CN116032741A - Equipment identification method and device, electronic equipment and computer storage medium - Google Patents

Abstract

The embodiment of the application discloses a device identification method, a device, an electronic device and a computer storage medium, wherein the method comprises the following steps: acquiring equipment information to be identified; inputting the information of the equipment to be identified into a preset identification model, wherein the preset identification model comprises a single character matching sub-model and a character string matching sub-model; identifying the equipment information to be identified by utilizing the single character matching sub-model, determining a first identification result, identifying the equipment information to be identified by utilizing the character string matching sub-model, and determining a second identification result; and determining a device identification result according to the first identification result and the second identification result. In this way, the single character matching sub-model and the character string matching sub-model are used for carrying out combined identification on the equipment information to be identified, so that the accuracy of identifying the equipment information can be improved, and meanwhile, the identification efficiency is also improved.

Description

A device identification method, device, electronic device and computer storage medium

technical field

The present application relates to the technical field of communications, and in particular to a device identification method, device, electronic device and computer storage medium.

Background technique

With the rapid development of the Internet and the Internet of Things, the network has become an indispensable part of people's lives, and the emergence of more and more smart terminals poses challenges to the management of devices connected to the gateway.

In recent years, with the rapid development of triple play integration and broadband, intelligent gateways have gradually entered people's daily lives and become the key to open intelligent life. How to dig out the potential value of intelligent gateways is a question worthy of consideration and research. The smart gateway is the "heart" of the smart home. With the smart gateway, users can easily control the smart devices at home. Therefore, identifying which devices a home user associates with a smart gateway has become an urgent problem to be solved.

Contents of the invention

The present application provides a device identification method, device, electronic device, and computer storage medium, which can accurately identify devices associated with an intelligent gateway, and improve identification efficiency and accuracy.

The technical scheme of the present application is realized like this:

In the first aspect, the embodiment of the present application provides a device identification method, the method includes:

Obtain information about the device to be identified;

Inputting the information of the device to be identified into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a character string matching sub-model;

Using the single-character matching sub-model to identify the device information to be identified, determine a first identification result, and use the character string matching sub-model to identify the device information to be identified, and determine a second identification result;

A device identification result is determined according to the first identification result and the second identification result.

In the second aspect, an embodiment of the present application provides an apparatus for identifying equipment, the apparatus for identifying equipment includes an acquiring unit, an identifying unit, and a determining unit, wherein,

The acquiring unit is configured to acquire information about the device to be identified;

The recognition unit is configured to input the information of the device to be recognized into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a string matching sub-model; and utilizes the single-character matching sub-model Identifying the information of the device to be identified, determining a first identification result, and using the character string matching sub-model to identify the information of the device to be identified, and determining a second identification result;

The determining unit is configured to determine a device identification result according to the first identification result and the second identification result.

In a third aspect, an embodiment of the present application provides an electronic device, where the electronic device includes a memory and a processor, wherein,

said memory for storing a computer program capable of running on said processor;

The processor is configured to execute the device identification method according to the first aspect when running the computer program.

In a fourth aspect, an embodiment of the present application provides a computer storage medium, where a computer program is stored in the computer storage medium, and when the computer program is executed by at least one processor, the device identification method as described in the first aspect is implemented.

A device identification method, device, electronic device, and computer storage medium provided in the embodiments of the present application obtain the information of the device to be identified; input the information of the device to be identified into a preset recognition model, wherein the preset recognition model includes single-character matching The sub-model and the string matching sub-model; use the single-character matching sub-model to identify the information of the device to be recognized, determine the first recognition result, and use the string matching sub-model to identify the information of the device to be recognized, and determine the second recognition result; according to the first The first recognition result and the second recognition result determine the device recognition result. In this way, the combination of the single-character matching sub-model and the character string matching sub-model to recognize the device information to be recognized, that is, to jointly determine the device recognition result of the device information to be recognized according to the recognition results of the two sub-models, can improve the accuracy of device information recognition. and improve the recognition efficiency.

Description of drawings

FIG. 1 is a schematic flowchart of a device identification method provided in an embodiment of the present application;

FIG. 2 is a schematic diagram of a training process of a preset recognition model provided by an embodiment of the present application;

Fig. 3 is a detailed flowchart of a device identification method provided in an embodiment of the present application;

FIG. 4 is a schematic flowchart of another device identification method provided in the embodiment of the present application;

FIG. 5 is a schematic flowchart of another device identification method provided in the embodiment of the present application;

FIG. 6 is a schematic diagram of a training framework of a preset recognition model provided by an embodiment of the present application;

FIG. 7 is a schematic flowchart of another device identification method provided in the embodiment of the present application;

FIG. 8 is a schematic diagram of the composition and structure of a device identification device provided by an embodiment of the present application;

FIG. 9 is a schematic diagram of the composition and structure of an electronic device provided by an embodiment of the present application;

FIG. 10 is a schematic diagram of the composition and structure of another electronic device provided by the embodiment of the present application.

Detailed ways

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. It should be understood that the specific embodiments described here are only used to explain the related application, not to limit the application. It should also be noted that, for the convenience of description, only the parts related to the relevant application are shown in the drawings.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

In the following description, references to "some embodiments" describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or a different subset of all possible embodiments, and Can be combined with each other without conflict.

It should be pointed out that the term "first\second\third" involved in the embodiment of this application is only to distinguish similar objects, and does not represent a specific ordering of objects. Understandably, "first\second\third" Where permitted, the specific order or sequencing may be interchanged such that the embodiments of the application described herein can be practiced in sequences other than those illustrated or described herein.

At present, there are very limited methods for identifying devices connected to the smart gateway. Most of the solutions are to create a device model database, and then perform string matching between the device information reported by the smart gateway and the keywords in the device model database. Obtain information such as the model and type of the device. However, this method is only suitable for identification of devices connected to smart gateways with a small amount of data and relatively fixed device types. This method has at least the following disadvantages: (1) When the keywords of the device information reported by the smart gateway are incomplete or the order of keywords is incomplete If there is a slight change, the identification effect of the device will be greatly reduced; (2) rely too much on the device model database, when a new device not included in the database appears, it will be discarded as invalid device information; (3) cannot In-depth analysis of the equipment information that has not been successfully matched leads to a low utilization rate of the equipment information reported by the intelligent gateway; (4) the equipment model database cannot be automatically updated according to the equipment information reported by the intelligent gateway, resulting in manual screening and expansion of identifiable type of device model.

That is to say, the connected device information reported by the smart gateway has a large amount of data, many types of devices (for example: smart TV, smart phone, smart air conditioner, smart camera, etc.), and the information format is not uniform (for example: Honor_7A-6036f0010 , T1-821w-ed7e90066d182, 15557ac28e_bbk-H8A-6173d, etc.), making it difficult to accurately identify the model of the device connected to the smart gateway.

Based on this, the embodiment of the present application provides a device identification method. The basic idea of the method is: obtain the information of the device to be identified; input the information of the device to be identified into the preset recognition model, wherein the preset recognition model includes a single-character match The model and the character string matching sub-model; use the single-character matching sub-model to identify the information of the device to be recognized, determine the first recognition result, and use the character string matching sub-model to identify the information of the device to be recognized, and determine the second recognition result; according to the first The recognition result and the second recognition result determine the device recognition result. In this way, the combination of the single-character matching sub-model and the character string matching sub-model is used to identify the device information to be recognized, that is, the device recognition result of the device information to be recognized is jointly determined according to the recognition results of the two sub-models, thereby improving the accuracy of identifying the device information. Accuracy, but also improve the recognition efficiency.

