TWI494780B - Method and system for sorting photos base on geographic position, and computer readable recording media - Google Patents

Description

Geographically based photo classification method and system thereof, and computer readable record media

The invention relates to a classification method and a system thereof, in particular to a photo classification method and a system thereof.

Due to the booming image capture technology, many people use their mobile phones or digital cameras to record their personal lives. Invisible, thousands of photos will be accumulated. Especially after the tour, the number of photos is even larger.

When organizing or browsing photos, it is not easy for users to find the photos they need because of the large number of photos. In addition, photos are usually sorted according to the year, month, and day of the filming. Therefore, when users organize thousands of photos, they can only quickly sort photos by "day". If the user wants to further classify the spots in the photo, it is usually time consuming to use thousands of photos to sort the photos.

Therefore, how to quickly and accurately classify photos based on existing photo information will solve the inconvenience of traditional users to manually sort thousands of photos, and save time in photo classification.

The invention provides a geographic location based photo classification method and system thereof The system and the computer can read the recording medium. The photo classification method and system thereof of the present invention utilizes a photo containing a two-dimensional geographic location (such as position latitude and longitude) of a photographing place, and a fuzzy clustering algorithm (Fuzzy C-Means clustering, FCM) to quickly take multiple sheets. The photos are grouped into groups and the photos from the same group are grouped into the same album. The photo classification method and system thereof of the present invention can save time for photo classification and increase the accuracy of photo classification.

In one embodiment of the present invention, the above-described geographic location-based photo classification method is used to classify a plurality of photos into a plurality of groups. The above photo classification method comprises the steps of: (A) receiving a plurality of photos and setting a group number of the plurality of groups. Each photo has a geographic location representing the location of the location, and each group has a group representative value. The geographic location and the group representative value are two-dimensional data. Step (B) arbitrarily sets the group representative value and the attribution value of the probability that each photo belongs to each group, and the probability of the attribution value of each photo is 1. Step (C) correcting the group representative value of each group according to the attribution value and the geographical position of each photo, and correcting the attribution value of each photo according to the corrected group representative value, and generating the target attribution according to the step value. Step (D) compares whether the difference between the target attribution value and the previous target attribution value is less than the first threshold to confirm the correctness of the attribution value. If no, the attribution value is incorrect and return to step (C). If yes, the attribution value is correct, and step (E) is performed. Step (E) determines whether any of the attribution values of each photo is greater than the second threshold. If so, each photo is sorted based on the current number of groups and the attribution value. If not, the target group number value is generated according to the attribution value of each photo and the geographical location and the group representative value of each group. And step (F) comparing whether the difference between the target group number value and the previous target group number value is less than the first threshold value to confirm the correctness of the group number. If so, the number of groups is correct, and each photo is classified according to the current number of groups and the attribution value. If not, the number of groups is incorrect, increase the number of groups by one, and return to step (B).

In one embodiment of the present invention, the geographic location based photo classification system is configured to classify a plurality of photos into a plurality of groups. Photo classification system The invention comprises a display unit, a storage unit and an operation processing unit. The display unit is configured to display a group number setting interface of the plurality of groups to further set the number of groups for the user. The storage unit is for storing a plurality of photos. Each photo has a geographical location that represents the location of the shot. The geographic location is the location latitude and longitude of the photo. The operation processing unit is configured to perform the following steps: step (A) receiving a plurality of photos and setting a group number of the plurality of groups. Each group has a group representative value. The group representative value is the two-dimensional center position of the group. Step (B) arbitrarily sets the group representative value and the attribution value of the probability that each photo belongs to each group. The probability of attribution of each photo is 1 in total. Step (C) correcting the group representative value of each group according to the attribution value and the geographical location of each photo, and correcting the attribution value of each photo according to the corrected group representative value, thereby generating the target attribution according to the step value. Step (D) compares whether the difference between the target attribution value and the previous target attribution value is less than the first threshold to confirm the correctness of the attribution value. If no, the attribution value is incorrect and return to step (C). If yes, the attribution value is correct, and step (E) is performed. Step (E) determines whether any of the attribution values of each photo is greater than the second threshold. If so, each photo is sorted based on the current number of groups and the attribution value. If not, the target group number value is generated according to the attribution value of each photo and the geographical location and the group representative value of each group. Step (F) compares whether the difference between the target group number value and the previous target group number value is less than the first threshold value to confirm the correctness of the group number. If so, the number of groups is correct, and each photo is classified according to the current number of groups and the attribution value. If not, the number of groups is incorrect, increase the number of groups by one, and return to step (B).

