WO2015135277A1 - 聚类方法及相关装置 - Google Patents
聚类方法及相关装置 Download PDFInfo
- Publication number
- WO2015135277A1 WO2015135277A1 PCT/CN2014/082884 CN2014082884W WO2015135277A1 WO 2015135277 A1 WO2015135277 A1 WO 2015135277A1 CN 2014082884 W CN2014082884 W CN 2014082884W WO 2015135277 A1 WO2015135277 A1 WO 2015135277A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- distance
- weight
- objects
- classes
- category
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Definitions
- the present disclosure relates to the field of computer technologies, and in particular, to a clustering method and related apparatus. Background technique
- Clustering is the process of dividing a collection of physical or abstract objects into multiple classes consisting of similar objects. It is the process of classifying objects into different classes or clusters. The objects in the same class have great similarities, different classes. The objects in the group are very different.
- the hierarchical clustering method is to hierarchically decompose a given object until some termination condition is met.
- the condensed hierarchical clustering approach is a bottom-up strategy that first treats each object as a class and then merges those classes into larger and larger classes until a termination condition is met. Most hierarchical clustering methods fall into this category, but the definitions of similarity between classes are different.
- the clustering method when used for classifying pictures, the pictures belonging to the same person are divided into one class, and the related clustering method only uses the distance between the class and the class to measure the similarity between the two faces, each The distance between objects has the same contribution to the similarity measure, which results in a low accuracy of clustering results.
- a clustering method including: a weight coefficient corresponding to a distance between objects for all classes to be merged Obtaining a weight distance between two classes, wherein the weight coefficient is determined according to a similarity between two objects corresponding to the distance between the objects; determining whether the existence exists according to a weight distance between the two classes and a preset distance threshold Classes that can be merged; When there are classes that can be merged, all the classes that can be merged are merged, and the execution returns for all the classes to be merged, and the weight distance between the two classes is obtained according to the weight coefficient corresponding to the distance between the objects.
- the method further includes: Corresponding relationship between the distance between the objects and the probability that the two objects are the same object according to the sample object statistics; determining, according to the correspondence relationship, a mapping relationship between the distance between the objects and the weight coefficient, the weight coefficient according to The probability is determined.
- determining, according to the corresponding relationship, a mapping relationship between the distance between the objects and a weight coefficient Method: Query the corresponding relationship, and obtain a probability that two objects corresponding to the distance between the objects are the same object; and determine that the probability is a weight coefficient corresponding to the distance between the objects.
- the weight distance is a weight distance between the first class and the second class; For all the classes to be merged, the weight distance between the two classes is obtained according to the weight coefficient corresponding to the distance between the objects, in the following manner: according to all objects of the first class and all objects of the second class a distance, and a corresponding weight coefficient, obtaining a first one-way weight distance of the first class to the second class; acquiring a second one-way weight distance of the second class to the first class; The first one-way weight distance and the second one-way weight distance obtain a weight distance of the first class and the second class.
- the distance between all objects of the first class and all objects of the second class and Corresponding weight coefficient obtaining the first one-way weight distance of the first class to the second class, in the following manner: acquiring between any object in the first class and all objects in the second class a distance with a maximum similarity, and a first weight coefficient corresponding to the distance with the largest similarity; obtaining an object and a target in the first class according to a product of the distance with the largest similarity and a corresponding first weight coefficient a minimum weight distance between all objects of the second class; obtaining an average weight of distance between the object in the first class and other objects in the second class other than the object having the largest similarity distance in the second class a distance according to the minimum weight distance and the average weight distance to obtain a weight distance between the object in the first class and the second class; according to all objects and objects in the first class Between the second type of heavy weight from the weight and distance weight corresponding to the weight coefficient to
- a clustering apparatus including: An obtaining unit, configured to obtain a weight distance between two classes according to a weight coefficient corresponding to a distance between objects, for the class to be merged, the weight coefficient being determined according to similarity between two objects corresponding to the distance between the objects a determining unit, configured to determine, according to the weight distance between the two classes and the preset distance threshold, whether there is a class that can be merged; the merging unit is configured to perform, when there are classes that can be merged, all the classes that can be merged Combining, the obtaining unit performs a step of obtaining a weight distance between two classes according to a weight coefficient corresponding to the distance between the objects for all the classes to be merged, until the number of the merged classes is the same as the number of the pre-merging classes, and the aggregation is obtained.
- the device further includes: a statistical unit, configured to obtain, according to the sample object statistics, a distance between the objects and a probability that the two objects are the same object And a determining unit, configured to determine, according to the correspondence, a mapping relationship between the distance between the objects and a weight coefficient, where the weight coefficient is determined according to the probability.
