Message pushing method, device and computer storage medium

Abstract

The application provides a message pushing method, device and computer storage medium. The message pushing method comprises the following steps: when a target event message reported by a target device is received, performing scene understanding on the target event message, determining target scene data corresponding to the target event message and a risk level; in the case that the risk level is less than a preset risk threshold, comparing the target event message with historical event messages based on the target scene data to obtain a comparison result; and in the case that the comparison result indicates that the target event message and the historical event messages are different event messages, pushing the target event message. Through the message pushing method, the accuracy of message filtering is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of Internet of Things, and in particular to a message pushing method, device and computer storage medium. BACKGROUND

[0002] With the rapid development of Internet of Things technology, Internet of Things devices have been widely applied in smart home, industrial monitoring, security and alarm, environmental sensing and other fields. Internet of Things devices can collect surrounding environment data in real time, identify abnormal intrusion, environmental parameter exceeding and other abnormal conditions, and push reminder messages to user terminals through the network, helping users to timely grasp key situations and take measures, greatly improving the convenience and safety of life and work.

[0003] In the identification process, due to device false triggering, repeated data collection and other reasons, the phenomenon of too frequent message reporting or a large number of the same reminder messages being pushed in a short time occurs, which interferes with the user. After the message filtering mechanism is used to screen the messages reported by the Internet of Things device, the messages are pushed to the user terminal.

[0004] The message filtering mechanism in related technologies is mainly based on message text, for example, by extracting keywords in the message and matching them with preset filtering rules, or calculating the text similarity of the previous and next two messages to realize repeated message deduplication. However, special situations in different scenarios may be described by different expressions of text, and it is difficult to accurately identify the core semantics of the message based on keywords or text similarity alone, which is easy to misjudge different expression messages of the same type of key event as non-repeated messages, resulting in low accuracy of message filtering. SUMMARY

[0005] To solve the above technical problems, the present application provides a message pushing method, device and computer storage medium.

[0006] To solve the above technical problems, the present application provides a message pushing method, which comprises:

[0007] When a target event message reported by a target device is received, the target event message is understood in a scene to determine target scene data corresponding to the target event message and a risk level;

[0008] In the case where the risk level is less than a preset risk threshold, the target event message is compared with a historical event message based on the target scene data to obtain a comparison result;

[0009] When the comparison result indicates that the target event message and the historical event message are different event messages, the target event message is pushed.

[0010] The target scene data includes a target scene label and a target scene description.

[0011] The target event message is compared with the historical event message based on the target scene data, including:

[0012] The target scene label is compared with a historical scene label corresponding to the historical event message.

[0013] When the target scene label is consistent with the historical scene label, the comparison result is determined according to a historical scene description corresponding to the historical event message and the target scene description.

[0014] The comparison result is determined according to the historical scene description corresponding to the historical event message and the target scene description, including:

[0015] A historical vector feature corresponding to the historical scene description and a target vector feature corresponding to the target scene description are obtained.

[0016] A feature similarity between the historical vector feature and the target vector feature is determined.

[0017] When the feature similarity is greater than or equal to a preset similarity, an image similarity of a target image corresponding to the target event message and a historical image corresponding to the historical event message is determined to determine the comparison result.

[0018] The method further includes:

[0019] When the feature similarity is less than the preset similarity, the comparison result is determined to be that the target event message and the historical event message are different event messages.

[0020] The comparison result is determined according to the image similarity of the target image and the historical image corresponding to the historical event message, including:

[0021] A target hash value corresponding to the target image and a historical hash value corresponding to the historical image are determined.

[0022] A Hamming distance of the target hash value and the historical hash value is determined.

[0023] The image similarity is determined according to the Hamming distance.

[0024] When the image similarity is greater than or equal to a preset image similarity, the comparison result is determined to be that the target event message and the historical event message are the same event message.

[0025] The target event message is compared with the historical event message based on the target scene data, further including:

[0026] obtain a historical event message set within a preset time window;

[0027] compare the target event message with each historical event message in the historical event message set one by one, to obtain one or more comparison results;

[0028] when the one or more comparison results all indicate that the target event message and the corresponding historical event message are different event messages, push the target event message.