Various embodiments of the present application will be described in detail below with reference to the accompanying drawings.

In an embodiment of the present application, refer to FIG. 1 , which shows a schematic flowchart of a device identification method provided in an embodiment of the present application. As shown in Figure 1, the method may include:

S101. Obtain information about a device to be identified.

It should be noted that the device identification method provided in the embodiment of the present application may be applied to a device identification device, or an electronic device integrated with the device. Here, the electronic device may be, for example, a computer, a smart phone, a tablet computer, a notebook computer, a palmtop computer, a personal digital assistant (Personal Digital Assistant, PDA), a navigation device, a server, etc., which are not specifically limited in this embodiment of the present application.

It should also be noted that the device identification method provided by the embodiment of the present application is mainly applied to accurately identify the information of the devices connected to the smart gateway in the home, so the device information to be identified is usually the device information reported by the smart gateway. It can be understood that the information of the device to be identified may also be obtained through other means, which is not specifically limited in this embodiment of the present application.

In addition, the information of the device to be identified is usually a series of messages that may include information related to the device, for example: wang-njaf-oppo-k9, 201-phicomm fwr706, netcore-mg-1200ac, dh-nvr2108hs-8p-s1- kdgj, hikvision ezviz cs-n1p-204 and so on.

S102. Input the information of the device to be identified into a preset recognition model, where the preset recognition model includes a single-character matching sub-model and a character string matching sub-model.

S103. Use the single-character matching sub-model to identify the information of the device to be recognized, determine a first recognition result, and use the character string matching sub-model to recognize the device information to be recognized, and determine a second recognition result.

It should be noted that, in this embodiment of the present application, the device information to be recognized is recognized by a preset recognition model to obtain the device recognition result, so as to determine the device label corresponding to the device information to be recognized, such as the device model, device brand, and device type, etc. Label Information.

Specifically, the preset recognition model may include a single-character matching sub-model and a character string matching sub-model. After inputting the device information to be recognized into the preset recognition model, the single-character matching sub-model and the character string matching The equipment information is identified to obtain a first identification result and a second identification result respectively, and then a final equipment identification result is determined according to the first identification result and the second identification result. Here, the single character matching sub-model and the character string matching sub-model may be a convolutional neural network structure or other neural network structures, which are not specifically limited in this embodiment of the present application.

Further, for the training process of the preset recognition model, refer to FIG. 2 , which shows a schematic diagram of a training flow of a preset recognition model provided by an embodiment of the present application. As shown in Figure 2, the process can include:

S201. Obtain a sample training set from a preset device model database.

It should be noted that when determining the preset recognition model, the sample training set used for model training is first obtained, and the sample training set is obtained from the preset device model database, that is, the sample training set is obtained from the preset model database It consists of the device information and the device label corresponding to the device information. The sample training set includes at least one device information and at least one device label corresponding to the device information, wherein the device label may at least include label information such as device model, device brand, and device type corresponding to the device information.

Further, for the preset device model database, in some embodiments, the method may also include:

Acquiring an original device information set, where the original device information set includes at least one piece of device information;

Perform data cleaning and data statistics on the equipment information in the original equipment information set, and obtain the intermediate equipment information set and the occurrence times of each equipment information in the intermediate equipment information set;

Comparing the number of occurrences of each device information in the intermediate device information set with a preset number of times threshold, and acquiring device information whose number of occurrences is greater than the preset number of times threshold;

Build a total database of device information based on the device information whose occurrence times are greater than the preset threshold;

The general database of equipment information is captured and processed to obtain a database of preset equipment models.

It should be noted that at least one piece of device information is included in the original device information set. Since the embodiment of the present application is mainly applied to the accurate identification of the devices attached to the smart gateway, when obtaining the original device information set, all the device information reported by the smart gateway can be obtained, and the device information can be collected based on the gateway plug-in.

In this way, after obtaining the original device information set, at least there may be some invalid device information, duplicate device information, etc. in the device information, so it is also possible to perform data cleaning and data statistics on the device information in the original device information Duplicated device information is eliminated, only valid device information is retained, and the number of occurrences of each valid device information is counted to obtain a set of intermediate device information and the number of occurrences of each device information in the set of intermediate device information.

In some specific embodiments, data cleaning and data statistics are performed on the device information in the original device information set to obtain the intermediate device information set and the number of occurrences of each device information in the intermediate device information set, which may include:

Eliminate invalid device information in the original device information set by using a preset expression to obtain at least one valid device information;

Counting the number of occurrences of each valid device information in the at least one valid device information, and performing deduplication processing on the at least one valid device information to obtain at least one target device information;

An intermediate device information set is constructed according to at least one target device information, and the occurrence times of each device information in the intermediate device information set are determined.

It should be noted that when using preset expressions to remove invalid device information, it usually refers to removing device information that is empty or anonymous; in addition, since device information usually includes Chinese characters, letters, and numbers, Other special characters usually mean that the device information is invalid, so they can also be removed. It should also be noted that with the development of technology, device information will also change accordingly, and the possibility of special characters in valid device information cannot be ruled out. Therefore, whether the device information containing special characters is invalid device information can also be combined with the actual The scene is determined.

It should also be noted that when removing invalid device information, the preset expression used may be a regular expression.

In this way, after the invalid device information in the original device information set is eliminated, at least one piece of valid device information is obtained. For the at least one valid device information, count the occurrence times of each valid device information, and perform deduplication processing on the valid device information, specifically, for multiple completely consistent valid device information, only keep one of them, so as to obtain at least A target device information and its corresponding occurrence count. In addition, it should be noted that when counting the number of occurrences of each valid device information, it can be realized through preset statistical tools. The preset statistical tools can include MapReduce, etc. Among them, MapReduce is a method for parallel computing of large-scale data sets. programming model.

In this way, the embodiment of the present application can obtain the cleaned and deduplicated at least one target device information, construct an intermediate device information set based on this, and determine the occurrence times of each device information in the intermediate device information set.

In this way, the number of occurrences of each device information in the intermediate device information set is compared with the preset number of times threshold, and the device information whose number of occurrences is greater than the preset number of times threshold is obtained, and the device information with the number of occurrences greater than the preset number of times threshold can be used. Construct a general database of equipment information.

It should also be noted that for the device information reported by the smart gateway, the greater the number of occurrences of the device information, the higher the frequency of use of the device. In this embodiment of the application, the number of occurrences in the intermediate device information set is greater than the preset threshold The device information is retained, and the setting of the preset times threshold can be determined in combination with actual usage requirements, which is not specifically limited in this embodiment of the present application.

It should also be noted that since the embodiment of the present application mainly focuses on the identification and analysis of the devices connected to the smart gateway in the home, the device whose device information appears only once or twice has a high probability that it does not belong to the member device of the family. If there are too many device information It will also increase the difficulty of training the preset recognition model for the first time, which will affect the recognition accuracy. Therefore, the embodiment of the present application only constructs the total device information database based on valid device information whose occurrence times are greater than the preset number threshold. In addition, if the application scenario is relatively simple and the amount of device information involved is small, all the device information can also be retained. However, since the device information is usually massive, it is necessary to perform a deletion operation.

In this way, after the general database of equipment information is determined, the general database of equipment information is captured, so as to obtain the database of preset equipment models. Specifically, in some embodiments, the general database of device information is captured to obtain the database of preset device models, which may include:

Obtaining the seed URL of each device information in the general device information database by means of a web crawler, and obtaining the web page information corresponding to each seed URL using web crawler technology;

Obtaining the URL of the seed uniform resource locator of each device information in the general database of device information respectively by means of a web crawler, and obtaining the webpage information corresponding to the seed URL of each device information by means of the web crawler;

From the web page information corresponding to the seed URL of each device information, determine the device label corresponding to each device information;

A preset device model database is constructed according to each piece of equipment information and the equipment label corresponding to each piece of equipment information.