- the determining unit includes: a query subunit, configured to query the corresponding relationship, to obtain the The probability that the two objects corresponding to the distance between the objects are the same person; the determining subunit is configured to determine the weight coefficient corresponding to the distance between the objects.
- the weight distance is the first type and the second The weighting distance between the classes
- the obtaining unit includes: a first acquiring subunit, configured to obtain a distance between the object in the first class and the object in the second class to be the most similar, And a first weighting coefficient corresponding to the distance with the largest similarity; a second acquiring subunit, configured to obtain, according to the distance with the largest similarity and the corresponding first weight coefficient, for all objects of the first class a first one-way weight distance of the first class to the second class; a third acquiring sub-unit, configured to acquire a second one-way weight distance of the second class to the first class; a subunit, configured to obtain, according to the first one-way weight distance and the second one-way weight distance, a weight distance of the first class and the second class.
- the second obtaining subunit includes: a fifth acquiring subunit, acquiring the first category a distance between the object and the object of the second class having the greatest similarity, and a first weight coefficient corresponding to the distance with the greatest similarity; a sixth obtaining subunit, configured to use the distance according to the maximum similarity And a minimum weighted distance between the object in the first class and all the objects in the second class; and a seventh acquiring subunit, configured to obtain the first class An average weight distance of the distance between the object and the object other than the object having the largest similarity distance in the second class; an eighth obtaining subunit, configured to use the minimum weight distance and the average weight The distance is obtained by the weight distance between the object in the first class and the second class; the ninth obtaining subunit is configured to be used according to all objects in the first class and the second class Weighted distance, weight and the weight corresponding to the distance weighting coefficient, to obtain a first type of the
- a terminal device including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: for all classes to be merged, according to The weight coefficient corresponding to the distance between the objects obtains a weight distance between the two classes, and the weight coefficient is determined according to the similarity between the two objects corresponding to the distance between the objects; according to the weight distance between the two classes and the preset
- the distance threshold is used to determine whether there are classes that can be merged; when the two classes can be merged, all the classes that can be merged are merged, and the execution returns for all the classes to be merged, and two weight coefficients are obtained according to the distance between the objects.
- the technical solution provided by the embodiment of the present disclosure may include the following beneficial effects: the clustering method obtains a weight distance between classes according to a weight coefficient corresponding to a distance between objects, and the weight coefficient is determined according to the similarity of two objects, that is, The distance between the objects is given a weight; then, the class whose weight is in accordance with the merge condition is terminated until the number of the merged classes is the same as the number of the pre-merged classes, and the clustering result is obtained.
- FIG. 1 is a flow chart showing a clustering method according to an exemplary embodiment.
- FIG. 2 is a flowchart of step S110 of FIG. 1 according to an exemplary embodiment;
- FIG. 3 is a flowchart of step S111 of FIG. 2 according to an exemplary embodiment;
- FIG. 4 is an exemplary implementation according to Illustrated flowchart of obtaining a weighting factor;
- FIG. 5 is a schematic diagram of a clustering apparatus according to an exemplary embodiment;
- FIG. 6 is a schematic diagram of another clustering apparatus according to an exemplary embodiment;
- 7 is a block diagram of a terminal device according to an exemplary embodiment;
- FIG. 8 is a block diagram of a server according to an exemplary embodiment.
- FIG. 1 is a flow chart of a clustering method according to an exemplary embodiment.
- the clustering method may include the following steps: In step S110, for all classes to be merged, weight distances between two classes are obtained according to weight coefficients corresponding to distances between objects, and the weights are The coefficients are determined based on the similarity between the objects. Assuming that the object is a face image, the clustering method provided by the present disclosure is capable of grouping images belonging to the same person to form a cluster. The features in the face image are converted into a set of vectors, so the distance between the objects is the distance between the vectors.
- the clustering method provided by the present disclosure can also be applied to other data than pictures.
- the weight coefficient may be determined according to whether the two objects corresponding to the distance between the objects are the same object.
- step S110 may include the following steps S111 to S113: in step S111, according to all objects of the first class and all objects of the second class The distance between the two, and the corresponding weight coefficient, obtain the first one-way weight distance of the first class to the second class.
- the step S111 may include the following steps S1111 to S1115: in step S1111, acquiring any object in the first class and the second class The distance with the greatest similarity between all the objects within, and the first weight coefficient corresponding to the distance with the greatest similarity.