In addition, an embodiment of the present invention further provides a computer readable recording medium for recording a set of computer executable programs. When the computer readable recording medium is read by the processor, the processor may perform the steps in the photo classification method.

In order to further understand the technology, method and effect of the present invention in order to achieve the intended purpose, reference should be made to the detailed description and drawings of the present invention. It’s The drawings and the annexes are to be considered as illustrative and not restrictive.

110‧‧‧Computer host

112‧‧‧Operation Processing Unit

116‧‧‧ storage unit

120‧‧‧Display unit

130‧‧‧Operating unit

S210, S220, S230, S240, S250, S260, S270, S280, S290‧‧ steps

G1, G2, G3, G4‧‧‧ groups

C1, C2, C3, C4‧‧‧ group representative values

Pi‧‧‧ Photos

1 is a schematic diagram of a photo classification system according to an embodiment of the present invention.

2 is a flow chart of a photo classification method according to an embodiment of the present invention.

FIG. 3 is a schematic diagram showing the classification of photos by the photo classification system of the embodiment of the present invention.

4 is a schematic diagram showing the classification of photos by the photo classification system of the embodiment of the present invention.

FIG. 5 is a schematic diagram showing the classification of photos by the photo classification system of the embodiment of the present invention.

First, please refer to Figure 1. 1 is a schematic diagram of a photo classification system according to an embodiment of the present invention. As shown in FIG. 1, the photo classification system of this embodiment is used to classify a plurality of photos into a plurality of groups. The photo classification system includes a display unit 120, a storage unit 116, and an operation processing unit 112. The display unit 120 displays a group number setting interface of a plurality of groups to provide a user to set the number of groups by using the operation unit 130, and transmits the set number of groups to the operation processing unit 112. In this embodiment, the number of groups is preset to be 2 groups. If the user does not set the number of groups, the photo classification system will automatically classify the photos by the number of groups of 2 groups. The operation unit 130 of this embodiment is a mouse, a keyboard or other operation unit capable of setting a group number. The arithmetic processing unit 112 and the storage unit 116 of the embodiment may be disposed in the computer host 110.

The storage unit 116 stores a plurality of photos stored. Each photo has a geographical location that represents the location of the shot. In this embodiment, the geographic location may be the position latitude and longitude at the time of taking the photo or other position at the time of taking the photo, and the present invention is not limited thereto. In addition, an exchangeable image file (EXIF) is stored in the photo to record attribute information and shooting data of the digital photo. Therefore, the position latitude and longitude of the embodiment can be exchangeable image files of photos Obtained in (Exchangeableimagefile format, EXIF). Of course, the geographic location of the photo can also be stored in a specific place (for example, the file name of the photo) to facilitate the operation processing unit 112 to obtain the geographical position of the photo, and further classify the photo, which is not limited by the present invention.

The operation processing unit 112 electrically connects the display unit 120 and the storage unit 116 and performs the following steps to classify a plurality of photos according to a photograph containing a geographical location of the photographing place and a Fuzzy C-Means clustering (FCM). For multiple groups. Referring to FIG. 2 at the same time, first, the operation processing unit 112 receives a plurality of photos and the number of groups. In this embodiment, if the operation processing unit 112 does not receive the number of groups, the number of groups will be preset to 2 groups. Each photo has a two-dimensional position of latitude and longitude and is located on a two-dimensional coordinate. Each group has a two-dimensional position of a group representative value, and each group representative value will represent the group to which it belongs. In this embodiment, the group representative value is the center position of the group. It may also be a specific location of each group (eg, a central location where the photos are densely distributed in the group), as long as it can represent the group to which it belongs, and the present invention does not limit this (step S210).