That is to say, for any device information in the total device information database, firstly, the web crawler is used to automatically obtain the seed Uniform Resource Locator (URL) related to the device information from the query results of the search engine, wherein the search engine The query results for are determined after searching device information in a search engine.

Then, for the captured seed URL, the web crawler technology is used to obtain the corresponding web page information again, and the corresponding device tags are screened out from the web page information, such as tag information such as device model, device brand, and device type.

Finally, a preset device model database is constructed according to each device information and each corresponding device label.

Further, in some embodiments, the method may also include:

If one of the device information cannot successfully obtain the seed URL, then store one of the device information in the invalid device information database; and/or,

If one of the equipment information cannot successfully determine the equipment label, one of the equipment information is stored in the invalid equipment information database.

It should be noted that, for the device information in the general database of device information, if after the search engine searches, the relevant information of some device information cannot be found, that is, for these device information, the seed URL cannot be obtained successfully, or for some For device information, although the seed URL is obtained, the device label cannot be successfully obtained when crawling again. For these device information, it may be because the device information is invalid or the corresponding device is not fully listed, etc., so that the device label cannot be obtained temporarily on the webpage. , it is stored in the invalid device information database.

In this way, after the preset device model database is constructed, a sample training set can be obtained from the preset device model database for model training.

S202. Use the sample training set to train the single-character matching sub-network to obtain a single-character matching sub-model, and use the sample training set to train the string matching sub-network to obtain a string matching sub-model.

It should be noted that, by using the sample training set to train the single-character matching sub-network and the string matching sub-network, the single-character matching sub-model and the string matching sub-model can be respectively obtained. In addition, the single-character matching sub-network and string matching sub-network usually choose convolutional neural network.

Further, for the training of the single-character matching sub-model and the string matching sub-model, in some embodiments, the method may also include:

Create a device tag database according to the device tags corresponding to at least one piece of device information;

Correspondingly, using the sample training set to train the single-character matching sub-network to obtain a single-character matching sub-model, which may include:

Perform single-character segmentation on the device information in the sample training set to determine the single-character lexicon;

Perform single-character-level vector conversion on the single-character thesaurus and sample training set to obtain a single-character training vector set;

Input the single-character training vector set into the single-character matching sub-network, and use the device label database for supervised iterative training to obtain the character matching sub-model;

Use the sample training set to train the string matching sub-network to obtain the string matching sub-model, which can include:

Carry out string segmentation on the device information in the sample training set, and determine the string lexicon;

Perform string-level vector conversion on the string thesaurus and sample training set to obtain a string training vector set;

The string training vector set is input into the string matching sub-network, and the device label database is used for supervised iterative training to obtain the string matching sub-model.

It should be noted that, when creating the device label database, all device labels may be extracted from the sample training set and deduplicated, and then the device label database may be created according to the deduplicated device labels. That is to say, in the device label database, only one of each kind of device label is reserved.

It should also be noted that, in the embodiment of the present application, the device label database is used to perform supervised iterative training on the single-character matching sub-network and the string matching sub-network respectively to obtain the single-character matching sub-model and the string matching sub-model.

Specifically, single-character segmentation or string segmentation is performed on the device information in the sample training set, so as to obtain the single-character thesaurus and string thesaurus respectively; and then the single-character-level vector Convert to obtain a single-character training vector set, perform string-level vector conversion on the string lexicon and sample training set, and obtain a string training vector set.

For example, for the device information: yunuo-oppo-k9, the result of single-character segmentation is: y u nu o o p p o k 9, a total of eleven characters; the result of string segmentation is : yunuo oppo k9, a total of three words.

Input the single-character training vector set into the single-character matching sub-network for key feature extraction and classification, use the device label database for supervised iterative training, and obtain the character matching sub-model; input the string training vector set into the string matching sub-network for further Key feature extraction and classification, using the device label database for supervised iterative training, to obtain a string matching sub-model.

It should also be noted that when iteratively training the two sub-models, two independent loss functions can be used for the single-character matching sub-network and the string matching sub-network, and the two independent loss functions can be the same loss Functions, such as cross-entropy loss functions. In this way, after the classification iterative training is completed, the single-character matching sub-model and the string matching sub-model can be respectively obtained.

S203. Determine a preset recognition model according to the single-character matching sub-model and the character string matching sub-model.

It should be noted that the preset recognition model is obtained by combining the single-character matching sub-model and the character string matching sub-model.

That is to say, in the embodiment of the present application, the preset recognition model can be obtained according to the above steps S201-S203. In this way, the single-character matching sub-model and the character string matching sub-model in the preset recognition model can be used to respectively recognize the device information to be recognized, and obtain the first recognition result and the second recognition result respectively.

For the first recognition result, in some embodiments, using the single-character matching sub-model to identify the device information to be recognized, and determining the first recognition result may include:

Using the single-character matching sub-model to perform feature extraction and classification on the equipment information to be identified, and determine the first probability that the equipment information to be identified belongs to each equipment label;

The maximum value is selected from the first probability, and the device label corresponding to the maximum value is determined as the first recognition result.

For the second recognition result, in some embodiments, using the character string matching sub-model to identify the device information to be recognized, and determining the second recognition result may include:

Using the character string matching sub-model to perform feature extraction and classification on the equipment information to be identified, and determine the second probability that the equipment information to be identified belongs to each equipment label;

The maximum value is selected from the second probability, and the device label corresponding to the maximum value is determined as the second recognition result.

It should be noted that the single-character matching sub-model and the character string matching sub-model are used to extract and classify the key features of the device information to be recognized, so that the device information to be recognized can be obtained through the single-character matching sub-model. Probability, the device label corresponding to the maximum value in the first probability is determined as the first recognition result; the second probability that the device information to be recognized belongs to each device label can be obtained through the string matching sub-model, and the second probability in the second probability is The device label corresponding to the maximum value is determined as the second recognition result. In this way, the first recognition result and the second recognition result are respectively obtained.

In addition, in some embodiments, after obtaining the information of the device to be identified, the method may further include:

Matching the device information to be identified with the device information in the invalid device information database;

If the matching is successful, it is determined that the device information to be identified is invalid device information;

If the matching is unsuccessful, the step of inputting the information of the device to be identified into the preset identification model is performed.

It should be noted that, since the information of the device to be identified may be invalid, in this embodiment of the present application, after obtaining the information of the device to be identified, it may first match the information of the device to be identified with the device information in the invalid device information database, if If the matching is successful, it can be determined that the information of the device to be identified is invalid. At this time, there is no need to use the preset identification model for identification; identify. In this way, the calculation pressure of the equipment identification device can also be reduced, and invalid identification can be avoided, resulting in waste of resources.

It should also be noted that, in the embodiment of the present application, if the device information to be identified is invalid device information, the invalid device information database may also be updated with the device information to be identified.

S104. Determine a device identification result according to the first identification result and the second identification result.

It should be noted that, according to the first identification result and the second identification result, the final device identification result can be determined. Specifically, in some embodiments, determining the device identification result according to the first identification result and the second identification result may include:

If the first identification result is equal to the second identification result, determining the first identification result as the device identification result;

If the first recognition result is not equal to the second recognition result, then determine the first classification probability corresponding to the first recognition result and the second classification probability corresponding to the second recognition result, and according to the comparison result of the first classification probability and the first judgment threshold And a comparison result between the second classification probability and the second judgment threshold to determine the device identification result.