[0029] wherein, obtaining the historical event message set within the preset time window comprises:

[0030] determining whether the cache space includes a plurality of historical event messages corresponding to the target device;

[0031] if yes, obtaining all historical event messages cached in the cache space to obtain a first set; determining a first timestamp according to a first preset time length corresponding to the cache space, and determining a second timestamp according to the preset time window;

[0032] when the first timestamp is later than the second timestamp, determining a target time period according to the preset time window and the first timestamp;

[0033] downloading historical event messages corresponding to the target time period in the database to obtain a second set;

[0034] determining the first set and the second set as the historical event message set.

[0035] wherein, the method further comprises:

[0036] obtaining a device identifier of the target device, and determining a target server corresponding to the device identifier;

[0037] sending the target event message to the target server.

[0038] To solve the above technical problems, the present application also proposes a message pushing device, which comprises a memory and a processor coupled with the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to realize the message pushing method as described above.

[0039] To solve the above technical problems, the present application also proposes a computer storage medium, which is used to store program data, and the program data is used to realize the message pushing method when executed by a computer.

[0040] Compared with the prior art, the beneficial effects of the present application are: through the embodiments of the present application, when the target event message reported by the target device is received, scene understanding is first performed, target scene data and a risk level corresponding to the target event message are obtained, the core semantics of the event message can be effectively captured, which is not limited by the limitations of keywords and text similarity, whether the target event message and the historical event message are the same event message can be accurately determined, and the accuracy of message filtering is improved. In addition, in the case where the risk level is less than a preset risk threshold, the target event message and the historical event message are compared based on the target scene data, low-risk events are ensured through risk level screening, only low-risk events are filtered, the timeliness and rationality of message pushing are taken into account, the interference of redundant messages on users is reduced, and key information is not missed. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0042] Figure 1 is an architecture schematic diagram of an Internet of Things system provided by the present application;

[0043] Figure 2 is a flow schematic diagram of an embodiment of a message pushing method provided by the present application;

[0044] Figure 3 is a space schematic diagram of storing historical event messages in the present application;

[0045] Figure 4 is a flow schematic diagram of obtaining a historical event message set provided by the present application;

[0046] Figure 5 is a flow schematic diagram of comparing a target event message with a historical event message provided by the present application;

[0047] Figure 6 is a flow schematic diagram of scene description comparison provided by the present application;

[0048] Figure 7 is a structure schematic diagram of an embodiment of a message pushing device provided by the present application;

[0049] Figure 8 is a structure schematic diagram of an embodiment of a message pushing device provided by the present application;

[0050] Figure 9 is a structure schematic diagram of an embodiment of a computer storage medium provided by the present application. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0052] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0053] The message push method of this application is applied to a message push device, wherein the message push device can be a server, a terminal device, or a system in which the server and the terminal device cooperate with each other. Accordingly, the various parts of the message push device, such as various units, sub-units, modules, and sub-modules, can all be set in the server, all in the terminal device, or separately in the server and the terminal device.

[0054] Furthermore, the aforementioned server can be either hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software programs or software modules, such as software or software modules used to provide distributed server functionality, or as a single software program or software module; no specific limitations are made here.

[0055] The message push method in this application is applied to an Internet of Things (IoT) system. Please refer to [link / reference]. Figure 1 , Figure 1 This is a schematic diagram of the architecture of the IoT system provided in this application, such as... Figure 1 As shown, the Internet of Things (IoT) system is divided into three layers: the device layer, the service layer, and the application layer. The device layer includes multiple IoT devices, the service layer includes multiple servers, and the user layer includes multiple user terminals.

[0056] The device layer Internet of Things device in the device layer is used to collect raw data such as environment and state in real time, the Internet of Things device built-in identification algorithm makes a preliminary judgment on the abnormal situation, generates an event message containing information such as event type, collection time, device identification, collected data, message text, and sends the event message to the server in the service layer through the network; the Internet of Things device can be a home intelligent device, a vehicle-mounted device, etc.

