Object tracking method, device and computer readable storage medium

By converting the characteristics of multiple data types into the same data type and generating a hash code, the limitations of hash functions for single data types are overcome, enabling stable tracking of complex clients and improving the adaptability and accuracy of tracking.

CN112887256BActive Publication Date: 2024-11-22HUAWEI CLOUD COMPUTING TECHNOLOGIES CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN201911201522.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-11-29
Publication Date
2024-11-22
Estimated Expiration
2039-11-29

AI Technical Summary

Technical Problem

In existing technologies, hash functions can only be designed for the characteristics of one type of data, which cannot effectively track clients with complex data types and lacks the ability to adapt to changes in characteristics, leading to tracking failure.

Method used

By converting features of various data types into the same data type and generating hash codes based on feature weights, the number of bits in the hash codes can be adjusted using weight learning, thus enabling the tracking of objects of different data types.

Benefits of technology

It improves the adaptability and accuracy of object tracking, can handle the characteristics of various data types, and ensures the stability and accuracy of tracking.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112887256B_ABST
    Figure CN112887256B_ABST
Patent Text Reader

Abstract

The present application provides an object tracking method, device and computer-readable storage medium. The method includes: a server obtains a feature instance of an object, the object is a client served by the server, the feature instance contains multiple features of different data types, and the multiple features are used to indicate the hardware and / or software information of the client; the server converts the data types of the multiple features into the same data type to obtain the encoding of the feature instance; the server converts the encoding of the feature instance into a hash code according to the weights of the multiple features, wherein the weight of each feature is positively correlated to the number of bits it occupies in the hash code; the server tracks the object according to the hash code. The technical solution provided by the present application can realize the tracking of objects with different data types based on hash technology.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of target tracking, and more specifically, to a method, device and computer-readable storage medium for object tracking. Background Art

[0002] In the server-client architecture, in some applications, the server needs to track the client as an object. In specific implementation, the server will track the client based on the client's feature instance. The client's feature instance usually contains multiple features of the client (for example, it can contain various software and / or hardware information of the client), and the data types of the multiple features are different.

[0003] In order to achieve fast tracking of objects, one possible implementation method is to use hash technology to track objects. However, in the traditional object tracking method based on hash technology, the hash function is designed for the characteristics of a data type and cannot track clients with complex data types. Summary of the invention

[0004] The present application provides a method, an apparatus, and a computer-readable storage medium for object tracking, which can process features of multiple different data types and implement tracking of objects with different data types.

[0005] In a first aspect, a method for object tracking is provided, the method comprising: a server side obtains a feature instance of an object, the object being a client served by the server side, the feature instance comprising a plurality of features of different data types, the plurality of features being used to indicate hardware and / or software information of the client side; the server side converts the different data types of the plurality of features into the same data type to obtain an encoding of the feature instance; the server side converts the encoding of the feature instance into a hash code according to the weights of the plurality of features, wherein the weight of each feature is positively correlated with the number of bits occupied by each feature in the hash code; the server side tracks the object according to the hash code.

[0006] By converting multiple types of different data types into the same data type, the server can process the characteristics of multiple different data types and track objects with different data types.

[0007] In combination with the first aspect, in some implementations of the first aspect, the encoding of the feature instance is binary encoding.

[0008] In combination with the first aspect, in certain implementations of the first aspect, the encoding of the feature instance includes the encoding corresponding to each feature of the multiple features, and the server converts the encoding of the feature instance into a hash code according to the weights of the multiple features, including: the server determines the number of bits occupied by each feature in the hash code according to the weight of each feature and the number of bits of the hash code; the server samples the encoding corresponding to each feature according to the number of bits occupied by each feature in the hash code to obtain a sampling result corresponding to each feature, the number of bits of the sampling result corresponding to each feature being equal to the number of bits occupied by each feature in the hash code; the server merges the sampling results of the multiple features to obtain the hash code.