- the first class is labeled A and the second class is labeled B. The weight distance between class A and class B is calculated.
- the distance between any object in class A and all objects in class B is calculated. ( ⁇ ,, ⁇ ), and get the maximum similarity between the object and all objects in the class.
- the cosine similarity is used in this embodiment. Therefore, the maximum cosine similarity d ⁇ Ai, B) corresponds to the greatest similarity between objects.
- the weight coefficient may be obtained according to a probability that two objects corresponding to the distance between the objects are the same object. As shown in FIG. 4, the following steps may be included: In step S100, according to the sample The object statistics obtain a correspondence between the distance between objects and the probability that two objects are the same object.
- the range of the cosine similarity co of two face images calculated from the high-dimensional features is [0, 1], which is obtained from a large number of face image statistics, when the cosine similarity is [ In the range of 0.45,1], the probability that two objects are the same person is basically 98%; when the cosine similarity is in the range of [0.35, 0.45), the probability that two objects are the same person is basically 70%; when the cosine is similar When the degree is in the range of [0.25, 0.35), the probability that two objects are the same person is basically 40%; when the cosine similarity is in the range of [0.15, 0.25), the probability that two objects are the same person is basically 10%; When the cosine similarity is in the range of [0, 0.15), the probability that two objects are the same person is basically 0.1%; According to the above statistical results, the relationship between the weight coefficient and the cosine similarity can be described by the following formula (1) -
- Equation (1) is a correspondence between the cosine similarity and the probability that two objects are the same person. Other types of distances are derived based on the relationship between the distances corresponding to the distances, and are not described here.
- step S200 a mapping relationship between the inter-object distance and the weight coefficient is determined according to the correspondence, and the weight coefficient is determined according to the probability.
- Corresponding weight coefficient is W (d ⁇ CA ⁇ B).
- step S1113 the distance between the object in the first class and the object other than the object corresponding to the distance with the greatest similarity in the second class is acquired.
- the average weight distance Assuming that the distance between the object in class A and the object b in class B is the largest, the average weight distance between the object and the other objects in class B other than object b is obtained according to formula (3). ) :
- ⁇ B) jeB j ⁇ b ⁇ ( 3 ) jeB, j ⁇ b, in step S1114, obtaining the object and the second in the first class according to the minimum weight distance and the average weight distance
- the weight distance between the classes; the minimum weight distance from the class B ⁇ ⁇ (A ⁇ B) and the average weight distance ⁇ ⁇ (AB), according to the public Equation (4) Get the weight distance of the object to class B ⁇ A ⁇ B):
- step S1115 according to the weight distance between all objects in the first class and the second class And a weight coefficient corresponding to the distance with the largest similarity, and obtaining a first one-way weight distance between the first class and the second class.
- the first one-way weight distance S (A, B) between the class and the class is obtained:
- W (d ⁇ CA ⁇ B) ) represents the maximum cosine similarity (minimum distance) between the object in class A and all objects in class B.
- d ⁇ A ⁇ B) The corresponding weight coefficient.
- step S112 the second unidirectional weight distance of the second class to the first class is obtained; the second unidirectional weight distance S (B, A) of the class B to the class A is calculated, according to the calculation class A.
- the first one-way weight distance to class B is similar, and will not be described here.
- the weight distances of the first class and the second class are obtained according to the first one-way weight distance and the second one-way weight distance.
- step S120 based on the weight distance between the two classes and the preset distance threshold, it is determined whether there is a class that can be merged.
- step S130 is performed; when there is no class that can be merged, step S140 is performed.
- the preset distance threshold may be set according to a data type of different objects, and according to the calculated type of distance between objects (for example, cosine similarity, Euclidean distance, etc.), for example, the object is a face image, And the distance between the objects is cosine similarity, the distance threshold can be set to 0.3 ⁇ 0.35.
- the type of distance between classes is different.
- the conditions for judging whether two classes can be combined are different: When the weight distance of the class is obtained according to the cosine similarity between the classes, the weight distance between the two classes is judged. Whether it is not less than the preset distance threshold, if the weight distance between the two classes is not less than the preset distance threshold, the table The similarity between the two classes can be combined. When the weight distance between the classes is calculated by the Euclidean distance or other distances, it is determined whether the weight distance between the two classes is not greater than a preset distance threshold, if the weight distance between the two classes is not greater than the preset distance threshold , indicating that the similarity of the two classes can be combined.
