CN104424476A - Photo classification method and system based on geographic position - Google Patents

Abstract

The invention provides a photo classification method and a photo classification system based on geographic positions, which are suitable for electronic devices. The classification method and the system thereof of the invention utilize the photos containing the two-dimensional geographic position of the shooting place and combine Fuzzy C-Means clustering, FCM to quickly classify a plurality of photos into a plurality of groups and generalize the photos of the same group into the same photo album. The photo classification method and the photo classification system can save the photo classification time and increase the accuracy of photo classification.

Description

Based on photo classification method and the system thereof in geographic position

Technical field

The present invention has about a kind of sorting technique and system thereof, refers to a kind of photo classification method and system thereof especially.

Background technology

Flourish due to image capture technology, a lot of people can utilize mobile phone or digital camera to record the life of individual.Virtually, photo file just can accumulate a thousand sheets.Especially, after tourism, numeral picture is huge especially.

In arrangement or when browsing photograph, because the quantity of photograph is too many, user is not easy to find required photograph.In addition, photograph all sorts according to the time of the year, month, day of shooting usually.Therefore during the photograph of user's thousand sheets on arranging, at most can only with " day " for unit does to classify to photograph fast.If user thinks to do to classify according to the sight spot in photograph further, usually need to use manpower to do to classify to the photograph of a upper thousand sheets, quite expend time in.

Therefore, how according to existing photographic information, photograph to be classified fast and accurately, the inconvenience that traditional user does the photograph of a upper thousand sheets with manpower to classify can be solved, more can save the time of photo classification.

Summary of the invention

The invention provides a kind of photo classification method based on geographic position and system thereof.Photo classification method of the present invention and system thereof are the photograph in the two-dimentional geographic position (as position longitude and latitude) utilized containing photograph shooting ground, simultaneously in conjunction with fuzzy clustering algorithm (Fuzzy C-Means clustering, FCM), being fast multiple group by multiple photo classifications, and the photograph of same group is summarized in same photo album.Photo classification method of the present invention and system thereof is made to be saved the time of photo classification and increase the accuracy of photo classification.

In one of them embodiment of the present invention, the above-mentioned photo classification method based on geographic position is in order to be most individual groups by a majority photo classification.Above-mentioned photo classification method comprises: step (A) receives most photographs and the group number of setting most groups.Each photograph has the geographic position on representative shooting ground, and each group has group's typical value.Geographic position and group's typical value are 2-D data.Step (B) arbitrarily setting group's typical value and each photograph belongs to the ascribed value of the probability of each group, and the probability of the ascribed value of each photograph sum total is 1.Step (C), according to the ascribed value of each photograph and geographic position, revises group's typical value of each group, and according to revised group typical value, revises the ascribed value of each photograph, and produces target ascribed value accordingly.Whether the relative difference of step (D) comparison object ascribed value and previous target ascribed value is less than the first threshold value, and to confirm the correctness of ascribed value, the previous target ascribed value of wherein initial target ascribed value is 0.If not, ascribed value is incorrect, gets back to step (C).Whether if so, ascribed value right value, perform step (E) step (E) and judge in the ascribed value of each photograph, have arbitrary ascribed value to be greater than the second threshold value.If have, do to classify to each photograph according to current group number and ascribed value.If nothing, according to group's typical value of the ascribed value of each photograph and geographic position, each group, produce target group number value.And whether the relative difference of step (F) comparison object group number value and previous target group number value is less than the first threshold value, to confirm the correctness of group number, the previous target group number value of wherein initial target group number value is 0.If so, group number is correct, does to classify to each photograph according to current group number and ascribed value.If not, group number is incorrect, and group number is added 1, and gets back to step (B).