[0009] In combination with the first aspect, in certain implementations of the first aspect, the server samples the code corresponding to each feature according to the number of bits occupied by each feature in the hash code, including: the server randomly samples the code corresponding to each feature according to the number of bits occupied by each feature in the hash code.

[0010] In combination with the first aspect, in some implementations of the first aspect, the weights of the multiple features are updated according to multiple feature instances stored in a database.

[0011] In combination with the first aspect, in certain implementations of the first aspect, updating the weights of the multiple features based on the multiple feature instances stored in the database includes: determining new weights of the multiple features based on the multiple feature instances stored in the database, so that the distance between the feature instances of the same object is smaller than the distance between the feature instances of different objects, wherein the distance between two feature instances in the multiple feature instances is the weighted sum of the distances of corresponding features in the two feature instances; if the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, updating the current weights of the multiple features.

[0012] By comparing the differences between the new weights and current weights of multiple features to timely update the current weights, and determining the number of bits occupied by each feature in the hash code based on the weights of multiple features, the obtained hash code can be adapted to the feature changes of the object, thereby improving the adaptive ability of tracking and improving the accuracy of tracking.

[0013] In a second aspect, a device for object tracking is provided, the device comprising: an acquisition module, used to acquire a feature instance of an object, the object being a client served by the server, the feature instance comprising multiple features of different data types, the multiple features being used to indicate hardware and / or software information of the client; a feature encoding module, used to convert different data types of the multiple features into the same data type to obtain an encoding of the feature instance; a hash encoding module, used to convert the encoding of the feature instance into a hash code according to the weights of the multiple features, wherein the weight of each feature is positively correlated with the number of bits occupied by each feature in the hash code; and a tracking module, used to track the object according to the hash code.

[0014] By converting multiple types of different data types into the same data type, the server can process the characteristics of multiple different data types and track objects with different data types.

[0015] In combination with the second aspect, in some implementations of the second aspect, the encoding of the feature instance is binary encoding.

[0016] In combination with the second aspect, in certain implementations of the second aspect, the encoding of the feature instance includes the encoding corresponding to each feature of the multiple features, and the hash coding module is specifically used to: determine the number of bits occupied by each feature in the hash code based on the weight of each feature and the number of bits of the hash code; sample the encoding corresponding to each feature based on the number of bits occupied by each feature in the hash code to obtain a sampling result corresponding to each feature, the number of bits of the sampling result corresponding to each feature being equal to the number of bits occupied by each feature in the hash code; and merge the sampling results of the multiple features to obtain the hash code.

[0017] In combination with the second aspect, in some implementations of the second aspect, the hash coding module is further used to: randomly sample the code corresponding to each feature according to the number of bits occupied by each feature in the hash code.

[0018] In combination with the second aspect, in some implementations of the second aspect, the device further includes: an updating module, configured to update the weights of the multiple features according to the multiple feature instances stored in the database.

[0019] In conjunction with the second aspect, in some implementations of the second aspect, the update module is further used to:

[0020] According to the multiple feature instances stored in the database, new weights of the multiple features are determined so that the distance between the feature instances of the same object is smaller than the distance between the feature instances of different objects, wherein the distance between two feature instances of the multiple feature instances is the weighted sum of the distances of corresponding features in the two feature instances; if the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, the current weights of the multiple features are updated.

[0021] By comparing the differences between the new weights and current weights of multiple features to timely update the current weights, and determining the number of bits occupied by each feature in the hash code based on the weights of multiple features, the obtained hash code can be adapted to the feature changes of the object, thereby improving the adaptive ability of tracking and improving the accuracy of tracking.

[0022] In a third aspect, a computer-readable storage medium is provided, comprising a computer program, which, when executed on a computer device, enables a processing unit in the computer device to execute the object tracking method in any possible implementation of any of the above aspects.

[0023] According to a fourth aspect, a computer program product is provided, comprising a computer program, which, when executed on a computer device, enables a processing unit in the computer device to execute the method according to the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the object tracking system architecture using hash coding technology.