- the weight distance provided by the embodiment of the present disclosure is obtained by using the minimum weight distance and the average weight distance between the classes, so that the class can be more partial to the object, and the class with less objects can be taken into consideration, which is in line with the characteristics of the face clustering. Therefore, clustering using the weight distance can improve the clustering accuracy.
- step S130 all the classes that can be merged are merged.
- step S140 it is judged whether the number of merged classes is less than the number of merged classes, and if so, the process returns to step S110; otherwise, the process proceeds to step S150.
- the process of returning the weights between the two classes according to the weight coefficients corresponding to the distance between the objects is performed for all the classes to be merged, until the number of merged classes is
- the number of pre-merging classes is the same, that is, there are no classes that can be merged, and the clustering result is obtained.
- step S150 a clustering result is obtained.
- Objects that are aggregated into a class are very similar and have little dissimilarity. Taking a face image as an object, a face image that is grouped into one type is an image of the same person.
- the clustering method provided by the embodiment of the present disclosure obtains the weight distance between the classes according to the weight coefficient corresponding to the distance between the objects, and the weight coefficient is determined according to the similarity of the two objects, that is, the distance between the objects is given a weight; then, the combination The class whose weight distance meets the merge condition is terminated until the number of merged classes is the same as the number of merged classes, and the clustering result is obtained. Since the weight distance is related to the similarity of the two objects, the similarity of the objects is different, and the corresponding contribution to the clustering result is different. The greater the similarity, the greater the contribution is, so the clustering is improved. The accuracy of the results.
- FIG. 5 is a schematic diagram of a clustering apparatus according to an exemplary embodiment.
- the apparatus includes an acquisition unit 100, a determination unit 200, and a merging unit 300.
- the obtaining unit 100 is configured to obtain a weight distance between two classes according to a weight coefficient corresponding to the distance between objects for all classes to be merged, and the weight coefficient is similar to two objects corresponding to the distance between the objects. Sexual determination.
- the obtaining unit 100 may include a first acquiring subunit, a second acquiring subunit, and a third acquiring subunit, and a fourth acquiring subunit; the first obtaining subunit is configured to acquire a distance with the greatest similarity between any object in the first class and all objects in the second class, and the similarity The first weighting coefficient corresponding to the largest distance; the second acquiring subunit is configured to obtain, according to the distance with the greatest similarity and the corresponding first weight coefficient, for all objects of the first class a first unidirectional weight distance to the second class; in an embodiment of the present disclosure, the second obtaining subunit may include: a fifth obtaining subunit, a sixth obtaining subunit, and a seventh obtaining a subunit, an eighth obtaining subunit, and a ninth obtaining subunit; the fifth obtaining subunit is configured to obtain the greatest similarity between any object in the first class and all objects in the second class a distance, and a first weight coefficient corresponding to the distance with the greatest similarity; the sixth acquisition subunit is configured
- the third obtaining subunit is configured to acquire the second unidirectional weight distance of the second class to the first class; the fourth acquiring subunit is configured to be according to the first unidirectional weight distance and The second one-way weight distance obtains a weight distance between the first class and the second class.
- the determining unit 200 is configured to determine whether the two classes can merge the classes according to the weight distance between the two classes and the preset distance threshold.
- the merging unit 300 is configured to combine all the classes that can be merged when the two classes can be merged, and cause the acquiring unit to perform the class for all the to be merged, and obtain two weight coefficients according to the distance between the objects.
- the obtaining unit obtains the weight distance between the classes according to the weight coefficient corresponding to the distance between the objects, and the weight coefficient is determined according to the similarity of the two objects, that is, the weight between the objects is weighted; After the judging unit judges the mergeable class according to the weight distance, the merge unit merges the weights that meet the merge condition until the number of the merged classes is the same as the number of the pre-merged classes, and the clustering result is obtained.
- FIG. 6 is a schematic diagram of another clustering device according to an exemplary embodiment.
- the apparatus includes a statistic unit 400, a determining unit 500, an obtaining unit 100, a judging unit 200, and a merging unit 300.
- the functions of the same elements as those of the embodiment shown in FIG. 5 are also the same, and will not be described again in this embodiment.
- the statistical unit 400 is configured to obtain a correspondence between the distance between the objects and the probability that the two objects are the same object according to the sample object statistics.
- the determining unit 500 is configured to determine a mapping relationship between the inter-object distance and a weight coefficient according to the correspondence, and the weight coefficient is determined according to the probability.
- the determining unit may include a query subunit and a determining subunit: the query subunit is configured to query the correspondence, and obtain a probability that two objects corresponding to the distance between the objects are the same person.
- the determining subunit is configured to determine that the probability is a weighting coefficient corresponding to the distance between the objects.