[0057] The server receives the event message reported by the Internet of Things device, executes filtering logic, and pushes the event message reserved after filtering to the user terminal in the application layer, and the user terminal can remind the user in the form of a pop-up window, a ringtone, a notification, etc. The user terminal includes a mobile phone, a tablet computer, etc.

[0058] Taking a camera as an Internet of Things device as an example, the camera obtains an image after photographing a collection area, and judges whether a preset event occurs by detecting the image. In the related technology, after it is determined that the preset event occurs, an event message including a message text is generated and sent to the server, and the message text is pushed to the user terminal by the server.

[0059] However, in the camera, the same event can correspond to different message text expressions. For example, the same camera detects a stranger staying in the living room, and the first reported message text is "a suspicious person stays in the living room area", and the second reported message text is "a non-family member appears in the shooting area". The two messages are essentially the same event description, but the text expressions are quite different. The filtering method based on keywords or text similarity cannot identify the core semantic association, and misjudges it as a non-repeated message, resulting in frequent push of redundant reminders.

[0060] Moreover, different core events can have high text similarity. For example, the camera detects a pet cat moving in the living room, and the reported message text is "an object moves in the living room area". The camera detects a stranger moving in the living room, and the reported message text is "a stranger object moves in the living room area". The text core keywords are highly coincident, and the text similarity is high, which is easy to be misjudged as a repeated message and filtered, resulting in missing report of a key security event.

[0061] The message pushing method of the present application is applied to the server of the Internet of Things system, and the filtering logic of the server in the related technology is improved.

[0062] For details, please refer to Figure 2 , Figure 2 is a flowchart of an embodiment of the message pushing method provided by the present application, as Figure 2 shown, and the specific steps are as follows:

[0063] Step S11: upon receiving the target event message reported by the target device, performing scene understanding on the target event message to determine target scene data corresponding to the target event message and a risk level.

[0064] In the embodiments of the present application, the execution subject is a server in an Internet of Things system, and the target device is any Internet of Things device in the Internet of Things system that is in network connection with the server. The built-in identification algorithm of the target device performs preliminary determination on abnormal situations, generates an event message, i.e., a target event message, containing information such as event type, collection time, device identifier, collected data, and message text, and forwards the target event message to the server. After receiving the target message, the server performs a message filtering operation, i.e., determines whether to push the target event message to a user terminal.

[0065] When performing the message filtering operation, the server first performs scene understanding on the target event message reported by the target device to obtain target scene data corresponding to the target event message and a risk level.

[0066] Instead of using the way of extracting keywords in the message text for matching or calculating the text similarity to perform message filtering when the server receives the target event message, scene understanding is performed on the data collected by the target device in the target event message, and the overall scene in which the target event occurs is analyzed. Taking the target device as a target camera as an example, scene understanding is performed on the image collected by the target camera in the target event message to obtain target scene data and a risk level.

[0067] The target scene data includes a scene label and a scene description. The scene label represents the subject of the event, such as a person, a pet, or a vehicle, and the scene description is a specific description of the event. The collected data, such as the target image, can be intelligently analyzed by an AI large model to obtain features such as the scene label, the scene description, and the risk level through preset prompt words.

[0068] Step S12: in a case where the risk level is less than a preset risk threshold, comparing the target event message with historical event messages based on the target scene data to obtain a comparison result.

[0069] Step S13: in a case where the comparison result indicates that the target event message and the historical event message are different event messages, pushing the target event message.

[0070] In the embodiments of the present application, when performing message filtering, first, it is determined whether the risk level is less than a preset risk threshold. If not, it indicates that the event is a high-risk event, and the user needs to be pushed immediately. For a high-risk event, the target event message is not further determined, and is directly pushed in full amount.

[0071] If yes, it indicates that the event is a low-risk event with low emergency level, and the target event message can be compared with the historical event message to determine whether to push the target event message.