[0025] Figure 2 It is a schematic diagram of an object tracking method according to an embodiment of the present application.

[0026] Figure 3 It is a schematic diagram of the system architecture of the object tracking method according to an embodiment of the present application.

[0027] Figure 4 It is a flowchart of the object tracking method of an embodiment of the present application.

[0028] Figure 5 is a schematic diagram of an object tracking device according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0030] The embodiments of the present application can be used for object tracking, such as web browser tracking, mobile application tracking, etc.

[0031] To facilitate understanding, the application scenarios of the embodiments of the present application are first briefly introduced.

[0032] In systems with a master-slave architecture such as server / client and server / browser, the server usually needs to uniquely identify a client (or browser) based on the client's (or browser's) hardware, software, and runtime information. Here, the served subjects such as clients and browsers are collectively referred to as objects, and the hardware, software, and runtime information of the objects are called features. During the operation of the system, the features of the object may change, such as hardware configuration changes, software upgrades, network connection changes, etc. After the object features change, the collected feature instances change when the server receives the object's request. In this case where the object features are variable, the server still needs to overcome the interference of feature changes and continuously and accurately track the object based on the feature instances sent to the server by the object.

[0033] When an object sends a request to the server and needs to be tracked, the server extracts features based on the data sent by the object, and further matches the feature instance with historical data. If the similarity match is successful, the current request is associated with the historical request, thus achieving the goal of tracking the object. To achieve similarity matching and fast tracking, the industry usually uses a database to store data and uses indexes to speed up queries. For example, Figure 1 This is a schematic diagram of the object tracking system architecture using hash coding technology. Figure 1 As shown, the object tracking system architecture includes an identification (ID) retrieval and generation system 110 and a data storage system 120. The ID retrieval and generation system 110 includes an anti-disturbance hash coding module 1101, a fast code retrieval module 1102 and a unique ID generation module 1103, and the data storage system 120 includes an original feature library 1201 and a code retrieval library 1202. After receiving the object request, the system extracts the feature instance, calculates the hash code through the anti-disturbance hash coding module 1101, and then queries the existing object ID associated with the hash code in the database through the fast code retrieval module 1102. Since the hash function has a fast calculation speed, data with high similarity will most likely get the same hash code, so fast similarity search and tracking can be achieved.

[0034] However, the hash coding technology in the prior art can only define a specific hash function for a specific distance measurement method. A distance measurement method is usually only suitable for one type of feature, and the feature value types of the current object are usually rich, including character type, enumeration (category) type, numerical type, etc. Therefore, the hash coding technology used in the prior art cannot complete the tracking of objects with multiple feature types. Moreover, the prior art has no adaptability to changes in object features, and its distance measurement method cannot be adjusted as the object's features change, which will lead to incorrect hash coding and make tracking invalid.

[0035] The embodiment of the present application converts different data types of multiple features into the same data type, so that the server can process features of multiple different data types, thereby achieving tracking of objects with different data types.

[0036] The following will be combined Figure 2 The object tracking method of the present application is described in detail.

[0037] Figure 2 Schematic diagram of an object tracking method according to an embodiment of the present application. Figure 2 As shown, the method 200 includes steps S210 to S240, which are described in detail below.

[0038] S210, the server obtains a feature instance of the object.

[0039] As an embodiment, the object may be a client served by the server, and the feature instance includes multiple features with different data types, and the multiple features are used to indicate hardware and / or software information of the client.

[0040] Optionally, the object may be a device-type object such as a browser or a mobile application, or may be an online text object such as the same news or the same post on different websites.

[0041] It should be understood that the hardware and / or software information of the client indicated by the various features of the object may include: hardware configuration, software version, etc. Optionally, the various features of the object may also be used to indicate runtime information, such as network connection, etc.

[0042] S220, the server converts different data types of the multiple features into the same data type to obtain the encoding of the feature instance.