- the acquisition unit 300 connected to the determination unit 500 obtains a weight distance between two classes according to a weight coefficient corresponding to the distance between the objects.
- the clustering apparatus provided by the embodiment of the present disclosure first determines the weight coefficient corresponding to the probability that the distance between the objects and the two objects are the same object according to a large number of sample objects, and then obtains the weight distance between the classes, and determines according to the weight distance.
- the merged class and merge the classes that can be merged, until the number of classes before and after the merge is unchanged, and the clustering result is obtained. Since the weight distance is associated with the probability that two objects are the same object, the probability that the two objects are the same object can be considered as the distance between the two objects changes, so that the contribution of the distance between the different objects is different, the probability The larger the value, the larger the contribution is. Therefore, the clustering method can improve the accuracy of the clustering result.
- FIG. 7 is a block diagram of a terminal device 800 for clustering, according to an exemplary embodiment.
- the terminal device 800 can be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like. Referring to FIG.
- the terminal device 800 can include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, and a sensor component 814. And a communication component 816.
- Processing component 802 typically controls the overall operation of terminal device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
- Processing component 802 can include one or more processors 820 to execute instructions to perform all or part of the steps of the above described methods.
- processing component 802 can include one or more modules to facilitate interaction between component 802 and other components.
- processing component 802 can include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.
- Memory 804 is configured to store various types of data to support operation at device 800. Examples of such data include instructions for any application or method operating on terminal device 800, contact data, phone book data, messages, pictures, videos, and the like. Memory 804 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Disk or Optical Disk.
- Power component 806 provides power to various components of terminal device 800.
- Power component 806 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for terminal device 800.
- the multimedia component 808 includes a screen that provides an output interface between the terminal device 800 and a user.
- the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user.
- the touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can sense not only the boundary of the touch or sliding action, but also the duration and pressure associated with the touch or slide operation.
- the multimedia component 808 includes a front camera and/or a rear camera.
- the audio component 810 is configured to output and/or input an audio signal.
- the audio component 810 includes a microphone (MIC) that, when the terminal device 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode, The microphone is configured to receive an external audio signal.
- the received audio signal may be further stored in memory 804 or transmitted via communication component 816.
- the audio component 810 also includes a speaker for outputting an audio signal.
- the 1/0 interface 812 provides an interface between the processing component 802 and the peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons can include, but are not limited to: Home button, Volume button, Start button, and Lock button.
- Sensor component 814 includes one or more sensors for providing terminal device 800 with a status assessment of various aspects. For example, sensor component 814 can detect an open/closed state of device 800, a relative positioning of components, such as the display and keypad of terminal device 800, and sensor component 814 can also detect terminal device 800 or terminal device 800 The position of the component changes, the presence or absence of contact of the user with the terminal device 800, the orientation or acceleration/deceleration of the terminal device 800, and the temperature change of the terminal device 800.
- Sensor assembly 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
- Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
- the sensor assembly 814 can also include an acceleration sensor, a gyro sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
- Communication component 816 is configured to facilitate wired or wireless communication between terminal device 800 and other devices.
- the terminal device 800 can access a wireless network based on a communication standard such as WiFi, 2G, 3G or 4G, or a combination thereof.
- communication component 816 receives broadcast signals or broadcast associated information from an external broadcast management system via a broadcast channel.
- the communication component 816 also includes a near field communication (NFC) module to facilitate short range communication.
- NFC near field communication
- the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
- terminal device 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), A gated array (FPGA), controller, microcontroller, microprocessor, or other electronic component implementation is used to perform the above methods.
- ASICs application specific integrated circuits
- DSPs digital signal processors
- DSPDs digital signal processing devices
- PLDs programmable logic devices
- FPGA gated array
- controller microcontroller, microprocessor, or other electronic component implementation is used to perform the above methods.
- non-transitory computer readable storage medium comprising instructions, such as a memory 804 comprising instructions executable by the processor 820 of the terminal device 800 to perform the above method.
- the non-transitory computer readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device.
- a non-transitory computer readable storage medium when instructions in the storage medium are executed by a processor of a mobile terminal, enabling the mobile terminal to perform a clustering method, the method comprising: For all the classes to be merged, the weight distance between the two classes is obtained according to the weight coefficient corresponding to the distance between the objects, and the weight coefficient is determined according to the similarity of the two objects corresponding to the distance between the objects; The weight distance and the preset distance threshold determine whether there is a class that can be merged; when there are classes that can be merged, all the classes that can be merged are merged, and the execution is performed for all the classes to be merged, according to the distance between the objects.