It should be noted that, when determining the device recognition result, if the first recognition result and the second recognition result are the same, the first recognition result (or the second recognition result) is directly determined as the final device recognition result.

If the first recognition result is different from the second recognition result, the first classification probability corresponding to the first recognition result (that is, the maximum value of the aforementioned first probability) and the second classification probability corresponding to the second recognition result (that is, the aforementioned first classification probability) are combined. The maximum value of the two probabilities), and the first judgment threshold and the second judgment threshold are further determined.

Wherein, the first judgment threshold is an optimal threshold for judging whether the recognition result of the single-character matching sub-model is reliable, and the second judgment threshold is an optimal threshold for judging whether the recognition result of the character string matching sub-model is reliable. In this embodiment of the present application, the same test set can be respectively input into the single-character matching sub-model and the character string matching sub-model, so as to obtain the first judgment threshold and the second judgment threshold respectively.

Specifically, in some embodiments, determining the device identification result according to the comparison result between the first classification probability and the first judgment threshold and the comparison result between the second classification probability and the second judgment threshold may include:

If the first classification probability is greater than or equal to the first judgment threshold, and the second classification probability is less than the second judgment threshold, then determining that the device identification result is the first identification result;

If the first classification probability is less than the first judgment threshold, and the second classification probability is greater than or equal to the second judgment threshold, then determining that the device identification result is the second identification result;

If the first classification probability is greater than or equal to the first judgment threshold and the second classification probability is greater than or equal to the second judgment threshold, or the first classification probability is less than the first judgment threshold and the second classification probability is less than the second judgment threshold, then the determination device The recognition result is not recognized.

It should be noted that, if the first recognition result is different from the second recognition result, if the first classification probability is greater than or equal to the first judgment threshold, and the second classification probability is smaller than the second judgment threshold, then the first recognition result Determined as a device recognition result. If the first classification probability is less than the first judgment threshold and the second classification probability is greater than or equal to the second judgment threshold, then the second recognition result is determined as the device recognition result. If the first classification probability is greater than or equal to the first judgment threshold, and the second classification probability is greater than or equal to the second judgment threshold; or, the first classification probability is less than the first judgment threshold, and the second classification probability is less than the second judgment threshold, then Make sure that the device recognition result is not recognized. In this way, the final device recognition result of the device information to be recognized is obtained.

Further, in some embodiments, after the device identification result is determined, the method may further include:

If the device identification result is unrecognized, the information of the device to be identified is added to the default device model database to obtain a new preset device model database;

The preset recognition model is updated to train with the new database of preset device models.

It should be noted that if the device recognition result is not recognized, it means that the device information to be recognized may be new device information. At this time, the device information to be recognized is added to the preset device model database to obtain a new preset The device model database, and use the new preset device model database to update and train the preset recognition model.

It should also be noted that when updating and training the preset recognition model, according to some of the steps described in the aforementioned steps S201-S203, the web crawler technology can be used to obtain the device label of the device information to be recognized and update the default device model database , and after updating the single-character thesaurus and the string thesaurus, the single-character matching sub-model and the string matching sub-model are updated and trained again to realize iterative updating of the preset recognition model.

It should also be noted that if the seed URL or device label is not successfully obtained during the web crawling process of the device information to be identified, as mentioned above, the device information to be identified can be added to the invalid device information database.

In addition, in the foregoing embodiments, the device information in the invalid device information library may be temporarily obtained from the network for its device label, and as the network information increases, the device information may be valid; for example, when the product is in the testing phase, Device-related information is not disclosed, and web crawlers cannot obtain its device labels. When the product is officially launched, device-related information is disclosed on the Internet, and the device information may be valid.

Therefore, in some embodiments, the method may also include:

Judging whether the device information in the invalid device information database is valid;

If it is determined that the device information in the invalid device information database is valid, the preset recognition model is updated and trained using the device information.

It should be noted that the embodiment of the present application may use web crawler technology to try to obtain the device label of the device information in the invalid device information database at intervals of a certain period of time or at a preset time. If it is successfully obtained, it means that the device information has become valid. The device information can be used to update and train the preset recognition model. The update method is the same as the above, and the corresponding device information is deleted from the invalid device information database, thereby avoiding misjudgment caused by untimely information updates.

This embodiment provides an identification method, by acquiring the information of the equipment to be identified; inputting the information of the equipment to be identified into the preset identification model, wherein the preset identification model includes a single-character matching sub-model and a character string matching sub-model; The matching sub-model identifies the information of the device to be recognized, determines the first recognition result, and uses the character string matching sub-model to recognize the information of the device to be recognized, and determines the second recognition result; according to the first recognition result and the second recognition result, determine the device recognition result. In this way, through the single character matching sub-model and the string matching sub-model to identify the equipment information to be identified, the equipment identification result of the equipment information to be identified is jointly determined according to the identification results of the two sub-models, thereby effectively improving the identification of equipment information. Accuracy, while also improving the recognition efficiency; even when the keywords of the device information are incomplete or the order of the keywords is changed, the device information can be accurately recognized, and the recognition effect is greatly improved; in addition, in the preset recognition model During the training and updating process, the device label corresponding to the device information is automatically obtained through the web crawler technology, which realizes the automatic and accurate acquisition of the device label and the automatic update of the preset recognition model. New device information can also be effectively identified.

In another embodiment of the present application, refer to FIG. 3 , which shows a detailed flowchart of a device identification method provided in an embodiment of the present application. As shown in Figure 3, the detailed process may include:

S301. The cleaning and statistics intelligent gateway reports device information.

It should be noted that, in the embodiment of the present application, the device information reported by the smart gateway may be used as raw data information for training the preset recognition model. Specifically, firstly, clean and count the device information reported by all smart gateways in preparation for model training.

Further, for the specific implementation process of step S301 , refer to FIG. 4 , which shows a detailed flowchart of another device identification method provided by the embodiment of the present application. As shown in Figure 4, the detailed process may include:

S301a. Store the device information reported by the smart gateway into the general device information database.

It should be noted that this step can create a total device information database based on Hadoop Distributed File System (Hadoop Distributed File System, HDFS), and store all device information reported by the smart gateway to the device General database of information. Since HDFS has advantages such as high reliability, high scalability, and high throughput, the embodiment of the present application utilizes HDFS to create a total device information database, but no specific limitation is made.

S301b. Use a regular expression to eliminate invalid equipment information in the general equipment information database.

It should be noted that regular expressions can be used in this step to eliminate special characters other than Chinese characters, letters, and numbers in the general equipment information database, as well as invalid equipment information in which the equipment information is empty or anonymous.

S301c. Use a preset statistical tool to count occurrence times of different device information, and complete deduplication.

It should be noted that in this step, a preset statistical tool (such as MapReduce) can be used to count the number of occurrences of different device information in the total device information database (that is, valid device information retained after eliminating invalid device information), and the total device information The device information in the database is deduplicated.

S301d. Filter effective and high-frequency device information according to the statistical times of the device information.

S301e. Update the general database of device information.

It should be noted that, according to the statistical results of the number of occurrences of the device information and the preset number threshold, the device information with the number of occurrences of the device information greater than the preset number threshold is used to update the total device information database, and only valid and High-frequency device information.

S302. The web crawler constructs/updates a preset device model database.

It should be noted that the preset device model database (also referred to as the device model database) may include: device information in the general device information database and its corresponding four types of data: device model, device brand, and device type.

Further, for the specific implementation process of step S302, refer to FIG. 5 , which shows a detailed flow chart of another device identification method provided by the embodiment of the present application. As shown in Figure 5, the detailed process may include:

S302a. The web crawler technology automatically obtains the seed URL of each device information in the general device information database.