[0043] It should be understood that different data types of various features of an object may include: character type, enumeration (category) type, numeric type, and the like.

[0044] As an embodiment, the encoding of the feature instance is binary encoding. The server encodes different data types of multiple features separately and converts them into binary encoding. Specifically, for example: when the data type of the feature is an integer type: if the value range is C, the encoding length is C, if the value is k, the first k bits are 1, and the rest are 0; when it is a character type: take the first few bits after md5 hashing, and the number of bits is configurable; when it is an enumeration type: one-hot encoding, that is, if there are C possible values, it is encoded as C-bit binary, when the value is k, the kth bit is 1, and the rest are 0; when it is a floating point type: if the value interval length is Z, and the value is x, then it is calculated as That is, normalize the interval from 1 to [0,1], divide the interval into C equal parts, round down, and then encode it as an integer.

[0045] By converting different data types of multiple features into the same data type, it can handle features of multiple different data types.

[0046] S230, the server converts the encoding of the feature instance into a hash code according to the weights of the multiple features.

[0047] Optionally, the weight of each feature may be positively correlated to the number of bits each feature occupies in the hash code.

[0048] It should be understood that the weight of each feature is obtained through feature weight learning.

[0049] As an embodiment, the encoding of the feature instance is binary encoding, and the server selects the corresponding sampling bits from the binary encoding of each feature according to the weight of each feature, and then concatenates the sampling bits of all features to form a hash code. Specifically, for a given hash code length N, the server determines the number of sampling bits to be taken from the binary encoding of the feature according to the weight of the feature, such as: if the weight of a feature is w (0<w<1), then the number of sampling bits of the feature in the corresponding binary encoding is u=round(N×w) (rounded to an integer). Optionally, the server randomly selects u binary bits in the corresponding binary encoding as the sampling bits of the feature according to the number of bits u of the sampling bits of the feature, and the selected sampling bits are used as the index number of the binary code. In the same way, the server selects the sampling bits of the feature in the binary codes corresponding to all features, and then concatenates the sampling bits of all features to form a hash code. For example, given the final hash code length N = 20, the binary code of a feature is 010010100101101, and its weight is w = 0.1, then u = N × w = 2 bits are randomly collected on this binary, and it is assumed that the results of the random selection are the 5th and 11th bits, and the binary bit index numbers here are 5 and 11. Optionally, once the index number is determined, it will not change until the system is updated.

[0050] S240: The server tracks the object according to the hash code.

[0051] As an embodiment, the server retrieves the generated hash code in the database. If the query result is not empty, the server reads the result to obtain the object ID (if there are multiple object IDs, conflicts are resolved according to the majority voting principle). If the query result is empty, a new object ID is generated by the unique ID generation module. The system binds this ID to the corresponding object to implement object tracking, and the process ends.

[0052] The embodiment of the present application converts multiple types of different data types into the same data type, so that the server can process features of multiple different data types; at the same time, by converting the encoding of the feature instance into a hash code according to the weights of multiple features, the adaptive capability of tracking can be improved, thereby improving the accuracy of tracking.

[0053] Figure 3 A schematic diagram of the system architecture of the object tracking method in the embodiment of the present application is given. Figure 3 As shown, the system architecture 300 includes a weight learning system 330, an ID retrieval and generation system 310, and a data storage system 320. These systems are described in detail below.

[0054] The weight learning system 330 is used to output feature weights through a learning algorithm based on historical data from the server.

[0055] The weight learning system 330 may further include a feature weight learning module 3301, which may be used to learn feature weights based on feature data of objects in a database.

[0056] As an embodiment, the feature weight learning module can define a meta-distance measurement method for each type of feature, and then use the labeled data to learn the distance parameters and learn the distance weights of multiple features. Specifically, when the data type of the feature is character type, data type and enumeration type, the definition method is:

[0057] Character type: the same distance is 0, different distance is 1; numeric type: the same distance is 0, different distance is the absolute value of the numerical difference / the absolute value of the maximum numerical difference; enumeration type: the same distance is 0, different distance is 1.