Then, the operation processing unit 112 arbitrarily sets the group representative value of each group as the initial value of the group representative value, and arbitrarily sets one of the probability values of each photo belonging to each group as the attribution value. Initial value. The probability of each attribution value of each photo is 1 in total. That is, if there are 50 photos and want to be divided into 3 groups, each photo will have 3 attribution values respectively to indicate the probability of each group to which each photo belongs. The probability of each of the three affiliation values of each photo is 1, which means that each photo must be classified into a group. Further, assuming that the second attribution value of a photo is greater than the first and third attribution values, the photo is sorted into the second group (step S220).

Next, the operation processing unit 112 corrects the group representative value c _{j of} each group by a group representative value correction function according to the attribution value and the geographical position of each photo. The group representative value correction function is as shown in Equation 1:

Where c _j is the group representative value of the jth group, N is the number of photos, u _ij is the attribution value, and represents the probability that the jth photo belongs to the i-th group, m is a fixed value of 4.5, x _i Geographic information for the ith photo. Through the operation of the group representative value correction function, the group representative value c _j will more and more represent the group to which it belongs. That is, if the group representative value c _{j is} set to the center position of the group, the group representative value c _j will gradually approach the center position of the group by the operation of the group representative value correction function.

Then, the operation processing unit 112 corrects the attribution value u _{ij of} each photo based on the corrected group representative value and a attribution value correction function. The attribution value correction function is as shown in Equation 2:

Where u _ij is the attribution value, representing the probability that the jth photo belongs to the i-th group, C is the number of groups, m is a fixed value of 4.5, x _i is the geographic information of the i-th photo, c _j is the j-th The group representative values of the groups, and c _k is the group representative value of the kth group. Through the operation of the attribution value correction function, the attribution value of each photo will be gradually corrected.

Thereafter, the operation processing unit 112 generates a target attribution value J according to the corrected group representative value c _j and the corrected attribution value u _ij and a target attribution value function to obtain the corrected group representative value. The relationship between c _j and the attribution value u _ij . The target attribution value function is shown in Equation 3:

Where N is the number of photos, C is the number of groups, u _ij is the attribution value, which represents the probability that the jth photo belongs to the i-th group, m is a fixed value of 4.5, and x _i is the geographic information of the i-th photo , c _j is a group representative value of the jth group (step S230).

Next, the operation processing unit 112 compares whether the difference between the comparison target attribution value and the previous target attribution value is less than the first threshold value, and performs a difference function to determine whether the corrected attribution value u _ij is correct. This means that each photo is sorted into the correct group. The difference function is shown in Equation 4:

Where J ( p ) is the target assignment value of the pth time, and J ( p -1) is the target attribution value of the p-1th (that is, the previous time). The correctness of the corrected attribution value u _ij is further confirmed by the operation of the difference function. In the present embodiment, the first threshold value is set to a fixed value of 0.5 (step S240). When the result of the difference function is greater than or equal to 0.5, it indicates that the corrected attribution value u _ij is incorrect. At this point arithmetic processing unit 112 back to the step S230, the home according to the corrected value, the correction re-groups of each group representative values c _j, a photo of each home and a target value u _ij home value J. When the result of the difference function is less than 0.5, it means that the attribution value u _ij at this time is correct. This means that each photo has been sorted into the correct group.

Next, the operation processing unit 112 determines whether any of the attribution values of each photo is greater than the second threshold value. In the present embodiment, the second threshold value is set to 0.9 (step S250). If so, it means that each photo has converged into a certain group. The operation processing unit 112 classifies each photo based on the current number of groups and the attribution value (step S260). If not, the operation processing unit 112 generates a target group number value V _kwon according to the attribution value of each photo and the geographical location, the group representative value of each group, and a target attribution value function. Whether the number of groups at this time has been converged. The target attribution value function is as shown in Equation 5:

Where C is the number of groups, N is the number of photos, u _ij is the attribution value, which represents the probability that the jth photo belongs to the i-th group, m is a fixed value of 4.5, and c _j is the group of the j-th group Group representative value, x _i is the geographic information of the i-th photo, c _i is the group representative value of the i-th group, The average of the values representative of the group for each group (step S270).