[0072] The preset risk threshold is pre-set, for example, the risk level is divided into 1-5 levels, level 1 is the lowest risk, level 5 is the highest risk, and level 3 is the preset risk threshold. When the risk level is greater than or equal to level 3, the corresponding target event message is not filtered and directly pushed in full amount.

[0073] The target event message is compared with the historical event message based on the target scene data to determine whether the target event message and the historical event message are the same event message. If yes, the target event message is filtered, i.e., the target event message is not pushed. If no, the target event message is pushed.

[0074] The historical event message is an event message reported by the target device before the target event message is reported, which can be specifically referred to as a previous event message of the target event message. At this time, the target event message and the previous event message are different event messages, and the target event message is pushed.

[0075] The historical event message can also be all event messages reported by the target device within a preset time window before the target event message is reported.

[0076] At this time, the target event message is compared with the historical event message based on the target scene data, which further includes: obtaining a historical event message set within a preset time window; comparing the target event message with each historical event message in the historical event message set one by one to obtain one or more comparison results; and when one or more comparison results all indicate that the target event message and the corresponding historical event message are different event messages, the target event message is pushed.

[0077] In the embodiments of the present application, the duration of the preset time window can be pre-set, for example, 10 minutes, 30 minutes, 1 hour, etc. Taking the duration of the preset time window as 10 minutes as an example, the 10 minutes before the current time is the preset time window, and all historical event messages reported by the target device within the 10 minutes before the current time form a historical event message set. The target event message is compared with each historical event message in the historical event message set one by one. When the target event message and each historical event message in the historical event message set are different event messages, i.e., one or more comparison results all indicate that the target event message and the corresponding historical event message are different event messages, the target event message is pushed.

[0078] In an optional embodiment, the historical event message set in the preset time window is acquired, including: judging whether the cache space includes a plurality of historical event messages corresponding to the target device.

[0079] If yes, all the historical event messages cached in the cache space are acquired to obtain a first set; a first timestamp is determined according to a first preset time length corresponding to the cache space, and a second timestamp is determined according to the preset time window; when the first timestamp is later than the second timestamp, a target time period is determined according to the preset time window and the first timestamp; the historical event messages corresponding to the target time period are downloaded from the database to obtain a second set; and the first set and the second set are determined as the historical event message set.

[0080] In the embodiments of the present application, please refer to Figure 3 , Figure 3 is a space diagram for storing historical event messages in the present application, as shown in Figure 3 , the cache space and the database are provided to store historical event messages, the cache space is a redis (RemoteDictionary Server, Remote Dictionary Server) cache, which is a temporary cache space for caching data; the database is a vector database, including but not limited to Milvus (an open source vector database based on vector retrieval), Elasticsearch (an open source search engine focusing on full-text retrieval), etc., which is a long-term database for caching data and does not actively delete saved data.

[0081] The server is connected to the cache space and the database through a network to realize uploading and downloading of event messages. The speed of downloading data from the cache space is faster than that of downloading data from the database.

[0082] The cache space caches the event messages reported by the target device. After an event message is reported to the server by the target device, the event message is cached in the cache space and removed from the cache space after a first preset time length. After an event message is reported to the server by the target device, it is also forwarded to the database for permanent storage of the event message in the database.

[0083] Please refer to Figure 4 , Figure 4 is a flowchart of acquiring a historical event message set provided by the present application, as shown in Figure 4 , including:

[0084] S21: judging whether the cache space has data, i.e., whether it includes a plurality of historical event messages, if yes, performing step S22, if not, performing step S27.

[0085] The historical event messages stored in the cache space are event messages reported within a first preset time period before the current time. If there are historical event messages corresponding to the target device in the cache space, it indicates that the target device reported event messages within the first preset time period before the current time, otherwise, it indicates that the time interval from the last time the target device reported event messages has exceeded the first preset time period.

[0086] S22: Obtain all historical event messages in the cache space to obtain a first set.

[0087] The historical event messages in the cache space are downloaded to the local server to obtain the first set.