[0058] The server then uses the labeled data to learn the distance parameters and the distance weights of multiple features. Based on the multiple feature instances stored in the database, the weights of multiple features are determined so that the distance between feature instances of the same object is smaller than the distance between feature instances of different objects, wherein the distance between two feature instances in the multiple feature instances can be the weighted sum of the distances of the corresponding features in the two feature instances. Specifically, w krepresents the weight of the kth feature, represents the meta-distance between the i-th object instance and the j-th object instance in the k-th feature dimension, then the distance between object instance i and object instance j is is the weighted sum of all feature dimensions, that is During training, there is known annotation information, that is, whether object instance i and object instance j are the same object. Here, t ij =1 means it is the same object, t ij =-1 means they are not the same object. So, given a training set, the algorithm adjusts the weights to make the loss function Minimum, that is: in,

[0059] It should be understood that the learned weights are output to the feature weight library, and if the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, the current weights of the multiple features are updated.

[0060] It should be understood that the feature weight learning module is timed-triggered. After being triggered, the server uses the feature instance set in the object database to learn the feature weight in the feature weight learning module.

[0061] Optionally, feature weight learning can be implemented using a gradient descent method or a Broyden-Fletcher-Goldfarb-Shannon (BFGS) algorithm.

[0062] The ID retrieval and generation system 310 is used to extract features from the hardware, software and runtime information sent by the object, form feature instances, and then perform corresponding retrieval and tracking. The system 310 also includes: a feature encoding module 3104, a hash encoding module 3101, a fast code retrieval module 3102 and a unique ID generation module 3103.

[0063] The feature encoding module 3104 can be used to encode different data types of object feature instances obtained by the server into the same data type, preferably, converting them into binary codes. The hash encoding module 3101 can be used to convert multiple codes of the same data type of multiple feature instances into hash codes based on feature weights, wherein the weight of each feature is positively correlated with the number of bits each feature occupies in the hash code. The fast code retrieval module 3102 can be used to query the code retrieval library based on the generated hash code, and return the object ID if the query is successful. The unique ID generation module 3103 can be used to generate a new object ID and return it, and the module ensures that the new object ID is a globally unique ID. Optionally, the unique object ID can be generated by recording the number of all known objects and using a self-increasing integer as the ID; or the current time of the system and the object features can be used as input, and a hash code calculated by a commonly used hash function can be used as the generated ID.

[0064] The data storage system 330 is responsible for storing the information that must be stored in the object tracking process, and mainly includes: an original feature library module 3201, a feature weight library module 3203, and a coding retrieval library module 3202. It should be understood that the database can be implemented using a key-value pair structured query language non-relational structured query language (NOSQL), an in-memory database, or a structured query language (SQL) table.

[0065] Optionally, the original feature library module 3201 can be used to store object IDs and features. One object ID can be associated with multiple sets of features. The original feature library module can be implemented using an SQL database. It should be understood that the system uses a unique ID to identify an object, called an object ID. The feature weight library module 3203 can be used to access feature weights. The feature weights can be numerical vectors and can be stored in an SQL table, where each column stores a one-dimensional feature weight. The encoding retrieval library module 3202 is used to store hash codes and object IDs, which can be stored in key-value pairs or tables.

[0066] Figure 4 4 is a flow chart of an object tracking method according to an embodiment of the present application. The method 400 includes steps S401 to S404, which are described in detail below.

[0067] S401, the server obtains the object instance and determines whether the object ID exists. Specifically, after obtaining the object instance, the server searches whether the feature instance contains the object ID. If the feature instance contains the object ID, it proceeds to step S4052, and the server returns a unique object ID. Optionally, the feature instance and its object ID can be added to the original feature library, that is, step S4011, and the tracking process ends; if the feature instance does not contain the object ID, it proceeds to step S402.