It should be noted that, in this step, the web crawler technology is used to automatically obtain the relevant seed URL of the device information from the query results of the search engine.

S302b. Grab webpage information corresponding to the seed URL.

It should be noted that, in this step, the web crawler technology is used again to obtain the corresponding web page information for the crawled seed URL.

S302c. Filter out three types of information, namely device model, device brand, and device type, from the web page information.

It should be noted that three types of information (ie, device label, also referred to as device model information) such as device model, device brand, and device type are screened out from the webpage information.

S302d. If the device label is captured successfully, add the device information and its corresponding device model, device brand, and device type into the device model database.

S302e. If the capture of the device label is unsuccessful, add the device information into the invalid device information database.

It should be noted that, for the device information for which the device tag can be successfully captured, the device information and the corresponding device tag (that is, the device model, device brand, and device type corresponding to the device information) are added to the device model database.

For the device information for which the device label cannot be successfully captured, the device information is added to the invalid device information database.

For each piece of equipment information in the general equipment information database, the above steps are taken to crawl the network to complete the construction of the equipment model database, and save the equipment information that the network crawler cannot obtain to the equipment model, equipment brand and equipment type to the invalid equipment information database.

S303. Preset recognition model training based on convolutional neural network.

It should be noted that after obtaining the device model database, model training can be performed according to the device model database to obtain a preset recognition model based on convolutional neural network (also called precise device recognition model).

It should be noted that, for step S303, the network structure and training process of model training can be referred to FIG. 6, which shows a schematic diagram of a training framework of a preset recognition model provided by the embodiment of the present application. As shown in FIG. 6 , the framework may include: a training vector set acquisition part 601 , a subnetwork training part 602 and a model combination part 603 .

Among them, in the training vector set acquisition part 601, the main functions include respectively using the single-character segmentation method and the character string segmentation method to obtain the single-character lexicon (also called the letter lexicon) and the character string After the lexicon (also known as the word lexicon), the single-character lexicon and the character lexicon and the device model database are respectively subjected to single-character-level vector conversion and string-level vector conversion to obtain the training vector set M1 and the training vector Set N1; wherein, M1 represents a single-character training vector set, and N1 represents a character string training vector set.

In subnetwork training part 602, comprise single-character matching subnetwork and character string matching subnetwork; Wherein, single-character matching subnetwork comprises convolutional layer C1, convolutional layer C2, convolutional layer C3, fully connected layer F1, dropout Layer, activation layer R1, fully connected layer F2, cross-entropy loss layer S1; the string matching subnetwork includes convolutional layer D1, convolutional layer D2, convolutional layer D3, fully connected layer T1, dropout layer, activation layer K1, Fully connected layer T2, cross-entropy loss layer S2. The main functions of the sub-network training part 602 include inputting M1 and N1 into the single-character matching sub-network and string matching sub-network to extract key features of device information, and the two sub-networks use two independent cross-entropy loss functions and Unified device tags (also called device tags, from the device tag database) perform iterative supervised training. After completing the device tag classification iterative training, the single-character matching sub-model (also called the device recognition letter model) and string Matching submodel (also known as device recognition word model).

In the model combining part 603, the main function includes combining the single-character matching sub-model and the character string matching sub-model to obtain a preset recognition model.

It should be noted that the classification training based on the convolutional neural network mainly adopts a network structure in which the single-character matching subnetwork and the string matching subnetwork are parallel. The constructed device model database extracts the device model, device brand, and device type corresponding to each device information in the device model database and performs a unified deduplication operation, using the deduplicated device model, device brand, and device type to construct Device label database. Then, using the single-character segmentation method and the string segmentation method, the automatic creation and update of the single-character thesaurus and the string thesaurus are realized based on the sample training set. Vector conversion and string-level vector conversion to obtain training vector set M1 and training vector set N1, respectively input M1 and N1 into the single-character matching sub-network and string matching sub-network for key feature extraction of device information, the two sub-networks use Two independent cross-entropy loss functions and a unified device label are used for iterative supervised training. After the classification iterative training is completed, the single-character matching sub-model and the string matching sub-model are respectively obtained, and the combination of the two sub-models is used as the preset recognition model.

S304. Use of a preset recognition model.

It should be noted that, first, the same test set is input into the two sub-models, and the optimal thresholds for the identification of the two sub-models (ie, the first judgment threshold and the second judgment threshold in the foregoing embodiment) are respectively obtained, and the information of the equipment to be identified is respectively Input it into the single-character matching sub-model and the string matching sub-model to extract and classify the key features of the device information, and arrange the probability of the device information belonging to each device label from large to small, and take the classification in the two sub-models respectively The device label with the highest probability is used as the recognition result of the sub-model. The recognition results of the two sub-models are compared with their optimal thresholds. According to the set threshold discrimination rules, the device recognition result (also called the precise device recognition result) is finally output.

It is determined that the device identification result follows the following threshold discrimination rules, see formula (1) for details:

In formula (1), T _L represents the first judgment threshold, T _W represents the second judgment threshold, _{and RL} represents the first recognition result of the device information to be recognized in the single-character matching sub-model, which belongs to the first recognition result of _RL . The classification probability is _PL , and R _W represents the second recognition result of the device information to be recognized in the string matching sub-model, and the second classification probability belonging to R _W is P _W , and R represents the device recognition result.

S305. Automatically identify and update the devices attached to the smart gateway.

It should be noted that, for the specific implementation process of step S305, reference may be made to FIG. 7, which shows a detailed flowchart of another device identification method provided by the embodiment of the present application. As shown in Figure 7, the detailed process may include:

S305a. The intelligent gateway reports the information of the device to be identified.

S305b. Determine whether the device information to be identified exists in the invalid device information database.

It should be noted that, if the judgment result is yes, execute step S305i; if the judgment result is no, execute step S305c.

S305c. Input the information of the equipment to be identified into a preset identification model for identification.

S305d. Determine whether the recognition result is unrecognized.

It should be noted that, if the judgment result is yes, execute step S305e; if the judgment result is no, execute step S305j.

S305e. Output the information of the device to be identified.

S305f. The web crawler updates the device model database.

S305g, updating the single-character thesaurus and the character string thesaurus.

S305h, iteratively training the preset recognition model based on the convolutional neural network.

It should be noted that if the device label of the device information to be identified can be successfully obtained by using a web crawler, then the device model library is updated using the device information and the corresponding device label, and the single-character thesaurus and string thesaurus are further updated. Finally, based on The preset recognition model of the convolutional neural network is iteratively trained to complete the iterative update of the preset recognition model.

S305i. Outputting that the device information to be identified is invalid device information.

It should be noted that if the device information to be identified exists in the invalid device information database, it is directly determined that the device information to be identified is invalid device information.

S305j. Output the identified device model, device brand and device type.

It should be noted that, in the identification process of the equipment information to be identified, firstly, the equipment information reported by the smart gateway is matched with the equipment information in the invalid equipment information database, and if the matching is successful, the identification ends, and the output: the reported information is Invalid device information, at the same time, the database of invalid device information can also be updated; if the matching is unsuccessful, continue the process of the aforementioned step S304, and use the preset recognition model to identify the device information to be recognized. When the recognition obtained by the preset recognition model When the result is "unrecognized", follow the above steps S302 and S303 in sequence to add the unrecognized newly added device information into the device model database, complete automatic crawling of device tags and update the preset recognition model.