Next, the operation processing unit 112 compares whether the difference between the comparison target group number value and the previous target group number value is less than the first threshold value, and performs a difference function to determine whether the group number has converged. . The difference function is shown in Equation 6:

Where V _kwon (q) is the qth target group number value, and V _kwon (q-1) is the q- _1th (ie previous time) target group number value. The operation of the difference function is used to further confirm whether the number of groups at this time is correct. In the present embodiment, the first threshold value is set to a fixed value of 0.5 (step S280). When the result of the difference function is greater than or equal to 0.5, the number of groups is incorrect. At this time, the operation processing unit 112 will return to step S220 and increment the number of groups to re-set the new group representative value and the new attribution value to classify the photos. When the result of the difference function is less than 0.5, it means that the group classification at this time is correct. At this point, even if each photo does not completely converge into a certain group. The operation processing unit 112 still classifies each photo based on the current number of groups and the attribution value (step S290).

Hereinafter, the photo classification system of the present invention will be used to describe 50 photos and the number of groups is set to 3 groups, and please refer to the flowchart of FIG. 2 at the same time. For convenience of explanation, in the two-dimensional coordinate of the embodiment, the X axis represents the position longitude, and the Y axis represents the position latitude. The photo is indicated by the "X" mark. Groups are represented by "closed dotted area". The group representative value of the group is represented by the "▲" mark.

As shown in FIG. 3, first, all 50 photos have position latitude and longitude and are located on a two-dimensional coordinate. The group representative values C1-C3 of the groups G1-G3 are arbitrarily set on the two-dimensional coordinates, and are respectively used to represent the center positions of the groups G1-G3. The group representative values C1-C3 arbitrarily set at this time will be represented as the initial group of the groups G1-G3. value.

Next, the photo classification system will arbitrarily set the probability that the 50 photos belong to the three groups G1-G3, and form three attribution values whose probability is 1 as the initial attribution value. For example, if the second attribution value is the largest among the three attribution values of the tenth photo, the tenth photo will be classified into the second group.

Then, the photo classification system will perform the operation of Equation 1 so that the group representative values C1-C3 gradually move to the center positions of the respective groups G1-G3. That is, the group representative values C1-C3 gradually converge to the correct center position of the group G1-G3. The photo classification system will perform the operation of Equation 2 so that 50 photos can correct the attribution value of each photo according to the modified group representative values C1-C3. This means that 50 photos are gradually sorted into the correct group G1-G3. As shown in FIGS. 3 and 4, the group representative values C1-C3 are gradually moved to the center positions close to the groups G1-G3. The photo Pi changes from group G1 to group G3.

Next, the photo classification system will judge the correctness of each of the 50 photos at this time through the operations of Equations 3 and 4. Assume that the result of the difference function of the target attribution value is greater than 0.5, indicating that the attribution value is incorrect in the 50 photos. At this time, the photo classification system will re-correct the group representative value c _j of each group, the attribution value u _{ij of} each photo, and the target attribution value J according to the current attribution value. In the present embodiment, the result of the difference function of the target attribution value is less than 0.5, indicating that the respective attribution values of the 50 photos have converged.

Then, the photo classification system will further determine whether any of the collocation values of the 50 photos are greater than 0.9. Assuming that each of the 50 photos has a attribution value greater than 0.9, it means that 50 photos have been classified into the correct group. As shown in FIG. 4, the photo classification system divides 50 photos into 3 groups G1-G3 according to the current number of groups and the attribution value, and completes the classification of 50 photos.

However, in this embodiment, none of the attribution values of the 50 photos is greater than 0.9, indicating that 50 photos are not converged and classified into a certain group. Among them. At this time, the photo classification system will further judge whether the number of groups at this time has converged through the calculations of Equations 5 and 6. In this embodiment, the result of the difference function of the number of target groups is less than 0.5, indicating that the group classification at this time is correct. The photo classification system will divide 50 photos into 3 groups G1-G3 according to the current number of groups and the attribution value, and complete the classification of 50 photos, as shown in FIG. Of course, the result of the difference function of the number of target groups is not less than 0.5, indicating that the number of groups at this time is incorrect. At this time, the photo classification system adds 1 to the number of groups, and resets the new group representative value and the new attribution value of 50 photos to re-categorize 50 photos. As shown in Figure 5, 50 photos are located on the same two-dimensional coordinates. The number of groups was changed from 3 groups to 4 groups. The group representative values C1-C4 of the groups G1-G4 are arbitrarily set on the two-dimensional coordinates, and are respectively used to represent the initial center positions of the groups G1-G4. The respective attribution values of the 50 photos will also be reset. Next, the photo classification system re-classifies 50 photos according to Equation 1 - Equation 6.