[0068] S402, the server converts the acquired features of different data types into codes of the same data type. Optionally, the server converts the features of different data types into binary codes.

[0069] S403, the server performs hash encoding on the encoding of the same data type obtained by feature encoding and obtains a hash code. Specifically, the server determines the number of bits occupied by each feature in the hash code based on the weights of multiple features. The weight of each feature is obtained through feature weight learning triggered by the server at a fixed time. Specifically, the server performs feature weight learning based on the feature instances in the original feature library, that is, step S4012. When the difference between the new weights of multiple features and the current weights is greater than a preset threshold, step S4013 is entered to update the current weights of multiple features. The server improves the adaptive tracking capability of the server by obtaining the hash code of the object based on the continuously updated weights of multiple features, thereby improving the tracking accuracy.

[0070] S404, the server queries the hash code in the code search library based on the above hash code and determines whether the object ID exists based on the hash code. Specifically, the server queries the hash code in the code search library. If the query result is not empty, the result is read to obtain the object ID (when there are multiple object IDs, conflicts are resolved according to the majority voting principle), that is, step S4042; if the query result is empty, a new object ID is generated by the unique ID generation module, that is, step S4041, and then the server binds the generated unique ID to the corresponding object to achieve object tracking.

[0071] Figure 5 Schematic diagram of an object tracking device according to an embodiment of the present application. Figure 5 As shown, the device 500 includes an acquisition module 510 , a feature encoding module 520 , a hash encoding module 530 and a tracking module 540 .

[0072] The acquisition module 510 is used to obtain a feature instance of an object, where the object is a client served by the server, and the feature instance includes multiple features of different data types, and the multiple features are used to indicate the hardware and / or software information of the client; a feature encoding module is used to convert the data types of the multiple features into the same data type to obtain the encoding of the feature instance; a hash encoding module is used to convert the encoding of the feature instance into a hash code according to the weights of the multiple features, wherein the weight of each feature is positively correlated to the number of bits it occupies in the hash code; a tracking module is used to track the object according to the hash code.

[0073] Optionally, the encoding of the feature instance is binary encoding.

[0074] Optionally, the encoding of the feature instance includes the encoding corresponding to each feature of the multiple features, and the hash coding module is specifically used to: determine the number of bits occupied by each feature in the hash code based on the weight of each feature and the number of bits of the hash code; sample the encoding corresponding to each feature based on the number of bits occupied by each feature in the hash code to obtain a sampling result corresponding to each feature, the number of bits of the sampling result corresponding to each feature is equal to the number of bits occupied by each feature in the hash code; and merge the sampling results of the multiple features to obtain the hash code.

[0075] Optionally, the hash coding module is further used to: randomly sample the code corresponding to each feature according to the number of bits occupied by each feature in the hash code.

[0076] Optionally, the device further includes: an updating module 550, configured to update the weights of the multiple features according to the multiple feature instances stored in the database.

[0077] Optionally, the updating module is also used to: determine new weights of the multiple features based on multiple feature instances stored in the database, so that the distance between feature instances of the same object is smaller than the distance between feature instances of different objects, wherein the distance between two feature instances of the multiple feature instances is the weighted sum of the distances of corresponding features in the two feature instances; if the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, update the current weights of the multiple features.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0079] The method in the embodiment of the present application, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The storage medium at least includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes. The above is only a specific implementation method of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field of the present application can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An object tracking method, characterized in that: include: The server obtains a feature instance of an object, the object being a client served by the server, the feature instance comprising a plurality of features of different data types, the plurality of features being used to indicate hardware and / or software information of the client; The server converts different data types of the multiple features into the same data type to obtain the encoding of the feature instance; The server converts the encoding of the feature instance into a hash code according to the weights of the multiple features, wherein the weight of each feature is positively correlated with the number of bits occupied by each feature in the hash code; The server tracks the object according to the hash code; The code of the feature instance includes the code corresponding to each feature in the plurality of features. The server converts the encoding of the feature instance into a hash code according to the weights of the multiple features, including: The server determines the number of bits occupied by each feature in the hash code according to the weight of each feature and the number of bits of the hash code; The server samples the code corresponding to each feature according to the number of bits occupied by each feature in the hash code to obtain a sampling result corresponding to each feature, where the number of bits of the sampling result corresponding to each feature is equal to the number of bits occupied by each feature in the hash code; The server combines the sampling results of the multiple features to obtain the hash code.