- the corresponding weight coefficient obtains the weight distance between the two classes, until the number of merged classes is the same as the number of the pre-merged classes, and the clustering result is obtained.
- the method further includes: determining, according to the sample object statistics, a correspondence between the distance between the objects and the probability that the two objects are the same object; determining, according to the correspondence, the distance between the objects and the weight coefficient The mapping relationship between the weighting coefficients is determined according to the probability.
- determining, according to the correspondence, a mapping relationship between the distance between the objects and a weight coefficient by using the following manner: querying the corresponding relationship, and obtaining whether two objects corresponding to the distance between the objects are the same The probability of an object; determining the probability is a weighting coefficient corresponding to the distance between the objects.
- the weight distance is a weight distance between the first class and the second class; and the weight distance between the two classes is obtained according to the weight coefficient corresponding to the distance between the objects for all the classes to be merged,
- the first unidirectional weight of the first class to the second class is obtained according to a distance between all objects of the first class and all objects of the second class, and a corresponding weight coefficient.
- FIG. 13 is a schematic structural diagram of a server in an embodiment of the present invention.
- the server 1900 can be configured or sexually Different differences can be made, including one or more central processing units (CPUs) 1922 (eg, one or more processors) and memory 1932, one or more storage applications 1942 or data.
- Storage medium 1930 of 1944 eg, one or one storage device in Shanghai.
- the memory 1932 and the storage medium 1930 may be short-term storage or persistent storage.
- the program stored on the storage medium 1930 may include one or more modules (not shown), each of which may include a series of instruction operations in the terminal device.
- central processor 1922 can be configured to communicate with storage medium 1930, executing a series of instruction operations in storage medium 1930 on server 1900.
- Server 1900 may also include one or more power supplies 1926, one or more wired or wireless network interfaces 1950, one or more input and output interfaces 1958, one or more keyboards 1956, and/or one or more operating systems 1941.
- power supplies 1926 For example, Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and more.
- non-transitory computer readable storage medium comprising instructions, such as a memory 1932 or a storage medium 1930, which may be executed by the processor 1922 of the terminal device to perform the above method.
- the non-transitory computer readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device.
- a non-transitory computer readable storage medium when instructions in the storage medium are executed by a processor of a terminal device, enabling the terminal device to perform a clustering method, the method comprising: targeting all classes to be merged Obtaining a weight distance between two classes according to a weight coefficient corresponding to the distance between the objects, wherein the weight coefficient is determined according to the similarity of two objects corresponding to the distance between the objects; according to the weight distance between the two classes and the pre- Set the distance threshold to determine whether there are classes that can be merged.
- the method further includes: determining, according to the sample object statistics, a correspondence between the distance between the objects and the probability that the two objects are the same object; determining, according to the correspondence, the distance between the objects and the weight coefficient The mapping relationship between the weighting coefficients is determined according to the probability.
- determining, according to the correspondence, a mapping relationship between the distance between the objects and a weight coefficient by using the following manner: querying the corresponding relationship, and obtaining whether two objects corresponding to the distance between the objects are the same The probability of an object; Determining the probability is a weight coefficient corresponding to the distance between the objects.