In addition, the present invention can also utilize a computer readable recording medium to store a computer program of the aforementioned line layout method to perform the aforementioned steps. The computer readable medium can be a floppy disk, a hard disk, a compact disk, a flash drive, a magnetic tape, a database accessible by the network, or a storage medium that can be easily thought of by the person skilled in the art.

In summary, the geographic location-based photo classification method and system thereof according to the embodiments of the present invention utilize a photo of a two-dimensional geographic location containing a photo shooting location, and combine the FCM algorithm and determine the attribution in the FCM algorithm. Values and the number of groups are appropriate to quickly categorize photos into groups. The photo classification method and system thereof of the present invention can save time for photo classification and increase the accuracy of photo classification.

The above description is only an embodiment of the present invention, and is not intended to limit the scope of the invention.

110‧‧‧Computer host

112‧‧‧Operation Processing Unit

116‧‧‧ storage unit

120‧‧‧Display unit

130‧‧‧Operating unit

Claims (20)

A location-based photo classification method for classifying a plurality of photos into a plurality of groups, comprising the steps of: (A) receiving the plurality of photos and setting a number of groups of the plurality of groups, each photo Having the geographic location representing the location of the location, and each group having a representative value of the group, the geographic location and the representative value of the group being two-dimensional data; (B) arbitrarily setting the representative value of the group and each photo belongs to One of the probability of each group is a vesting value, and the probability of each of the affiliation values of each photo is 1; (C) correcting each group according to the affiliation value of each photo and the geographic location The group represents a value, and according to the corrected representative value of the group, corrects the attribution value of each photo to generate a target attribution value; (D) comparing the target attribution value with the previous target attribution value Whether the difference is less than a first threshold, if not, returning to step (C), and if so, performing step (E); (E) determining whether the attribution value is present in each of the attribution values of each photo Greater than the second threshold, if any, according to the current group The number of groups and the attribution value are used to classify each photo. If not, a target group number value is generated according to the attribution value of each photo and the geographic location, the group representative value of each group; and F) comparing whether the difference between the target group number value and the previous target group number value is less than the first threshold value, and if so, classifying each photo according to the current number of the group and the attribution value, If not, the number of the group is incremented by one and returns to step (B). The photo classification method of claim 1, wherein the geographic location is a location latitude and longitude when the photo is taken, and the group representative value is a central location of the group. The photo classification method of item 1 of the claim item further includes a group representative value correction function in the step (C): For correcting the group representative value of each group, wherein c _j is the group representative value of the jth group, N is the number of the photos, u _{ij is} the attribution value, and the jth photo belongs to The probability of the i-th group, m is a fixed value of 4.5, and x _i is the geographic information of the i-th photo. The photo classification method of item 1 of the claim item further includes a attribution value correction function in the step (C): For correcting the attribution value of each photo, where u _{ij is} the attribution value, representing the probability that the jth photo belongs to the i-th group, C is the number of the group, m is a fixed value of 4.5, and x _i is The geographic information of the i-th photo, c _j is the representative value of the group of the j-th group, and c _k is the representative value of the group of the k-th group. The photo classification method of item 1 of the claim item further includes a target attribution value function in the step (C): For generating the target attribution value, where N is the number of the photos, C is the number of the group, u _{ij is} the attribution value, and represents the probability that the jth photo belongs to the i-th group, and m is a fixed value 4.5, x _i is the geographic information of the i-th photo, and c _j is the representative value of the group of the j-th group. The photo classification method of item 5 of the claim, wherein in the step (D), the difference function of the target attribution value is further included: For confirming the correctness of the attribution value, where J ( p ) is the target assignment value of the pth time, and J ( p -1) is the target attribution value of the p-1th time. The photo classification method of item 1 of the claim item further includes a target group number value function in the step (E): For generating the target group number value, where C is the number of the group, N is the number of the photos, u _{ij is} the attribution value, and represents the probability that the jth photo belongs to the i-th group, m is The value 4.5, c _j is the representative value of the group of the jth group, x _i is the geographic information of the i-th photo, and c _i is the representative value of the group of the i-th group. The average of the values represented by the group for each group. The photo