{M}(i,j)/\[1+{(i-j)}^2\]$

(e) correlativity: ${\mathrm{\γ}}_{G5}=-\underset{i=0}{\overset{L-1}{\Σ}}\underset{j=0}{\overset{L-1}{\Σ}}ij\hat{M}(i,j)-{\mathrm{\μ}}_1{\mathrm{\μ}}_2/{\mathrm{\σ}}_1^2{\mathrm{\σ}}_2^2,$ Wherein $\begin{array}{cc}{\mathrm{\μ}}_1=\underset{i=0}{\overset{L-1}{\Σ}}i\underset{j=0}{\overset{L-1}{\Σ}}\hat{M}(i,j),& {\mathrm{\μ}}_2=\underset{i=0}{\overset{L-1}{\Σ}}j\underset{i=0}{\overset{L-1}{\Σ}}\hat{M}(i,j)\end{array},$ $\begin{array}{cc}{\mathrm{\σ}}_1^2=\underset{i=0}{\overset{L-1}{\Σ}}{(i-{\mathrm{\μ}}_1)}^2\underset{j=0}{\overset{L-1}{\Σ}}\hat{M}(i,j),& {\mathrm{\σ}}_2^2=\underset{j=0}{\overset{L-1}{\Σ}}{(j-{\mathrm{\μ}}_1)}^2\underset{i=0}{\overset{L-1}{\Σ}}\hat{M}(i,j).\end{array}$

(5.3) neighbour's propagation clustering module.This module using all sample datas as initial representative sample, and each sample data is as potential cluster centre (exemplar), Attraction Degree and the degree of membership of each point is constantly updated, until automatically produce several class centers by iterative process.Concrete steps are as follows

(5.3.1) initial similarity matrix S, the i-th row jth column element S (i, j) wherein represents the similarity between sample point i and sample point j, to P assignment.

(5.3.2) the attraction angle value between sample point is calculated

$R(i,k)\←S(i,k)-\underset{j\≠k}{\max }\[S(i,j)+A(i,j)\],$

Wherein A (i, j) represents the ownership angle value of sample point j for sample point i.

(5.3.3) the ownership angle value between sample point is calculated

$A(i,k)\←\min \{0,R(k,k)+\underset{j\≠i,k}{\max }\[0,R(j,k)\]\},$

$A(k,k)\←\underset{j\≠k}{\max }\[0,R(j,k)\]$

(5.3.4) Attraction Degree and ownership angle value is upgraded

$R_{i+1}(i,k)=\mathrm{\ρ}\·R_i(i,k)+(1-\mathrm{\ρ})\·R_{i+1}^{old}(i,k),\mathrm{\ρ}\∈\[0.5,1)$

$A_{i+1}(i,k)=\mathrm{\ρ}\·A_i(i,k)+(1-\mathrm{\ρ})\·A_{i+1}^{old}(i,k),\mathrm{\ρ}\∈\[0.5,1)$

Wherein ρ is convergence coefficient, is mainly used in the stability regulating convergence of algorithm speed and iterative process.

A _i+1(k,k)＝P(k)-max[A _i+1(k,j)+S _i+1(k,j)],j∈{1,2,…,N},j≠k

If (5.3.5) iterations exceedes the maximal value of setting or stops when cluster centre does not change in several times iteration calculating, determine class center and all kinds of sample points; Otherwise return (5.3.2).

Referring to Fig. 6, touch display systems by the holographic interaction of a kind of multi-source information input of the present invention, Intelligent Recognition process, the local of its many group display systems and wide area annexation figure, its primary structure is divided into following two parts.

(6.1) local area control structure.A computing machine can drive one or more touch display device, and each display device can be equipped with one group of sensor senses equipment.Multisensor awareness apparatus can be camera, infrared sensor, microwave remote sensor, ultrasonic sensor, laser sensor, magnetic switch sensor, sound transducer and cigarette sense sensor.

(6.2) wide area syndeton.Multiple stage computing machine can network consisting system, can transmit in a wired or wireless manner, can form bus-type, star-like, ring-like, the network topology structure such as tree-shaped, mixed type.In addition, by various computing machine, external unit and database link up mutually the information handling system of composition, can realize video display file management, application software is shared, show the functions such as schedule, Email and facsimile service service.

The present invention is the input of a kind of multi-source information, the holographic interaction of Intelligent Recognition process touches display systems, touch display device by many sensings input equipment, computing equipment and holographic interaction to form, modular construction, is convenient to client and carries out textural association and software adjustment as required.The feature of its existing billboard, can reach again the effect of film advertisement, and audient can also be made to participate in carrying out effective interaction, is desirable integrated communication media system.It the feature according to audient under complex background environment effectively can push information, according to the mood of audient and can also touch perception information self-adaptative adjustment information.It fully can be shown and lead audient to understand information more comprehensively, increases the acceptance of audient.

Above-described embodiment is only for illustration of the present invention, but not limitation of the invention, all should drop in the scope of the claims in the present invention the conversion of above-described embodiment and modification in spirit of the present invention.