- the weight distance is a weight distance between the first class and the second class; and the weight distance between the two classes is obtained according to the weight coefficient corresponding to the distance between the objects for all the classes to be merged,
- the first unidirectional weight of the first class to the second class is obtained according to a distance between all objects of the first class and all objects of the second class, and a corresponding weight coefficient.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
Description
Claims
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| MX2014011016A MX346698B (es) | 2014-03-14 | 2014-07-24 | Metodo y dispositivo de agrupamiento. |
| RU2015129671A RU2630373C2 (ru) | 2014-03-14 | 2014-07-24 | Способ и устройство для кластеризации |
| KR1020147026522A KR101639502B1 (ko) | 2014-03-14 | 2014-07-24 | 클러스터링 방법, 관련장치, 프로그램 및 기록매체 |
| JP2016506779A JP6051336B2 (ja) | 2014-03-14 | 2014-07-24 | クラスタリング方法、クラスタリング装置、端末装置、プログラム及び記録媒体 |
| US14/526,477 US20150262033A1 (en) | 2014-03-14 | 2014-10-28 | Method and terminal device for clustering |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410096608.9A CN103927545B (zh) | 2014-03-14 | 2014-03-14 | 聚类方法及相关装置 |
| CN201410096608.9 | 2014-03-14 |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US14/526,477 Continuation US20150262033A1 (en) | 2014-03-14 | 2014-10-28 | Method and terminal device for clustering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2015135277A1 true WO2015135277A1 (zh) | 2015-09-17 |
Family
ID=51145761
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2014/082884 Ceased WO2015135277A1 (zh) | 2014-03-14 | 2014-07-24 | 聚类方法及相关装置 |
Country Status (7)
| Country | Link |
|---|---|
| EP (1) | EP2919136A1 (zh) |
| JP (1) | JP6051336B2 (zh) |
| KR (1) | KR101639502B1 (zh) |
| CN (1) | CN103927545B (zh) |
| MX (1) | MX346698B (zh) |
| RU (1) | RU2630373C2 (zh) |
| WO (1) | WO2015135277A1 (zh) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111415151A (zh) * | 2020-03-10 | 2020-07-14 | 支付宝(杭州)信息技术有限公司 | 连锁商户的识别方法、装置、电子设备及存储介质 |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927545B (zh) * | 2014-03-14 | 2017-10-17 | 小米科技有限责任公司 | 聚类方法及相关装置 |
| CN105488467B (zh) * | 2015-11-26 | 2019-07-23 | 小米科技有限责任公司 | 图片分类方法及装置 |
| CN107092618A (zh) * | 2016-10-27 | 2017-08-25 | 北京小度信息科技有限公司 | 一种信息处理方法及装置 |
| CN109469919B (zh) * | 2018-11-12 | 2020-07-28 | 南京工程学院 | 一种基于权重聚类的电站空预器堵灰监测方法 |
| CN111310834B (zh) * | 2020-02-19 | 2024-05-28 | 深圳市商汤科技有限公司 | 数据处理方法及装置、处理器、电子设备、存储介质 |
| CN112132111A (zh) * | 2020-10-10 | 2020-12-25 | 安徽江淮汽车集团股份有限公司 | 泊车典型场景的提取方法、设备、存储介质及装置 |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214293A (zh) * | 2010-04-09 | 2011-10-12 | 索尼公司 | 面部群集设备、面部群集方法和程序 |
| US20120039515A1 (en) * | 2007-01-04 | 2012-02-16 | Samsung Electronic Co. Ltd. | Method and system for classifying scene for each person in video |
| US20120294540A1 (en) * | 2011-05-17 | 2012-11-22 | Microsoft Corporation | Rank order-based image clustering |
| CN103927545A (zh) * | 2014-03-14 | 2014-07-16 | 小米科技有限责任公司 | 聚类方法及相关装置 |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100318512B1 (ko) * | 1998-02-14 | 2002-04-22 | 이계철 | 두그룹간의유사도계산방법 |
| US7664735B2 (en) * | 2004-04-30 | 2010-02-16 | Microsoft Corporation | Method and system for ranking documents of a search result to improve diversity and information richness |
| US7809192B2 (en) * | 2005-05-09 | 2010-10-05 | Like.Com | System and method for recognizing objects from images and identifying relevancy amongst images and information |
| KR100824698B1 (ko) * | 2006-08-25 | 2008-04-24 | 인하대학교 산학협력단 | 이미지 검색기의 적합성 피드백을 위한 공간 위치정보 기반영역 가중치 방법 |
| RU2345414C1 (ru) * | 2007-08-10 | 2009-01-27 | Общество с ограниченной ответственностью "Рекогмишн" | Способ построения системы индексирования для поиска объектов на цифровых изображениях |
| WO2011152821A1 (en) * | 2010-06-01 | 2011-12-08 | Hewlett-Packard Development Company, L.P. | Clustering images |