To sum up, the embodiment of the present application provides a device identification method, and the implementation of the method can be divided into five steps: S301 cleaning and counting the device information reported by the intelligent gateway, S302 constructing/updating the device model database by a web crawler, and S303 based on volume The training of the preset recognition model of the product neural network, the use of the preset recognition model in S304, and the automatic recognition and update of the equipment connected to the new intelligent gateway in S305.

Step S301 is briefly described as follows: create a general device information database based on the Hadoop distributed file system, and store the device information reported by all intelligent gateways in the general device information database.

Use regular expressions to eliminate special characters other than Chinese characters, letters, and numbers in the general database of equipment information, as well as invalid equipment information where the equipment information is empty or anonymous.

Use MapReduce to count the number of occurrences of different device information in the total device information database, and deduplicate the device information.

According to the above statistical results, combined with the preset frequency threshold, the original device information total database is updated with the device information whose occurrence frequency is greater than the preset frequency threshold, and only effective and high-frequency device information is retained. The flow chart of this process is shown in Figure 4.

Step S302 is briefly described as follows: the device model database includes: device information in the device information general database and its corresponding device model, device brand, and device type.

First, the web crawler technology is used to automatically obtain the relevant seed URL of the device information from the query results of the search engine.

Then use the web crawler technology again to obtain the corresponding web page information on the captured seed URL.

Finally, three types of information, such as device model, device brand, and device type, are screened out from the web page information.

For each valid device information reported by the smart gateway, take the above steps to crawl the network to complete the construction of the device model database, and save the device information reported by the smart gateway that the crawler cannot obtain to the device model, brand and type to the invalid device information database. The process is shown in Figure 5 in detail.

Step S303 is briefly described as follows: The device model classification training based on the convolutional neural network mainly adopts a parallel network structure of a single-character matching subnetwork and a character string matching subnetwork. The network structure is shown in Figure X, and the sample training set used for training is The device model database built by the web crawler extracts the device model, device brand, and device type corresponding to each device information in the device model database, and then performs a unified deduplication operation, and uses the deduplicated device model and device brand , device type to build a device label database, and then use the single-character segmentation method and string segmentation method to automatically create and update the single-character thesaurus and string thesaurus according to the training data set. The single-character thesaurus and the string thesaurus are respectively With the device model database, the training vector set M1 and the training vector set N1 are obtained by using single-character-level vector conversion and string-level vector conversion, and M1 and N1 are respectively input into the single-character matching sub-network and string matching sub-network for device information For key feature extraction, the two sub-networks use two independent cross-entropy loss functions and unified device labels for iterative supervised training. After completing the iterative training for device model classification, a single-character matching sub-model and a string The combination of two sub-models is used as the preset recognition model. The process is shown in Figure 6 in detail.

Step S304 is briefly described as follows: first, input the same test set into the two sub-models, respectively obtain the optimal thresholds for device identification of the two sub-models, and input the information of the device to be identified into the single-character matching sub-model and the character string matching sub-model respectively Extract and classify the key features of equipment information, and arrange the probability that the equipment information belongs to each equipment label from large to small, respectively take the label with the highest classification probability in the two models as the recognition result of the model, and combine the two The recognition result of the model is compared with its threshold, and according to the set threshold discrimination rules, the accurate recognition result of the device is finally output. Determine the threshold discrimination rules followed by the device recognition results, see formula (1)

Step S305 is briefly described as follows: first, match the newly added device information to be identified with the information in the invalid device information database, and if the matching is successful, then end the identification, and output: the reported information is invalid device information; if the matching is unsuccessful, continue to the step S304, when the result obtained through the preset recognition model is "unrecognized", proceed to step S302 and step S303 in sequence, add the unrecognized device information to the device model database, complete automatic crawling of device labels and update the preset recognition Model. The process is shown in Figure 7 in detail.

This embodiment provides a device identification method. The implementation of the foregoing embodiments is described in detail through the above-mentioned embodiments. For example, the combination of big data analysis technology, web crawler and convolutional neural network classification algorithm has realized the accurate identification of multi-dimensional information such as the type, brand and model of the connected equipment. Use web crawler technology to automatically capture the device model, device brand, and device type corresponding to the device information reported by the smart gateway in the general device information database, and build a preset device model database and an invalid device information database; use the single character segmentation method The single-character thesaurus and the string thesaurus are respectively constructed with the device model database using the single-character-level vector conversion and the string-level vector conversion to obtain the corresponding training vector sets. Input the two training vector sets into the corresponding single-character matching sub-network and string matching sub-network to extract key features of device information. The two sub-networks use two independent cross-entropy loss functions and unified device labels for iteration. With supervised training, the single-character matching sub-model and string matching sub-model are finally obtained, and the combination of the two sub-models is used as the preset recognition model; when the smart gateway reports new device information to be recognized, the preset recognition model is used to identify the device to be recognized Extract key features and classify and identify the information. When the preset recognition model cannot be accurately classified, the web crawler is automatically started to update the device model database, and the updated device model database is used to conduct preset recognition model training again to obtain updated accurate device recognition. The model not only realizes the automatic and accurate identification of the devices attached to the smart gateway, but also realizes the automatic identification of new device labels and updates the device model database while improving the recognition accuracy of the devices connected to the smart gateway.

In yet another embodiment of the present application, refer to FIG. 8 , which shows a schematic diagram of the composition and structure of a device identification device 80 provided in the embodiment of the present application. As shown in FIG. 8, the apparatus for identifying equipment 80 may include an acquiring unit 801, an identifying unit 802, and a determining unit 803, wherein,

The obtaining unit 801 is configured to obtain the information of the device to be identified;

The recognition unit 802 is configured to input the information of the device to be recognized into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a string matching sub-model; and uses the single-character matching sub-model to pair Identifying the equipment information to be identified, determining a first identification result, and using the character string matching sub-model to identify the equipment information to be identified, and determining a second identification result;

The determining unit 803 is configured to determine a device identification result according to the first identification result and the second identification result.

In some embodiments, the determining unit 803 is further configured to match the device information to be identified with the device information in the invalid device information database; and if the matching is successful, determine that the device information to be identified is invalid device information; And if the matching is not successful, execute the step of inputting the information of the device to be identified into a preset identification model.

In some embodiments, referring to FIG. 8 , the apparatus for identifying equipment 80 may further include a training unit 804 configured to acquire a sample training set from a preset equipment model database; wherein, the sample training set includes at least one piece of equipment information and A device tag corresponding to each of the at least one piece of device information, the device tag at least including a device model, a device brand, and a device type; and using the sample training set to train a single-character matching subnetwork to obtain the single-character matching subnetwork model, and use the sample training set to train the character string matching subnetwork to obtain the character string matching submodel; and determine the preset according to the single character matching submodel and the character string matching submodel Identify the model.

In some embodiments, the training unit 804 is further configured to acquire an original device information set, the original device information set including at least one device information; and perform data cleaning and data statistics on the device information in the original device information set, Obtaining the set of intermediate device information and the number of occurrences of each piece of equipment information in the set of intermediate device information; the device information with the preset number of times threshold; and according to the device information with the number of occurrences greater than the preset number of times threshold, construct a total device information database; and perform capture processing on the device information total database to obtain the preset Device model database.

In some embodiments, the training unit 804 is specifically configured to use a preset expression to eliminate invalid device information in the original device information set to obtain at least one valid device information; and count the at least one valid device information the number of occurrences of each valid device information, and perform deduplication processing on the at least one valid device information to obtain at least one target device information; and construct the intermediate device information set according to the at least one target device information, and determine the The number of occurrences of each device information in the above intermediate device information set.