[0088] S23: Determine a first timestamp and a second timestamp.

[0089] The time interval between the first timestamp and the current time is a first preset time period, and the first timestamp is before the current time. The second timestamp is the start time of a preset time window, the end time of the preset time window is the current time, and the time interval of the preset time window is a second preset time period.

[0090] S24: Determine whether the first timestamp is later than the second timestamp. If yes, perform step S25, otherwise, perform step S28.

[0091] By comparing the first timestamp and the second timestamp, it is determined whether all historical event messages in the cache space cover all event messages within the preset time window.

[0092] When the first preset time period is configured to be less than the second preset time period, the first timestamp is later than the second timestamp, and all historical event messages in the cache space cannot cover all event messages within the preset time window.

[0093] When the second preset time period is configured to be greater than the second preset time period, the first timestamp is not later than the second timestamp, and all historical event messages in the cache space cover all event messages within the preset time window.

[0094] S25: Determine a target time period according to the preset time window and the first timestamp; download historical event messages corresponding to the target time period from the database to obtain a second set.

[0095] When the first timestamp is later than the second timestamp, only part of the event messages within the preset time window are cached in the cache space. In order to ensure that all event messages within the preset time window are obtained, the target time period corresponding to the missing event messages in the cache space needs to be determined, the start time of the target time period is the second timestamp, the end time of the target time period is the first timestamp, and the historical event messages corresponding to the target time period are downloaded from the database to obtain the second set.

[0096] S26: determine the first set and the second set as the historical event message set.

[0097] S27: download the corresponding historical event messages in the preset time window from the database to obtain the historical event message set.

[0098] If there is no data in the cache space, i.e., the target device does not report event messages for a long time before reporting the target event message, at this time, the cache space does not contain historical event messages, but only contains the target event message currently reported. At this time, the historical event messages corresponding to the preset time window need to be downloaded from the database.

[0099] S28: determine the first set as the historical event message set.

[0100] If the first timestamp is not later than the second timestamp, the historical event messages in the cache space can cover all historical event messages in the preset time window, and therefore, the event messages do not need to be downloaded from the database, and the first set is directly determined as the historical event message set.

[0101] Through the above embodiments, the cache mechanism of the cache space and the database stores the historical event messages. Based on the multi-level cache mechanism, a multi-level historical event message downloading mechanism is provided, which preferentially downloads the historical event messages from the cache space and then downloads the historical event messages from the database, and then obtains the historical event messages. Since the speed of downloading the event messages from the cache space is faster than that of downloading the event messages from the database, through the multi-level historical event message downloading mechanism, the aggregation speed of the event information can be accelerated, and the aggregation of the event information refers to determining the historical event message set.

[0102] Optionally, the embodiments of the present application also provide a multi-level cache mechanism of the server local, the cache space and the database. The historical event message set downloaded from the cache space and the database is saved in the server local for a third preset time length.

[0103] If the target device reports event messages multiple times in a short time, the historical event message set downloaded when filtering the previous event message is still cached in the server local when filtering the target event message corresponding to the current time. At this time, the historical event messages do not need to be downloaded from the cache space and the database, and only the first historical event message set when filtering the previous event message (denoted as the first event message) saved in the server local is added to the first event message to obtain the historical event message set corresponding to the target event message. Through the multi-level cache mechanism of the server local, the cache space and the database, the number of data downloads is reduced, and the message filtering efficiency is improved.

[0104] In an optional embodiment, a device identifier of the target device is acquired, and a corresponding target server is determined according to the device identifier; and the target event message is sent to the target server.

[0105] In the embodiments of the present application, in the case of distributed deployment of servers, after the current server receives the target event message reported by the target device, the device identifier of the target device is acquired, and all servers perform consistent hash allocation based on the device identifier to allocate a corresponding unique server for the corresponding device, that is, all event messages reported by the same device are sent to the corresponding server, and the corresponding server performs filtering logic.