In one of them embodiment of the present invention, the above-mentioned photo classification system based on geographic position is in order to be most individual groups by a majority photo classification.Above-mentioned photo classification system comprises a display unit, a storage element and an operation processing unit.Display unit is the group number setting interface showing most groups, to set group number for user further.Storage element stores most photographs.Each photograph has the geographic position on representative shooting ground.And geographic position is position longitude and latitude when taking photograph.Operation processing unit performs the following step: step (A) receives most photographs and the group number of setting most groups.Each group has group's typical value.And group's typical value is the two-dimensional center position of group.Step (B) arbitrarily setting group's typical value and each photograph belongs to the ascribed value of the probability of each group.And the probability sum total of the ascribed value of each photograph is 1.Step (C), according to the ascribed value of each photograph and geographic position, revises group's typical value of each group, and according to revised group typical value, revises the ascribed value of each photograph, to produce target ascribed value accordingly.Whether the relative difference of step (D) comparison object ascribed value and previous target ascribed value is less than the first threshold value, to confirm the correctness of ascribed value.The previous target ascribed value of wherein initial target ascribed value is 0.If not, ascribed value is incorrect, gets back to step (C).If so, ascribed value is correct, performs step (E).Whether step (E) judges in the ascribed value of each photograph, have arbitrary ascribed value to be greater than the second threshold value.If have, do to classify to each photograph according to current group number and ascribed value.If nothing, according to group's typical value of the ascribed value of each photograph and geographic position, each group, produce target group number value.Whether the relative difference of step (F) comparison object group number value and previous target group number value is less than the first threshold value, to confirm the correctness of group number.The previous target group number value of wherein initial target group number value is 0.If so, group number is correct, does to classify to each photograph according to current group number and ascribed value.If not, group number is incorrect, and group number is added 1, and gets back to step (B).

In order to technology, method and effect that the present invention takes for reaching set object further can be understood, refer to following detailed description for the present invention, graphic, believe object of the present invention, feature and feature, when being goed deep into thus and concrete understanding, but institute's accompanying drawings and annex only provide with reference to and use is described, be not used for the present invention's in addition limitr.

Accompanying drawing explanation

Fig. 1 is the photo classification system schematic of the embodiment of the present invention.

Fig. 2 is the photo classification method flow diagram of the embodiment of the present invention.

Fig. 3 is that the photo classification system of the embodiment of the present invention makes classification schematic diagram to photograph.

Fig. 4 is that the photo classification system of the embodiment of the present invention makes classification schematic diagram to photograph.

Fig. 5 is that the photo classification system of the embodiment of the present invention makes classification schematic diagram to photograph.

Wherein, description of reference numerals is as follows:

110: main frame

112: operation processing unit

116: storage element

120: display unit

130: operating unit

S210, S220, S230, S240, S250, S260, S270, S280, S290: step

G1, G2, G3, G4: group

C1, C2, C3, C4: group's typical value

Pi: photograph

Embodiment

First, please refer to Fig. 1.Fig. 1 is the photo classification system schematic of the embodiment of the present invention.As shown in Figure 1, the photo classification system of the present embodiment is in order to being multiple group by multiple photo classifications.Photo classification system comprises display unit 120, storage element 116 and operation processing unit 112.Display unit 120 shows the group number setting interface of multiple group, to provide user to utilize operating unit 130 sets itself group number, and the group number after setting is sent to operation processing unit 112.In the present embodiment, group number is preset as 2 groups.If user does not set group number, photo classification system starts to classify to photograph by being automatically 2 groups with group number.The operating unit 130 of the present embodiment is mouse, keyboard or other can set the operating unit of group number.Operation processing unit 112 and the storage element 116 of the present embodiment can be arranged in main frame 110.

Storage element 116 stores and stores multiple photographs.Each photograph has the geographic position on representative shooting ground.In the present embodiment, geographic position can be shooting photograph time position longitude and latitude or other representative shooting photograph time position, the present invention is not restricted this.In addition, exchangeable image file (Exchangeable image file format, EXIF) is stored in photograph, to record attribute message and the photographed data of digital photo.Therefore, the position longitude and latitude of the present embodiment can obtain by the exchangeable image file of photograph (Exchangeable image file format, EXIF).Certainly, the geographic position of photograph also can be stored in specific place (e.g., the filename of photograph), and with the geographic position facilitating operation processing unit 112 to obtain photograph, and classify to photograph further, the present invention is not restricted this.