Claims (8)

1. a holographic touch interactive exhibition system for multi-source input, information intelligent optimization process, is characterized in that, comprises the multisensor, computing equipment and the holographic interaction that gather environmental information and touches display device,

(1) open multisensor awareness apparatus monitoring surrounding enviroment, whether detection infrared sensor, sound transducer exist signal; When the equal no signal of infrared sensor, sound transducer, open the generation of cigarette propagated sensation sensor monitoring of environmental fire news feelings; When infrared sensor or sound transducer have signal, start camera and interactive touch display device;

(2) reading images frame or sequence of image frames from video information: Face datection and location, specifically determines whether there is face in input picture, if there is face, then determines position and the size of one or more face; Facial feature extraction, the face detected extracts the feature relevant to sex and age; According to the facial characteristics extracted, estimation is made to age of face in image and sex;

(3) intelligent information mining algorithm and neighbour's propagation clustering algorithm is adopted to carry out INTELLIGENT IDENTIFICATION to human face expression and are classified; The characteristic sum intelligent information mining algorithm extracted by color co-occurrence matrix, is combined into a step by the searching area-of-interest of expression recognition and feature extraction, utilizes neighbour's propagation clustering algorithm to carry out cluster to face;

(4) information customization throw in and information recommend throw in: inquire about face whether mate with the face in image library, have no touch perceptual signal and human face expression whether positive; Do not mate with image library at face, no touch perceptual signal, and the recommendation carrying out information when human face expression is positive is play; When there being touch perception information, stop playing current information, the touch inquire request or the broadcasting audient that respond audient fast specify information; When face mates with image library, do not interrupting preferential its information interested of broadcasting under the current prerequisite just at information broadcasting; When human face expression is not positive, dynamic conditioning shows information classification;

(5) if in the unclosed situation of current presentation information, continue to show current information, otherwise, return step (1).

2. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, the multisensor of described collection environmental information comprises: multiple in camera, infrared sensor, microwave remote sensor, ultrasonic sensor, laser sensor, magnetic switch sensor, sound transducer or cigarette propagated sensation sensor.

3. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, display systems comprises following two parts hardware configuration: one, local area control structure a: computing machine drives one or more touch display device, and each display device is at least equipped with one group of sensor; Two, wide area syndeton: multiple stage computing machine network consisting system, transmits in a wired or wireless manner, forms bus-type, star-like, ring-like, tree-shaped or heterogeneous network topological structure.

4. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, in described step (2), Face datection and location, feature extraction comprises the following steps: first, determine whether there is face in input picture, if there is face, then the one or more face of quick position position in the picture and scope; Secondly, pretreatment image, makes the face location in facial image, size and image normalization, and overcome block, attitude, illumination, jewelry factor impact; Again, the face detected extracts the feature relevant to sex and age; Conventional feature comprises with the Luminance Distribution information global feature that is representative and the local feature that is representative with the position of face and shape profile information; Finally, according to the facial characteristics extracted, the sex of face to be identified is judged and estimates the age.

5. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, in described step (3), the feature that color co-occurrence matrix extracts comprises and represents redness in RGB color space model and green component, the chromatic component in hsv color spatial model and gray component; For above-mentioned arbitrary component, comprise multiple characteristic quantity, these characteristic quantities are one or more in the contrast of color, energy, entropy, consistance or correlativity.

6. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, in described step (3), intelligent information mining algorithm, the disaggregation of intelligent information mining algorithm is made up of multiple candidate solution, each candidate solution represents by area-of-interest component and characteristic component, and its detailed process is as follows:

(61) random initializtion disaggregation;

(62) the evaluation function value of each candidate solution is calculated;

(63) evaluation function value is greater than the candidate solution of average ratings functional value as input, calls reverse cloud model, obtain the numerical characteristic of Multidimensional Cloud Model;

(64) self-adaptative adjustment numerical characteristic;

(65) using revised numerical characteristic as input, call forward cloud model, generate new candidate solution, form disaggregation of new generation;

(66) if meet end condition, elite's candidate solution and evaluation function value thereof is exported; Otherwise turn to step (62).

7. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, it is characterized in that, in described step (3), neighbour's propagation clustering algorithm using all sample datas as initial representative sample, and each sample data is as potential cluster centre, Attraction Degree and the degree of membership of each point is constantly updated, until automatically produce several class centers by iterative process; Concrete steps are as follows:

(71) initial similarity matrix;

(72) the attraction angle value between sample point is calculated;

(73) the ownership angle value between sample point is calculated;

(74) Attraction Degree and ownership angle value is upgraded;

(75) if iterations exceedes the maximal value of setting or stops when cluster centre does not change in several times iteration calculating, class center and all kinds of sample points is determined; Otherwise return (72).

8. the holographic touch interactive exhibition system of multi-source input as claimed in claim 1, information intelligent optimization process, is characterized in that, in described step (4), when information customization is thrown in and information recommends to throw in: one, responds touch signal at any time; Two, do not interrupting under the current prerequisite just at information broadcasting, preferential answering image library matched signal, carrying out information customization according to match information and show; Three, carry out information broadcasting according to human facial expression information; When human face expression is positive, continue the broadcasting of current information, and recommend the recommendation of next round kind attributes information to play; When human face expression is not positive, dynamic conditioning shows information classification.