[0106] The server corresponding to the target device is a target server, and any server distributedly deployed determines the target server corresponding to the target device after receiving the target event message reported by the target device. When the current server is not the target server, the target event message is forwarded to the target server, and the target server performs a series of operations such as scene understanding, comparison, filtering or pushing. If the current server is the target server, a series of operations such as scene understanding, comparison, filtering or pushing are completed through the current server.

[0107] In combination with the multi-level cache mechanism of the server local, cache space and database, the historical event message set downloaded when filtering the previous event message is still cached in the server local, but it is necessary to ensure that the servers processing the previous event message and the target event message are the same server, so as to avoid downloading the historical event message set from the cache space and the database, and the event message cached in the server local can be directly used.

[0108] Especially when multiple event messages are triggered in a short time, processing in the same server can accelerate the filtering speed by using the historical event messages in the server, without the need to download the historical event message set from the cache space and / or the database again. If the event messages reported each time are allocated to different servers, for example, the previous event message is allocated to the first server for processing, and the target event message corresponding to the current time is allocated to the second server for processing. The data downloaded when filtering the previous event message is in the first server local, and there is no data in the second server local. Therefore, the second server needs to download from the cache space and the database when processing the target event message, which increases the data download pressure and reduces the filtering speed.

[0109] Through the above embodiments, all event messages reported by the same device are filtered by the same server, which avoids the allocation of event messages reported by the same device to different servers, thereby reducing the data download pressure of the server and improving the filtering speed of the event message.

[0110] Please refer to Figure 5 ,Figure 5 is a flowchart of the comparison between the target event message and the historical event message provided in the present application, as shown in Figure 5 , comprising:

[0111] S31, determining whether the target scene label is consistent with the historical scene label, if yes, executing step S32, if not, executing step S33.

[0112] In the comparison between the target event message and the historical event message, the final comparison result is determined through multi-level feature comparison. First, the scene labels are compared, and the target scene label corresponding to the target event message is compared with the historical scene label corresponding to the historical event message.

[0113] S32, determining the comparison result according to the historical scene description corresponding to the historical event message and the target scene description.

[0114] When the target scene label and the historical scene label are consistent, the final comparison result can be determined only through the comparison of the next level feature, that is, the comparison result of the event message is determined according to the historical scene description corresponding to the historical event message and the target scene description of the event message.

[0115] S33, determining that the comparison result is that the target event message and the historical event message are different event messages.

[0116] When the target scene label and the historical scene label are inconsistent, it indicates that the target event message is necessarily in a new scene, that is, the target event message and the historical event message are different event messages, and at this time, the target event message is pushed.

[0117] Further, the comparison result is determined according to the historical scene description corresponding to the historical event message and the target scene description, comprising: obtaining historical vector features corresponding to the historical scene description and target vector features corresponding to the target scene description; determining the feature similarity between the historical vector features and the target vector features; when the feature similarity is greater than or equal to a preset similarity, determining the comparison result according to the image similarity of the target image corresponding to the target event message and the historical image corresponding to the historical event message; when the feature similarity is less than the preset similarity, determining that the comparison result is that the target event message and the historical event message are different event messages.

[0118] In the embodiments of the present application, please refer to Figure 6 , Figure 6 is a flowchart of the comparison between the target event message and the historical event message provided in the present application, as shown in Figure 6 , comprising:

[0119] S321, acquire a historical vector feature corresponding to the historical scene description and a target vector feature corresponding to the target scene description.

[0120] The historical scene description and the target scene description can be text messages. The scene description is embedded to extract a vector feature, the semantic information in the text is learned, and the corresponding vector feature is obtained. The vector feature corresponding to the historical scene description is the historical vector feature, and the target vector feature corresponding to the target scene description is the target vector feature.

[0121] It should be noted that the AI large model is used to understand the scene of the input image (for example, the target image) and extract the description feature of the scene corresponding to the target image, that is, the scene description. The Embedding algorithm is used to extract the vector feature of the scene description, and the scene description is converted from high-dimensional data to low-dimensional continuous vectors. The similarity of the scene description is determined according to the similarity of the vector feature, and the calculation speed of the similarity of the scene description is accelerated.