Operation processing unit 112 is electrically connected display unit 120 and storage element 116 and performs the following step, with the photograph in the geographic position of basis containing shooting ground and fuzzy clustering algorithm (FuzzyC-Means clustering, FCM), be multiple group by multiple photo classifications.Please also refer to Fig. 2, first operation processing unit 112 receives multiple photograph and group number.In the present embodiment, if operation processing unit 112 does not receive group number, group number will be preset as 2 groups.Every sheet photo has the two-dimensional position of position longitude and latitude, and is located on two-dimensional coordinate.Each group has the two-dimensional position of a group class value, and each group's typical value is by the group belonging to representative.In the present embodiment, group's typical value is the center of group.Also can be the ad-hoc location (e.g., in group, the center of photograph dense distribution) of each group, as long as can represent affiliated group, the present invention is not restricted this (step S210).

Come, operation processing unit 112 will set arbitrarily the initial value of group's typical value as group's typical value of each group again, and each photograph of setting belongs to an ascribed value of the probability of each group using the initial value as ascribed value arbitrarily.And the probability sum total of each ascribed value of each photograph is 1.If this means, have 50 sheet photos and for being divided into 3 groups, every sheet photo will have 3 ascribed value respectively, to represent that every sheet photo belongs to the probability of which group respectively.And the probability sum total of 3 ascribed value of every sheet photo is 1, represent that every sheet photo is bound to be classified in certain group.In addition, suppose that the 2nd ascribed value of a certain photograph is greater than the 1st and the 3rd ascribed value, this sheet photo will be classified into the 2nd group (step S220).

Next, operation processing unit 112, by according to the ascribed value of each photograph and geographic position, with a group class value correction function, revises group's typical value of each group.Group's typical value correction function is as shown in Equation 1:

$c_j=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ij}}^m\×x_i}{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{u}_{\mathrm{ij}}^m}$ (formula 1)

Wherein, c _jfor group's typical value of jGe group, N is the quantity of photograph, u _ijfor ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, and the preferred values of m is 4.5, is the geography information of i-th photograph.Through the computing of group's typical value correction function, group's typical value more and more will can represent affiliated group.If this means, group typical value c _jbe set as the center of group, through the computing of group's typical value correction function, group typical value c _jthe center of group will be moved closer to.

Come, operation processing unit 112 according to revised group typical value, and with an ascribed value correction function, will revise the ascribed value u of each photograph again _ij.Ascribed value correction function is as shown in Equation 2:

$u_{\mathrm{ij}}=\frac{1}{\underset{k=1}{\overset{C}{\mathrm{\Σ}}}{\left(\frac{\left|\right|x_i-c_j\left|\right|}{\left|\right|x_i-c_k\left|\right|}\right)}^{\frac{2}{m-1}}}$ (formula 2)

Wherein, u _ijfor ascribed value, represent the probability that a jth photograph belongs to i-th group, C is group number, and m is definite value, and the preferred values of m is 4.5, x _ibe the geography information of i-th photograph, c _jfor group's typical value of jGe group, c _kfor group's typical value of kGe group.Through the computing of ascribed value correction function, the ascribed value of each photograph will be corrected gradually.

Afterwards, operation processing unit 112 according to revised group typical value and revised ascribed value, and with a target ascribed value function, will produce a target ascribed value J, to obtain the relation between revised group typical value and ascribed value.Target ascribed value function is as shown in Equation 3:

$J=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{C}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|x_i-c_j\left|\right|}^2,1\&le;m<\&infin;$ (formula 3)

Wherein, N is the quantity of photograph, and C is group number, u _ijfor ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, and the preferred values of m is 4.5, x _ibe the geography information of i-th photograph, c _jfor group's typical value (step S230) of jGe group.

Next, whether the relative difference of comparison object ascribed value and previous target ascribed value is less than the first threshold value by operation processing unit 112, and carries out computing with a difference functions, to judge that whether revised ascribed value is correct.This means, whether each photograph is classified in correct group.Difference functions is as shown in Equation 4:

$\left|\right|\frac{J\left(p\right)-J(p-1)}{J(p-1)}\left|\right|,p\&GreaterEqual;1$ (formula 4)

Wherein, J (p) is the target ascribed value of the p time, and J (p-1) is the target ascribed value of the p-1 time (namely once), and the numerical value of J (0) is 0.Through the computing of difference functions, to confirm revised ascribed value u further _ijcorrectness.In the present embodiment, the first threshold value is set as definite value 0.5 (step S240).When the result of difference functions is more than or equal to 0.5, represent revised ascribed value u _ijincorrect.Now operation processing unit 112 gets back to step S230, and according to revised ascribed value, again revises group's typical value of each group, the ascribed value u of each photograph _ijand target ascribed value J.When the result of difference functions is less than 0.5, represent ascribed value u now _ijcorrectly.Meaning and each photograph have been classified in correct group all.