classification method of item 7 of the claim item further includes a difference function of the target group number value in the step (F): For confirming the correctness of the number of groups, wherein V _kwon (q) is the number of target groups of the qth time, and V _kwon (q-1) is the number of the target group of the q- _{1th time} . The photo classification method of item 6 or item 8 of the claim, wherein the first threshold is 0.5. The photo classification method of claim 1, wherein the second threshold is 0.9. A location-based photo classification system for classifying a plurality of photos into a plurality of groups, comprising: a display unit for displaying a group number setting interface of the plurality of groups; and a storage unit for Storing the plurality of photos, each photo having a geographic location representing the location of the location, the geographic location being the location latitude and longitude of the photo; an arithmetic processing unit configured to perform the following steps: (A) receiving the plurality of photos and Setting a number of groups of the plurality of groups, each group having a group representative value, the group representative value is a two-dimensional center position of the group; (B) arbitrarily setting the group representative value and each One photo belongs to one of the probability of belonging to each group, and the probability of each of the photos is a total of 1; (C) correct each group according to the attribution value of each photo and the geographic location The group represents a value, and according to the corrected representative value of the group, corrects the attribution value of each photo to generate a target attribution value; (D) comparing the target attribution value with the previous time Target attribution value Whether the difference is less than a first threshold, if not, returning to step (C), and if so, performing step (E); (E) determining, in each of the affiliation values of each photo, whether the attribution value is greater than a second threshold, if any, sorting each photo based on the current number of the group and the attribution value, if not, based on the attribution value of each photo and the geographic location, the group of each group a representative value, generating a target group number value; and (F) comparing whether the difference between the target group number value and the previous target group number value is less than the first threshold value, and if so, according to the current group The number of groups and the attribution value are used to classify each photo. If not, the number of the groups is incremented by one and the process returns to step (B). The photo classification system of claim 11 further includes a group representative value correction function in the step (C): For correcting the group representative value of each group, where c _j is the group representative value of the jth group, N is the number of the photos, u _{ij is} the attribution value, and represents the jth photo The probability of belonging to the i-th group, m is a fixed value of 4.5, and x _i is the geographic information of the i-th photo. The photo classification system of claim 11 further includes a attribution value correction function in the step (C): For correcting the attribution value of each photo, where u _{ij is} the attribution value, representing the probability that the jth photo belongs to the i-th group, C is the number of the group, m is a fixed value of 4.5, and x _i is The geographic information of the i-th photo, c _j is the representative value of the group of the j-th group, and ck is the representative value of the group of the k-th group. The photo classification system of claim 11 further includes a target attribution value function in the step (C): The target value for generating a home, where, N for the number of photos, C for the number of group, u _ij for the attribution value that represents the j-th photo belonging to the i-th probability groups, m is a constant value 4.5 , x _i is the geographic information of the i-th photo, and c _j is the representative value of the group of the j-th group. The photo classification system of claim 14 further includes, in the step (D), a difference function of the target attribution value: For confirming the correctness of the attribution value, where J ( p ) is the target assignment value of the pth time, and J ( p -1) is the target attribution value of the p-1th time. The photo classification system of claim 11 further includes a target group number value function in the step (E): For generating the target group number value, where C is the number of the group, N is the number of the photos, u _{ij is} the attribution value, and represents the probability that the jth photo belongs to the i-th group, m is The value 4.5, c _j is the representative value of the group of the jth group, x _i is the geographic information of the i-th photo, c _i is the representative value of the group of the i-th group, and c is each group The group represents the average of the values. The photo classification system of claim 16 further includes, in the step (F), a difference function of the target group number value: For confirming the correctness of the number of groups, wherein V _kwon (q) is the number of target groups of the qth time, and V _kwon (q-1) is the number of the target group of the q- _{1th time} . The photo classification system of claim 15 or 17, wherein the first threshold is 0.5. The photo classification system of claim 11, wherein the second threshold is 0.9. A computer readable recording medium, wherein the computer readable recording medium records a set of computer executable programs, and when the computer readable recording medium is read by a processor, the processor executes the computer executable program To implement the steps as described in item 1 of the claim.