2. The method according to claim 1, characterized in that The encoding of the feature instance is binary encoding.

3. The method according to claim 1, characterized in that The server samples the code corresponding to each feature according to the number of bits occupied by each feature in the hash code, including: The server randomly samples the code corresponding to each feature according to the number of bits occupied by each feature in the hash code.

4. The method according to any one of claims 1 to 3, characterized in that Also includes: According to the plurality of feature instances stored in the database, the weights of the plurality of features are updated.

5. The method according to claim 4, characterized in that The updating of the weights of the multiple features according to the multiple feature instances stored in the database includes: Determining new weights of the multiple features based on the multiple feature instances stored in the database so that a distance between feature instances of the same object is smaller than a distance between feature instances of different objects, wherein a distance between two feature instances of the multiple feature instances is a weighted sum of distances of corresponding features in the two feature instances; If the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, the current weights of the multiple features are updated.

6. An object tracking device, characterized in that: include: An acquisition module, used to acquire a feature instance of an object, the object being a client served by a server, the feature instance comprising a plurality of features of different data types, the plurality of features being used to indicate hardware and / or software information of the client; A feature encoding module, used to convert different data types of the multiple features into the same data type to obtain the encoding of the feature instance; A hash coding module, used for converting the coding of the feature instance into a hash code according to the weights of the multiple features, wherein the weight of each feature is positively correlated with the number of bits occupied by each feature in the hash code; A tracking module, configured to track the object according to the hash code; The encoding of the feature instance includes the encoding corresponding to each feature of the multiple features, and the hash encoding module is specifically used for: Determine the number of bits occupied by each feature in the hash code according to the weight of each feature and the number of bits of the hash code; According to the number of bits occupied by each feature in the hash code, the code corresponding to each feature is sampled to obtain a sampling result corresponding to each feature, where the number of bits of the sampling result corresponding to each feature is equal to the number of bits occupied by each feature in the hash code; The sampling results of the multiple features are combined to obtain the hash code.

7. The device according to claim 6, characterized in that The encoding of the feature instance is binary encoding.

8. The device according to claim 6, characterized in that The hash coding module is also used to: According to the number of bits occupied by each feature in the hash code, the code corresponding to each feature is randomly sampled.

9. The device according to any one of claims 6 to 8, characterized in that The device also includes: The updating module is used to update the weights of the multiple features according to the multiple feature instances stored in the database.

10. The device according to claim 9, characterized in that The update module is also used for: Determining new weights of the multiple features based on the multiple feature instances stored in the database so that a distance between feature instances of the same object is smaller than a distance between feature instances of different objects, wherein a distance between two feature instances of the multiple feature instances is a weighted sum of distances of corresponding features in the two feature instances; If the difference between the new weights of the multiple features and the current weights of the multiple features is greater than a preset threshold, the current weights of the multiple features are updated.

11. A computer-readable storage medium, characterized in that: The invention comprises a computer program which, when executed on a computer device, causes a processing unit in the computer device to execute the method according to any one of claims 1 to 5.

Citation Information

Patent Citations

  • Self-adaptive query method based on Hash code weighting ranking

    CN104731882A

  • Image retrieval method based on variable-length depth hash learning

    CN105512273A

  • Feature extraction method, model generation method and malicious code detection method

    CN109308413A

  • Image Retrieval Method Based on Variable-Length Deep Hash Learning

    US20180276528A1

  • Systems and methods for identifying a returning web client

    US9165124B1