[0122] S322, determine the feature similarity between the historical vector feature and the target vector feature.

[0123] S323, determine whether the feature similarity is greater than or equal to a preset similarity. If yes, execute step S324, if not, execute step S325.

[0124] S324, determine the comparison result according to the image similarity between the target image corresponding to the target event message and the historical image corresponding to the historical event message.

[0125] When the feature similarity between the historical vector feature and the target vector feature is greater than or equal to the preset similarity, it cannot be directly determined that the target event message and the historical event message are the same event message. For example, the scene descriptions of the target event message and the historical event message are both that there is a black cat in the target area, but in fact they are two different cats. At this time, the scene descriptions are the same, the feature similarity between the historical vector feature and the target vector feature is greater than or equal to the preset similarity, but the target event message and the historical event message are different event messages. Therefore, the next level of feature comparison is performed, that is, the features of the images at the time of the event are directly compared, that is, the comparison result is determined according to the image similarity between the target image and the historical image.

[0126] Specifically, the determining the comparison result according to the image similarity of the historical image corresponding to the target image and the historical event message comprises: determining a target hash value corresponding to the target image and a historical hash value corresponding to the historical image; determining a Hamming distance of the target hash value and the historical hash value; determining the image similarity according to the Hamming distance; and determining that the comparison result is that the target event message and the historical event message are the same event message when the image similarity is greater than or equal to a preset image similarity.

[0127] In the embodiment of the present application, when determining the image similarity of the target image and the historical image, each image is converted into a hash value by a difference hash algorithm, the hash value corresponding to the target image is a target hash value, and the hash value corresponding to the historical image is a historical hash value. The Hamming distance of the target hash value and the historical hash value is calculated, the image similarity is determined according to the Hamming distance, the comparison result is determined according to the image similarity, the comparison result is determined to be that the target event message and the historical event message are the same event message when the image similarity is greater than or equal to a preset image similarity, and the comparison result is determined to be that the target event message and the historical event message are different event messages when the image similarity is less than the preset image similarity. At this time, the target event message is pushed.

[0128] S325, determining that the comparison result is that the target event message and the historical event message are different event messages.

[0129] Through the embodiment of the present application, when the target event message reported by the target device is received, scene understanding is first performed, target scene data and a risk level corresponding to the target event message are obtained, the core semantics of the event message can be effectively captured, which is not limited by the limitations of keywords and text similarity, whether the target event message and the historical event message are the same event message can be accurately determined, and the accuracy of message filtering is improved. In addition, in the case that the risk level is less than a preset risk threshold, the target event message and the historical event message are compared based on the target scene data, low-risk events are ensured through risk level screening, the timeliness and rationality of message pushing are taken into account, the interference of redundant messages on users is reduced, and key messages are not missed.

[0130] Those skilled in the art can understand that, in the above method of the specific implementation manner, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined according to its function and possible internal logic.

[0131] To implement the above message pushing method, the present application further provides a message pushing device. For details, please refer to Figure 7 , Figure 7 is a structural schematic diagram of an embodiment of the message pushing device provided by the present application.

[0132] The message pushing device 500 of the embodiment comprises:

[0133] a scene understanding module 51, configured to perform scene understanding on a target event message reported by a target device, to determine target scene data corresponding to the target event message and a risk level;

[0134] a comparison module 52, configured to compare the target event message with historical event messages based on the target scene data, to obtain a comparison result, in a case where the risk level is less than a preset risk threshold;

[0135] a pushing module 53, configured to push the target event message, in a case where the comparison result indicates that the target event message and the historical event messages are different event messages.

[0136] To implement the message pushing method, the application further provides a message pushing device, please refer to Figure 8 , Figure 8 is a structural schematic diagram of an embodiment of the message pushing device provided by the application.

[0137] The message pushing device 400 of the embodiment comprises a processor 41, a memory 42, an input / output device 43 and a bus 44.