Whether next, operation processing unit 112 will judge in the ascribed value of each photograph, have arbitrary ascribed value to be greater than the second threshold value.In the present embodiment, the second threshold value is set to 0.9 (step S250).If have, represent that each photograph has converged in a certain group all.Operation processing unit 112 is done to classify (step S260) to each photograph by according to current group number and ascribed value.If nothing, operation processing unit 112 by the group's typical value according to the ascribed value of each photograph and geographic position, each group, and with a target ascribed value function, produces a target group number value, using the use whether restrained as the group number judged now.Target ascribed value function is as shown in Equation 5:

$V_{\mathrm{kwon}}=\frac{\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|c_j-x_i\left|\right|}^2+\frac{1}{c}\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}{\left|\right|c_i-\stackrel{\&OverBar;}{c}\left|\right|}^2}{\underset{i\&NotEqual;j}{\min }{\left|\right|c_i-c_j\left|\right|}^2}$ (formula 5)

Wherein, C is group number, and N is the quantity of photograph, u _ijfor ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, and the preferred values of m is 4.5, c _jfor group's typical value of jGe group, x _ibe the geography information of i-th photograph, c _ibe group's typical value of i-th group, for the mean value (step S270) of group's typical value of each group.

Next, whether the relative difference of comparison object group number value and previous target group number value is less than the first threshold value by operation processing unit 112, and carries out computing with a difference functions, to judge whether group number restrains.Difference functions is as shown in Equation 6:

$\left|\right|\frac{V_{\mathrm{kwon}}\left(q\right)-V_{\mathrm{kwon}}(q-1)}{V_{\mathrm{kwon}}(q-1)}\left|\right|,q\&GreaterEqual;1$ (formula 6)

Wherein, V _kwonq () is the target group number value of the q time, V _kwon(q-1) be the target group number value of the q-1 time (namely once), V _kwon(0) numerical value is 0.Through the computing of difference functions, whether correct to confirm group number now further.In the present embodiment, the first threshold value is set as definite value 0.5 (step S280).When the result of difference functions is more than or equal to 0.5, represent that group number is incorrect.Now operation processing unit 112 will be got back to step S220 and group number will be added 1, do to classify to photograph to reset new group's typical value and new ascribed value.When the result of difference functions is less than 0.5, represent that group's classification is now correct.Now, even if each photograph is not among Complete Convergence to a certain group.Operation processing unit 112 still can be done to classify (step S290) to each photograph according to current group number and ascribed value.

Below will utilize photo classification system of the present invention, and 50 sheet photos and group number will be set to 3 groups explain, and please also refer to the process flow diagram of Fig. 2.For convenience of description, in the two-dimensional coordinate of the present embodiment, X-axis represents position longitude, and Y-axis represents position latitude.Photograph is done to represent with " X " mark.Group does to represent with " closing dashed region ".Group's typical value of group is then done to represent with " ▲ " mark.

As shown in Figure 3, first, 50 sheet photos all have position longitude and latitude, and are located on two-dimensional coordinate.And the group typical value C1-C3 of group G1-G3 is for be arranged on two-dimensional coordinate arbitrarily, be used for representing the center of group G1-G3 respectively.The group typical value C1-C3 now arranged arbitrarily is using the original group typical value as group G1-G3.

Next, photo classification system belongs to the probability of 3 group G1-G3 respectively by setting arbitrarily 50 sheet photos, and to form 3 probability sum total be that the ascribed value of 1 is as initial home value.For example, if in 3 ascribed value of the 10th sheet photo, the 2nd ascribed value is maximum, and the 10th sheet photo will be classified into the 2nd group.

Come, photo classification system, by the computing of through mode 1, makes group typical value C1-C3 shift to the center of respective group G1-G3 gradually again.Yi Ji group typical value C1-C3 converges to the correct center of group G1-G3 gradually.And photo classification system is by the computing of through mode 2, make 50 sheet photos can revise the ascribed value of each photograph according to amended group typical value C1-C3.Meaning i.e. 50 sheet photos are classified into correct group G1-G3 gradually.As shown in Figure 3 and 4, group typical value C1-C3 moves to the center close to group G1-G3 gradually.And photograph Pi becomes group G3 from group G1.