In some embodiments, the training unit 804 is also specifically configured to respectively obtain the seed URL of each device information in the general device information database by using a web crawler, and obtain the seed URL of each device information by using the web crawler The web page information corresponding to the URL; and from the web page information corresponding to the seed URL of each device information, determine the device tag corresponding to each device information; and according to each device information and the device tag corresponding to each device information, construct The preset device model database.

In some embodiments, the training unit 804 is further configured to store one of the device information in the invalid device information database if one of the device information cannot be successfully obtained to the seed URL; and/or, if one of the device information cannot be obtained If the device label is determined successfully, the information of one of the devices is stored in the invalid device information database.

In some embodiments, the training unit 804 is further specifically configured to create a device tag database according to the device tags corresponding to the at least one piece of device information; and perform single-character segmentation on the device information in the sample training set to determine the single-character Thesaurus; performing single-character level vector conversion to the single-character thesaurus and the sample training set to obtain a single-character training vector set; inputting the single-character training vector set into the single-character matching subnetwork, using the performing supervised iterative training on the device label database to obtain the character matching sub-model; and performing character string segmentation on the device information in the sample training set to determine a character string lexicon; The set is converted to string-level vectors to obtain a string training vector set; the string training vector set is input into the string matching subnetwork, and the device label database is used for supervised iterative training to obtain the string matching submodel.

In some embodiments, the determining unit 803 is specifically configured to use the single-character matching sub-model to perform feature extraction and classification on the device information to be recognized, and determine the first probability that the device information to be recognized belongs to each device label and selecting a maximum value from the first probability, and determining the device label corresponding to the maximum value as the first identification result.

In some embodiments, the determining unit 803 is specifically configured to use the character string matching sub-model to perform feature extraction and classification on the device information to be identified, and determine the second probability that the device information to be identified belongs to each device label and selecting a maximum value from the second probability, and determining the device label corresponding to the maximum value as the second identification result.

In some embodiments, the determining unit 803 is specifically configured to determine the first recognition result as the device recognition result if the first recognition result is equal to the second recognition result; and if the first If the recognition result is not equal to the second recognition result, the first classification probability corresponding to the first recognition result and the second classification probability corresponding to the second recognition result are determined, and according to the first classification probability and the first classification probability A comparison result of the judgment threshold and a comparison result of the second classification probability and the second judgment threshold determine the device identification result.

In some embodiments, the determining unit 803 is specifically configured to determine that the device The identification result is the first identification result; and if the first classification probability is less than the first judgment threshold and the second classification probability is greater than or equal to the second judgment threshold, then determine the device identification result is the second identification result; and if the first classification probability is greater than or equal to the first judgment threshold and the second classification probability is greater than or equal to the second judgment threshold, or, the first classification probability If it is smaller than the first judgment threshold and the second classification probability is smaller than the second judgment threshold, it is determined that the device recognition result is not recognized.

In some embodiments, referring to FIG. 8 , the device identification device 80 may further include an update unit 805 configured to add the information of the device to be identified to the preset device model database if the device identification result is unrecognized. , to obtain a new default device model database; and use the new preset device model database to update and train the preset recognition model.

In some embodiments, the update unit 805 is further configured to judge whether the device information in the invalid device information database is valid; and if it is judged that the device information in the invalid device information database is valid, use the device information to perform update training on the preset recognition model .

It can be understood that, in this embodiment, a "unit" may be a part of a circuit, a part of a processor, a part of a program or software, etc., of course it may also be a module, or it may be non-modular. Moreover, each component in this embodiment may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software function modules.

If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or It is said that the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product, the computer software product is stored in a storage medium, and includes several instructions to make a computer device (which can It is a personal computer, a server, or a network device, etc.) or a processor (processor) that executes all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read only memory (Read Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other various media that can store program codes.

Therefore, this embodiment provides a computer storage medium, where the computer storage medium stores a computer program, and when the computer program is executed by at least one processor, the device identification method described in any one of the preceding embodiments is implemented.

Based on the above-mentioned composition of an apparatus for identifying an apparatus 80 and a computer storage medium, refer to FIG. 9 , which shows a schematic diagram of the composition and structure of an electronic apparatus 90 provided by an embodiment of the present application. As shown in FIG. 9 , an electronic device 90 may include: a communication interface 901 , a memory 902 and a processor 903 ; each component is coupled together through a bus system 904 . It can be understood that the bus system 904 is used to realize connection and communication between these components. In addition to the data bus, the bus system 904 also includes a power bus, a control bus and a status signal bus. However, for clarity of illustration, the various buses are labeled as bus system 904 in FIG. 9 . Among them, the communication interface 901 is used for receiving and sending signals during the process of sending and receiving information with other external network elements;

Memory 902, used to store computer programs that can run on the processor 903;

The processor 903 is configured to, when running the computer program, execute:

Obtain information about the device to be identified;

Inputting the information of the device to be identified into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a character string matching sub-model;

Using the single-character matching sub-model to identify the device information to be identified, determine a first identification result, and use the character string matching sub-model to identify the device information to be identified, and determine a second identification result;

A device identification result is determined according to the first identification result and the second identification result.

It can be understood that the memory 902 in the embodiment of the present application may be a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memories. Among them, the non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electronically programmable Erase Programmable Read-Only Memory (Electrically EPROM, EEPROM) or Flash. The volatile memory can be Random Access Memory (RAM), which acts as external cache memory. By way of illustration and not limitation, many forms of RAM are available, such as Static Random Access Memory (Static RAM, SRAM), Dynamic Random Access Memory (Dynamic RAM, DRAM), Synchronous Dynamic Random Access Memory (Synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (Double Data RateSDRAM, DDRSDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous chain dynamic random access memory (Synchronous link DRAM, SLDRAM) And Direct Memory Bus Random Access Memory (Direct Rambus RAM, DRRAM). Memory 902 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

The processor 903 may be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method may be completed by an integrated logic circuit of hardware in the processor 903 or instructions in the form of software. The above-mentioned processor 903 may be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Various methods, steps, and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. The steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory 902, and the processor 903 reads the information in the memory 902, and completes the steps of the above method in combination with its hardware.

It should be understood that the embodiments described herein may be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (Application Specific Integrated Circuits, ASIC), digital signal processor (Digital Signal Processing, DSP), digital signal processing device (DSP Device, DSPD), programmable logic Device (Programmable Logic Device, PLD), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA), general-purpose processor, controller, microcontroller, microprocessor, other electronic units for performing the functions described in this application or a combination thereof.

For a software implementation, the techniques described herein can be implemented through modules (eg, procedures, functions, and so on) that perform the functions described herein. Software codes can be stored in memory and executed by a processor. Memory can be implemented within the processor or external to the processor.

Optionally, as another embodiment, the processor 903 is further configured to execute the device identification method described in any one of the foregoing embodiments when running the computer program.

Based on the composition and hardware structure schematic diagram of the device identification device 80 described above, refer to FIG. 10 , which shows a schematic composition structure diagram of another electronic device 90 provided by the embodiment of the present application. As shown in FIG. 10 , the electronic device 90 includes at least the device identification device 80 described in any one of the foregoing embodiments.

For the electronic device 90, since the single-character matching sub-model and the character string matching sub-model are combined to recognize the device information to be recognized, the recognition result of the device information to be recognized is jointly determined according to the recognition results of the two sub-models, thereby effectively improving the recognition of the device information. The accuracy of equipment information identification is improved, and the identification efficiency is also improved.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the protection scope of the present application.

It should be noted that in this application, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements , but also includes other elements not expressly listed, or also includes elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element.

The serial numbers of the above embodiments of the present application are for description only, and do not represent the advantages and disadvantages of the embodiments.