| WO2012089902A1 (en) * | 2010-12-30 | 2012-07-05 | Nokia Corporation | Method, apparatus, and computer program product for image clustering |
| CN102881032B (zh) * | 2012-09-14 | 2015-09-09 | 北京理工大学 | 一种基于改进算法的图片拼接方法 |
| CN103136355B (zh) * | 2013-03-05 | 2016-01-06 | 电子科技大学 | 一种基于自动阈值鱼群算法的文本聚类方法 |
| CN103544255B (zh) * | 2013-10-15 | 2017-01-11 | 常州大学 | 基于文本语义相关的网络舆情信息分析方法 |
-
2014
- 2014-03-14 CN CN201410096608.9A patent/CN103927545B/zh not_active Expired - Fee Related
- 2014-07-24 RU RU2015129671A patent/RU2630373C2/ru active
- 2014-07-24 JP JP2016506779A patent/JP6051336B2/ja active Active
- 2014-07-24 WO PCT/CN2014/082884 patent/WO2015135277A1/zh not_active Ceased
- 2014-07-24 MX MX2014011016A patent/MX346698B/es active IP Right Grant
- 2014-07-24 KR KR1020147026522A patent/KR101639502B1/ko active Active
-
2015
- 2015-02-06 EP EP15154051.5A patent/EP2919136A1/en not_active Ceased
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120039515A1 (en) * | 2007-01-04 | 2012-02-16 | Samsung Electronic Co. Ltd. | Method and system for classifying scene for each person in video |
| CN102214293A (zh) * | 2010-04-09 | 2011-10-12 | 索尼公司 | 面部群集设备、面部群集方法和程序 |
| US20120294540A1 (en) * | 2011-05-17 | 2012-11-22 | Microsoft Corporation | Rank order-based image clustering |
| CN103927545A (zh) * | 2014-03-14 | 2014-07-16 | 小米科技有限责任公司 | 聚类方法及相关装置 |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111415151A (zh) * | 2020-03-10 | 2020-07-14 | 支付宝(杭州)信息技术有限公司 | 连锁商户的识别方法、装置、电子设备及存储介质 |
Also Published As
| Publication number | Publication date |
|---|---|
| EP2919136A1 (en) | 2015-09-16 |
| CN103927545B (zh) | 2017-10-17 |
| KR101639502B1 (ko) | 2016-07-22 |
| MX2014011016A (es) | 2016-02-19 |
| RU2630373C2 (ru) | 2017-09-07 |
| JP2016517110A (ja) | 2016-06-09 |
| KR20150117201A (ko) | 2015-10-19 |
| CN103927545A (zh) | 2014-07-16 |
| JP6051336B2 (ja) | 2016-12-27 |
| RU2015129671A (ru) | 2017-04-27 |
| MX346698B (es) | 2017-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103914518B (zh) | 聚类方法及相关装置 | |
| WO2015135277A1 (zh) | 聚类方法及相关装置 | |
| KR101646887B1 (ko) | 클러스터링 방법, 장치 및 단말 장치 | |
| WO2022198853A1 (zh) | 任务调度方法及装置、电子设备、存储介质和程序产品 | |
| CN110019888A (zh) | 一种搜索方法及装置 | |
| CN107390932B (zh) | 边缘防误触方法及装置、计算机可读存储介质 | |
| CN106484284B (zh) | 切换单手模式的方法和装置 | |
| WO2022110614A1 (zh) | 手势识别方法及装置、电子设备和存储介质 | |
| WO2018133387A1 (zh) | 指纹识别方法及装置 | |
| WO2017107419A1 (zh) | 屏幕解锁方法、装置及终端 | |
| JP2016517110A5 (zh) | ||
| CN107704190B (zh) | 手势识别方法、装置、终端及存储介质 | |
| WO2019237870A1 (zh) | 目标匹配方法及装置、电子设备和存储介质 | |
| CN106572268B (zh) | 信息显示方法及装置 | |
| CN110083266B (zh) | 信息处理方法、装置及存储介质 | |
| WO2020103353A1 (zh) | 多波束选取方法及装置 | |
| CN105678266A (zh) | 一种合并人脸相册的方法及装置 | |
| WO2017088238A1 (zh) | 触摸手势确定方法及装置 | |
| CN107292306A (zh) | 目标检测方法及装置 | |
| CN107423757A (zh) | 聚类处理方法及装置 | |
| CN111610921A (zh) | 手势识别方法及装置 | |
| CN111368161A (zh) | 一种搜索意图的识别方法、意图识别模型训练方法和装置 | |
| CN113312475B (zh) | 一种文本相似度确定方法及装置 | |
| US20150262033A1 (en) | Method and terminal device for clustering | |
| CN106201192A (zh) | 点击事件响应方法、装置及终端 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| ENP | Entry into the national phase |
Ref document number: 2016506779 Country of ref document: JP Kind code of ref document: A |
|
| WWE | Wipo information: entry into national phase |
Ref document number: MX/A/2014/011016 Country of ref document: MX |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 1020147026522 Country of ref document: KR |
|
| ENP | Entry into the national phase |
Ref document number: 2015129671 Country of ref document: RU Kind code of ref document: A |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 14885748 Country of ref document: EP Kind code of ref document: A1 |
|
| REG | Reference to national code |
Ref country code: BR Ref legal event code: B01A Ref document number: 112014024337 Country of ref document: BR |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 14885748 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 112014024337 Country of ref document: BR Kind code of ref document: A2 Effective date: 20140930 |