Next, photo classification system, by the computing of through mode 3 and formula 4, judges the correctness of each ascribed value of now 50 sheet photos.The result of the difference functions of hypothetical target ascribed value is greater than 0.5, represents in 50 sheet photos and has ascribed value incorrect.Now photo classification system is by according to current ascribed value, again revises group's typical value of each group, the ascribed value of each photograph and target ascribed value J.And in the present embodiment, the result of the difference functions of target ascribed value is less than 0.5, represent that each ascribed value of 50 sheet photos has restrained complete.

Come, whether photo classification system in each ascribed value of 50 sheet photos, has any one ascribed value to be greater than 0.9 by judging further again.Suppose in each ascribed value of 50 sheet photos, all have an ascribed value to be greater than 0.9, represent that 50 sheet photos have been categorized among correct group all.As shown in Figure 4, photo classification system, by according to current group number and ascribed value, is divided into 3 group G1-G3 to 50 sheet photos, completes the classification of 50 sheet photos.

But in the present embodiment, in each ascribed value of 50 sheet photos, be greater than 0.9 without any an ascribed value, represent that 50 sheet photos are not restrained and are categorized among a certain group.Now, photo classification system, by the computing of through mode 5 and formula 6, judges whether group number now restrains further.In the present embodiment, the result of the difference functions of target group number is less than 0.5, represents that group's classification is now correct.Photo classification system according to current group number and ascribed value, will be divided into 3 group G1-G3 to 50 sheet photos, completes the classification of 50 sheet photos, as shown in Figure 4.Certainly, the result of the difference functions of hypothetical target group number is not less than 0.5, represents that group number is now incorrect.Now, group number is added 1 by photo classification system, and resets new group's typical value and the new ascribed value of 50 sheet photos, again to do to classify to 50 sheet photos.As shown in Figure 5,50 sheet photos are seated on same two-dimensional coordinate.Group number changes 4 groups into by 3 groups.And the group typical value C1-C4 of group G1-G4 is for be arranged on two-dimensional coordinate arbitrarily, be used for representing the initial center position of group G1-G4 respectively.Each ascribed value of 50 sheet photos also will reset.Then, photo classification system, again according to formula 1-formula 6, is done to classify to 50 sheet photos again.

In sum, the photo classification method based on geographic position that the embodiment of the present invention provides and system thereof, utilize the photograph in the two-dimentional geographic position containing photograph shooting ground, simultaneously in conjunction with FCM algorithm, and judge in FCM algorithm ascribed value and group number whether suitable, being fast multiple group by photo classification.Photo classification method of the present invention and system thereof is made to be saved the time of photo classification and increase the accuracy of photo classification.

The foregoing is only embodiments of the invention, it is also not used to limit to the scope of the claims of the present invention.

Claims (19)

1., based on the photo classification method in geographic position, in order to be most groups by a majority photo classification, to it is characterized in that, comprise the steps:

(A) receive this majority photograph and set a group number of this majority group, each photograph has this geographic position on representative shooting ground, and each group has a group class value, and this geographic position and this group's typical value are 2-D data;

(B) this group's typical value of setting and each photograph belong to an ascribed value of the probability of each group arbitrarily, and the probability of respectively this ascribed value of each photograph sum total is 1;

(C) according to this ascribed value and this geographic position of each photograph, revise this group's typical value of each group, and according to this group's typical value revised, revise this ascribed value of each photograph, to produce a target ascribed value accordingly;

(D) whether the relative difference comparing this target ascribed value and this target ascribed value previous is less than one first threshold value, this target ascribed value previous of this wherein initial target ascribed value is 0, if not, get back to step (C), if so, step (E) is then performed;

(E) in respectively this ascribed value of each photograph, judge whether that this ascribed value is greater than the second threshold value, if have, do to classify to each photograph according to this current group number and this ascribed value, if nothing, according to this group's typical value of this ascribed value of each photograph and this geographic position, each group, produce a target group number value; And

(F) whether the relative difference comparing this target group number value and this target group number value previous is less than this first threshold value, this target group number value previous of this wherein initial target group number value is 0, if, do to classify to each photograph according to this current group number and this ascribed value, if not, this group number adds 1, and gets back to step (B).