The methods disclosed in several method embodiments provided in this application can be combined arbitrarily to obtain new method embodiments under the condition of no conflict.

The features disclosed in several product embodiments provided in this application can be combined arbitrarily without conflict to obtain new product embodiments.

The features disclosed in several method or device embodiments provided in this application can be combined arbitrarily without conflict to obtain new method embodiments or device embodiments.

The above is only a specific implementation of the application, but the scope of protection of the application is not limited thereto. Anyone familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the application. Should be covered within the protection scope of this application. Therefore, the protection scope of the present application should be determined by the protection scope of the claims.

Claims (16)

1. A device identification method, characterized in that the method comprises: Obtain information about the device to be identified; Inputting the information of the device to be identified into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a character string matching sub-model; Using the single-character matching sub-model to identify the device information to be identified, determine a first identification result, and use the character string matching sub-model to identify the device information to be identified, and determine a second identification result; A device identification result is determined according to the first identification result and the second identification result. 2. The method according to claim 1, characterized in that, after the acquisition of the device information to be identified, the method further comprises: matching the device information to be identified with the device information in the invalid device information database; If the matching is successful, determining that the device information to be identified is invalid device information; If the matching is not successful, the step of inputting the information of the device to be identified into a preset identification model is performed. 3. The method according to claim 1, characterized in that the method further comprises: Obtain a sample training set from a preset device model database; wherein, the sample training set includes at least one device information and a device tag corresponding to each of the at least one device information, and the device tag includes at least a device model, a device brand and Equipment type; Using the sample training set to train the single-character matching sub-network to obtain the single-character matching sub-model, and using the sample training set to train the character string matching sub-network to obtain the character string matching sub-model; The preset recognition model is determined according to the single-character matching sub-model and the character string matching sub-model. 4. method according to claim 3, is characterized in that, described method also comprises: Acquire an original device information set, where the original device information set includes at least one piece of device information; Perform data cleaning and data statistics on the device information in the original device information set, and obtain the intermediate device information set and the occurrence times of each device information in the intermediate device information set; Comparing the number of occurrences of each piece of device information in the intermediate device information set with a preset number of times threshold, and acquiring device information whose number of occurrences is greater than the preset number of times threshold; Constructing a total database of device information according to the device information whose occurrence times are greater than the preset number of times threshold; The general database of equipment information is captured to obtain the database of preset equipment models. 5. The method according to claim 4, characterized in that, performing data cleaning and data statistics on the device information in the original device information set to obtain the intermediate device information set and each of the intermediate device information sets Occurrences of device information, including: Eliminate invalid device information in the original device information set by using a preset expression to obtain at least one valid device information; Counting the number of occurrences of each valid device information in the at least one valid device information, and performing deduplication processing on the at least one valid device information to obtain at least one target device information; Constructing the intermediate device information set according to the at least one target device information, and determining the occurrence times of each device information in the intermediate device information set. 6. The method according to claim 4, characterized in that, the capturing of the general database of equipment information to obtain the database of preset equipment models comprises: Obtaining the URL of the seed uniform resource locator of each device information in the general database of device information respectively by means of a web crawler, and obtaining the webpage information corresponding to the seed URL of each device information by means of the web crawler; From the webpage information corresponding to the seed URL of each device information, determine the device label corresponding to each device information; The preset device model database is constructed according to each device information and the device label corresponding to each device information. 7. The method according to claim 6, further comprising: If one of the device information cannot successfully obtain the seed URL, then store the one of the device information in the invalid device information database; and/or, If one of the equipment information cannot successfully determine the equipment label, the one of the equipment information is stored in the invalid equipment information database. 8. The method according to claim 3, further comprising: Create a device tag database according to the device tags corresponding to the at least one piece of device information; Correspondingly, said using said sample training set to train the single-character matching sub-network to obtain said single-character matching sub-model, including: performing single-character segmentation on the device information in the sample training set to determine a single-character lexicon; performing single-character level vector conversion on the single-character thesaurus and the sample training set to obtain a single-character training vector set; Inputting the single-character training vector set into the single-character matching sub-network, using the device label database to perform supervised iterative training to obtain the character matching sub-model; Said using said sample training set to train the character string matching sub-network to obtain said character string matching sub-model, including: performing character string segmentation on the equipment information in the sample training set, and determining the character string lexicon; performing string-level vector conversion on the string thesaurus and the sample training set to obtain a string training vector set; The string training vector set is input into the string matching subnetwork, and the device label database is used for supervised iterative training to obtain the string matching submodel. 9. The method according to claim 1, wherein the identifying the device information to be identified by using the single-character matching sub-model, and determining the first identification result comprises: Using the single-character matching sub-model to perform feature extraction and classification on the equipment information to be identified, and determine a first probability that the equipment information to be identified belongs to each equipment label; Selecting a maximum value from the first probability, and determining the device label corresponding to the maximum value as the first identification result. 10. The method according to claim 1, wherein the identifying the device information to be identified by using the character string matching sub-model to determine the second identification result comprises: Using the character string matching sub-model to perform feature extraction and classification on the device information to be identified, and determine a second probability that the device information to be identified belongs to each device label; Selecting a maximum value from the second probability, and determining the device label corresponding to the maximum value as the second identification result. 11. The method according to claim 1, wherein the determining the device identification result according to the first identification result and the second identification result comprises: If the first identification result is equal to the second identification result, determining the first identification result as the device identification result; If the first recognition result is not equal to the second recognition result, then determine the first classification probability corresponding to the first recognition result and the second classification probability corresponding to the second recognition result, and according to the first A comparison result of the classification probability with the first judgment threshold and a comparison result of the second classification probability with the second judgment threshold determine the device identification result. 12. The method according to claim 11, characterized in that, according to the comparison result between the first classification probability and the first judgment threshold and the comparison result between the second classification probability and the second judgment threshold, determine the The identification results of the above devices include: If the first classification probability is greater than or equal to the first judgment threshold, and the second classification probability is smaller than the second judgment threshold, then determine that the device identification result is the first identification result; If the first classification probability is less than the first judgment threshold, and the second classification probability is greater than or equal to the second judgment threshold, then determining that the device identification result is the second identification result; If the first classification probability is greater than or equal to the first judgment threshold and the second classification probability is greater than or equal to the second judgment threshold, or, the first classification probability is less than the first judgment threshold and the If the second classification probability is less than the second judgment threshold, it is determined that the device recognition result is not recognized. 13. The method according to any one of claims 1 to 12, characterized in that, after the device identification result is determined, the method further comprises: If the device identification result is unrecognized, adding the information of the device to be identified to the preset device model database to obtain a new preset device model database; The preset recognition model is updated and trained by using the new preset device model database. 14. A device identification device, characterized in that the device identification device includes an acquisition unit, an identification unit and a determination unit, wherein, The acquiring unit is configured to acquire information about the device to be identified; The recognition unit is configured to input the information of the device to be recognized into a preset recognition model, wherein the preset recognition model includes a single-character matching sub-model and a string matching sub-model; and utilizes the single-character matching sub-model Identifying the information of the device to be identified, determining a first identification result, and using the character string matching sub-model to identify the information of the device to be identified, and determining a second identification result; The determining unit is configured to determine a device identification result according to the first identification result and the second identification result. 15. An electronic device, characterized in that the electronic device comprises a memory and a processor, wherein, said memory for storing a computer program capable of running on said processor; The processor is configured to execute the device identification method according to any one of claims 1 to 13 when running the computer program. 16. A computer storage medium, wherein the computer storage medium stores a computer program, and when the computer program is executed by at least one processor, the device identification method according to any one of claims 1 to 13 is implemented.

Images