[0138] The processor 41, the memory 42 and the input / output device 43 are connected with the bus 44 respectively, the memory 42 stores program data, and the processor 41 is configured to execute the program data to implement the message pushing method of the above-mentioned embodiments.

[0139] In the embodiment of the application, the processor 41 can also be referred to as a CPU (Central Processing Unit). The processor 41 can be an integrated circuit chip with signal processing capability. The processor 41 can also be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor 41 can also be any conventional processor.

[0140] The application further provides a computer storage medium, please continue to refer to Figure 9 , Figure 9is a structural schematic diagram of an embodiment of the computer storage medium provided in the present application. The computer storage medium 600 stores a computer program 61. When the computer program 61 is executed by a processor, the computer program 61 is used to implement the message pushing method of the above-described embodiment.

[0141] When the embodiments of the present application are implemented in the form of software functional units and sold or used as independent products, the software functional units can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions 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 enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0142] The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent flow transformation based on the content of the specification and drawings, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A message push method, characterized in that, The message push method includes: Upon receiving a target event message reported by the target device, the target event message is subjected to scenario understanding to determine the target scenario data and risk level corresponding to the target event message. The target scenario data includes a target scenario tag and a target scenario description. If the risk level is less than a preset risk threshold, the target event message is compared with historical event messages based on the target scenario data to obtain the comparison result; When the comparison result indicates that the target event message and the historical event message are different event messages, the target event message will be pushed. The step of comparing the target event message with historical event messages based on the target scene data includes: comparing the target scene tag with the historical scene tag corresponding to the historical event message; when the target scene tag and the historical scene tag are consistent, obtaining the historical vector feature corresponding to the historical scene description and the target vector feature corresponding to the target scene description; determining the feature similarity between the historical vector feature and the target vector feature; when the feature similarity is greater than or equal to a preset similarity, determining the comparison result based on the image similarity between the target image corresponding to the target event message and the historical image corresponding to the historical event message, wherein the target image is the image captured by the target device when the target event occurs.

2. The message push method according to claim 1, characterized in that, The method further includes: When the feature similarity is less than the preset similarity, the comparison result is determined to be that the target event message and the historical event message are different event messages.

3. The message push method according to claim 2, characterized in that, Determining the comparison result based on the image similarity between the target image and the historical image corresponding to the historical event message includes: Determine the target hash value corresponding to the target image and the historical hash value corresponding to the historical image; Determine the Hamming distance between the target hash value and the historical hash value; The image similarity is determined based on the Hamming distance; When the image similarity is greater than or equal to a preset image similarity, the comparison result is determined to be that the target event message and the historical event message are the same event message.

4. The message push method according to claim 1, characterized in that, The comparison of the target event message with historical event messages based on the target scene data also includes: Retrieve the set of historical event messages within a preset time window; The target event message is compared one by one with each historical event message in the historical event message set to obtain one or more comparison results; When one or more of the comparison results indicate that the target event message and the corresponding historical event message are different event messages, the target event message will be pushed.

5. The message push method according to claim 4, characterized in that, Retrieve the set of historical event messages within a preset time window, including: Determine whether the cache space contains several historical event messages corresponding to the target device; If so, obtain all historical event messages cached in the cache space to get a first set; determine a first timestamp according to the first preset duration corresponding to the cache space, and determine a second timestamp according to the preset time window; When the first timestamp is later than the second timestamp, the target time period is determined based on the preset time window and the first timestamp; Download historical event messages corresponding to the target time period from the database to obtain a second set; The first set and the second set are determined as the historical event message set.

6. The message push method according to claim 1, characterized in that, The method further includes: Obtain the device identifier of the target device, and determine the corresponding target server based on the device identifier; Send the target event message to the target server.

7. A message push device, characterized in that, The message push device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the message push method as described in any one of claims 1 to 6.

8. A computer storage medium, characterized in that, The computer storage medium is used to store program data, which, when executed by the computer, is used to implement the message push method as described in any one of claims 1 to 6.

Images