2. photo classification method according to claim 1, wherein, position longitude and latitude when this geographic position is this photograph of shooting, this group's typical value is the center of this group.

3. photo classification method according to claim 1, wherein, in this step (C), also comprises a group class value correction function: $c_j=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m\&times;x_i}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m}$

In order to revise this group's typical value of each group, wherein, c _jfor this group's typical value of jGe group, N is the quantity of this photograph, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, x _iit is this geography information of i-th photograph.

4. photo classification method according to claim 1, wherein, in this step (C), also comprises an ascribed value correction function:

$u_{\mathrm{ij}}=\frac{1}{\underset{k=1}{\overset{C}{\mathrm{\&Sigma;}}}{\left(\frac{\left|\right|x_i-c_j\left|\right|}{\left|\right|x_i-c_k\left|\right|}\right)}^{\frac{2}{m-1}}}$

In order to revise this ascribed value of each photograph, wherein, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, C is this group number, and m is definite value, x _ibe this geography information of i-th photograph, c _jfor this group's typical value of jGe group, c _kfor this group's typical value of kGe group.

5. photo classification method according to claim 1, wherein, in this step (C), also comprises a target ascribed value function:

$J=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{C}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|x_i-c_j\left|\right|}^2,1\&le;m<\&infin;$

In order to produce this target ascribed value, wherein, N is the quantity of this photograph, and C is this group number, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, x _ibe this geography information of i-th photograph, c _jfor this group's typical value of jGe group.

6. photo classification method according to claim 5, wherein, in this step (D), also comprises a difference functions of this target ascribed value:

$\left|\right|\frac{J\left(p\right)-J(p-1)}{J(p-1)}\left|\right|,p\&GreaterEqual;1$

In order to confirm the correctness of this ascribed value, wherein, J (p) is this target ascribed value of the p time, and J (p-1) is this target ascribed value of the p-1 time, and the numerical value of J (0) is 0.

7. photo classification method according to claim 1, wherein, in this step (E), also comprises a target group number value function:

$V_{\mathrm{kwon}}=\frac{\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|c_j-x_i\left|\right|}^2+\frac{1}{c}\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}{\left|\right|c_i-\stackrel{\&OverBar;}{c}\left|\right|}^2}{\underset{i\&NotEqual;j}{\min }{\left|\right|c_i-c_j\left|\right|}^2}$

In order to produce this target group number value, wherein, C is this group number, and N is the quantity of this photograph, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, c _jfor this group's typical value of jGe group, x _ibe this geography information of i-th photograph, c _ibe this group's typical value of i-th group, for the mean value of this group's typical value of each group.

8. photo classification method according to claim 7, wherein, in this step (F), also comprises a difference functions of this target group number value:

$\left|\right|\frac{V_{\mathrm{kwon}}\left(q\right)-V_{\mathrm{kwon}}(q-1)}{V_{\mathrm{kwon}}(q-1)}\left|\right|,q\&GreaterEqual;1$

In order to confirm the correctness of this group number, wherein, V _kwonq () is this target group number value of the q time, V _kwon(q-1) be this target group number value of the q-1 time, V _kwon(0) numerical value is 0.

9. according to the photo classification method of claim 6 or 8, wherein, this first threshold value is 0.5.

10. photo classification method according to claim 1, wherein, this second threshold value is 0.9.

11. 1 kinds of photo classification systems based on geographic position, in order to be most groups by a majority photo classification, to is characterized in that, comprising:

One display unit, in order to show a group number setting interface of this majority group;

One storage element, in order to store this majority photograph, each photograph has this geographic position on representative shooting ground, position longitude and latitude when this geographic position is this photograph of shooting;

One operation processing unit, in order to perform the following step:

(A) receive this majority photograph and set a group number of this majority group, each group has a group class value, and this group's typical value is the two-dimensional center position of this group;

(B) this group's typical value of setting and each photograph belong to an ascribed value of the probability of each group arbitrarily, and the probability of respectively this ascribed value of each photograph sum total is 1;

(C) according to this ascribed value and this geographic position of each photograph, revise this group's typical value of each group, and according to this group's typical value revised, revise this ascribed value of each photograph, to produce a target ascribed value accordingly;

(D) whether the relative difference comparing this target ascribed value and this target ascribed value previous is less than one first threshold value, this target ascribed value previous of this wherein initial target ascribed value is 0, if not, get back to step (C), if so, step (E) is then performed;

(E) in respectively this ascribed value of each photograph, judge whether that this ascribed value is greater than the second threshold value, if have, do to classify to each photograph according to this current group number and this ascribed value, if nothing, according to this group's typical value of this ascribed value of each photograph and this geographic position, each group, produce a target group number value; And

(F) whether the relative difference comparing this target group number value and this target group number value previous is less than this first threshold value, this target group number value previous of this wherein initial target group number value is 0, if, do to classify to each photograph according to this current group number and this ascribed value, if not, this group number adds 1, and gets back to step (B).

12. photo classification systems according to claim 11, wherein, in this step (C), also comprise a group class value correction function:

$c_j=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m\&times;x_j}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m}$

In order to revise this group's typical value of each group, wherein, c _jfor this group's typical value of jGe group, N is the quantity of this photograph, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, x _iit is this geography information of i-th photograph.

13. photo classification systems according to claim 11, wherein, in this step (C), also comprise an ascribed value correction function:

$u_{\mathrm{ij}}=\frac{1}{\underset{k=1}{\overset{C}{\mathrm{\&Sigma;}}}{\left(\frac{\left|\right|x_i-c_j\left|\right|}{\left|\right|x_i-c_k\left|\right|}\right)}^{\frac{2}{m-1}}}$

In order to revise this ascribed value of each photograph, wherein, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, C is this group number, and m is definite value, x _ibe this geography information of i-th photograph, c _jfor this group's typical value of jGe group, c _kfor this group's typical value of kGe group.

14. photo classification systems according to claim 11, wherein, in this step (C), also comprise a target ascribed value function:

$J=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{C}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|x_i-c_j\left|\right|}^2,1\&le;m<\&infin;$

In order to produce this target ascribed value, wherein, N is the quantity of this photograph, and C is this group number, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, x _ibe this geography information of i-th photograph, c _jfor this group's typical value of jGe group.

15. photo classification systems according to claim 14, wherein, in this step (D), also comprise a difference functions of this target ascribed value:

$\left|\right|\frac{J\left(p\right)-J(p-1)}{J(p-1)}\left|\right|,p\&GreaterEqual;1$

In order to confirm the correctness of this ascribed value, wherein, J (p) is this target ascribed value of the p time, and J (p-1) is this target ascribed value of the p-1 time, and the numerical value of J (0) is 0.

16. photo classification systems according to claim 11, wherein, in this step (E), also comprise a target group number value function:

$V_{\mathrm{kwon}}=\frac{\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{\left|\right|c_j-x_i\left|\right|}^2+\frac{1}{c}\underset{i=1}{\overset{C}{\mathrm{\&Sigma;}}}{\left|\right|c_i-\stackrel{\&OverBar;}{c}\left|\right|}^2}{\underset{i\&NotEqual;j}{\min }{\left|\right|c_i-c_j\left|\right|}^2}$

In order to produce this target group number value, wherein, C is this group number, and N is the quantity of this photograph, u _ijfor this ascribed value, represent the probability that a jth photograph belongs to i-th group, m is definite value, c _jfor this group's typical value of jGe group, x _ibe this geography information of i-th photograph, c _ibe this group's typical value of i-th group, for the mean value of this group's typical value of each group.

17. photo classification systems according to claim 16, wherein, in this step (F), also comprise a difference functions of this target group number value:

$\left|\right|\frac{V_{\mathrm{kwon}}\left(q\right)-V_{\mathrm{kwon}}(q-1)}{V_{\mathrm{kwon}}(q-1)}\left|\right|,q\&GreaterEqual;1$

In order to confirm the correctness of this group number, wherein, V _kwonq () is this target group number value of the q time, V _kwon(q-1) be this target group number value of the q-1 time, V _kwon(0) numerical value is 0.

18. according to the photo classification system of claim 15 or 17, and wherein, this first threshold value is 0.5.

19. photo classification systems according to claim 11, wherein, this second threshold value is 0.9.