Alarm push optimization method and device for AI identification drowning prevention and medium

Abstract

The invention discloses an alarm pushing method and device for AI identification drowning prevention and a medium, and relates to the technical field of image identification. The method comprises the following steps: identifying and analyzing a real-time video stream based on a preset anti-drowning alarm pushing model to detect whether an anti-drowning event exists; when it is detected that the anti-drowning event exists, continuously generating an alarm message of which an alarm corresponds to a preset number of coherent video frames, and determining an alarm type corresponding to the alarm message; in the process of continuously generating the alarm messages, continuously judging whether adjacent alarm messages meet message merging logic or not; wherein the message merging logic is that the time interval of the adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of the adjacent alarm messages are the same; and in the process of continuously generating the alarm messages, if the alarm messages meeting the message merging logic reach the merging upper limit number, merging the alarm messages with the merging upper limit number into an upper limit alarm packet, and pushing the upper limit alarm packet to an alarm platform.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to an alarm push method, device and medium for AI recognition and drowning prevention. Background Art

[0002] With the rapid development of science and technology, artificial intelligence (AI) technology has been widely used in various fields, especially in the field of image recognition, which has made significant progress. In water conservancy safety management, AI image recognition technology is gradually becoming an important means to prevent safety accidents such as drowning. However, in actual application, the traditional AI alarm message push method has exposed a series of problems that need to be solved urgently.

[0003] Traditional AI alarm message push methods often have drawbacks such as repeated alarms and frequent pushes. Since monitoring equipment continuously captures image data and performs real-time identification and analysis, once abnormal behavior is detected (such as people wandering, entering the water, fishing, etc.), an alarm message will be triggered immediately. However, in actual scenarios, the same abnormal behavior may continue for a period of time, causing the system to push the same alarm message multiple times in a row. This repeated push not only wastes valuable computing and network resources, but may also cause unnecessary troubles and misjudgments for managers.

[0004] In addition, the traditional push method lacks flexibility and intelligence. It cannot adjust the push strategy according to different types of alarms and scenarios, nor can it effectively integrate and process alarm messages. This, to a certain extent, limits the application space and application effect of AI image recognition technology in water conservancy safety management. Summary of the invention

[0005] The embodiments of the present application provide a method, device and medium for pushing alarms for AI identification and drowning prevention, which are used to solve the following technical problems: In water conservancy safety management, how to reduce repeated push and improve the accuracy of alarm messages for AI identification and drowning prevention has become a technical problem that needs to be solved urgently.

[0006] In a first aspect, an embodiment of the present application provides an alarm push method for AI identification and drowning prevention, characterized in that the method includes: identifying and analyzing a real-time video stream based on a preset drowning prevention alarm push model to detect whether there is a drowning prevention event; when a drowning prevention event is detected, continuously generating an alarm message corresponding to a preset number of consecutive video frames, and determining the alarm type corresponding to the alarm message; wherein the alarm types include: alarms for wandering people in dangerous waters, alarms for entering water in dangerous waters, and alarms for fishing in dangerous waters; in the process of continuously generating alarm messages, continuously judging whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the upper limit number of merging, the alarm messages of the upper limit number of merging are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

[0007] In one implementation of the present application, the method also includes: in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and the message type of the current alarm message generated is different from the message type of several previous alarm messages, merging the previous alarm messages into a segmented alarm package, and pushing the segmented alarm package to the alarm platform; and, using the current alarm message as the initial alarm message that is continuously generated, monitoring the upper limit of the number of alarm messages that can be merged.

[0008] In one implementation of the present application, the method also includes: in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and the time interval between the current alarm message generated and the previous alarm message is greater than the merging time threshold, merging several previous alarm messages into a segmented alarm package, and pushing the segmented alarm package to the alarm platform; and, using the current alarm message as the initial alarm message that is continuously generated, monitoring the upper limit of the number of alarm messages that can be merged.

[0009] In one implementation of the present application, the method also includes: in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and there are no subsequent alarm messages, several previous alarm messages are merged into a termination alarm package, and the termination alarm package is pushed to the alarm platform.

[0010] In one implementation of the present application, before identifying and analyzing the real-time video stream based on a preset anti-drowning alarm push model, the method also includes: obtaining sample video data to be labeled, and labeling the sample video data to be labeled to extract scene feature values ​​under corresponding scenes; wherein the sample video data to be labeled includes: scene data of people wandering in dangerous waters, scene data of people entering water in dangerous waters, and scene data of fishing in dangerous waters; the scene feature values ​​are input into a preset neural network model for training until a converged anti-drowning alarm push model is obtained.

[0011] In one implementation of the present application, sample video data to be labeled are labeled to extract scene feature values ​​in corresponding scenes, specifically including: labeling each frame of sample video data to be labeled frame by frame to determine whether there is a drowning prevention event in each frame; for video frames with drowning prevention events, identifying and labeling the event subject, event behavior and event scene; wherein the event subject includes people, objects or waters, event behaviors include wandering, entering the water or fishing, and the event scene is a dangerous water area; based on the labeling results, using a feature extraction algorithm to extract scene feature values ​​in the corresponding scene; wherein the scene feature values ​​include but are not limited to personnel features, behavior features, water features and object features.

[0012] In one implementation of the present application, after determining the alarm type corresponding to the alarm message, the method also includes: determining the alarm level according to the alarm type, and assigning a corresponding alarm level identifier to the alarm message; wherein the alarm level includes three levels: high, medium, and low; and adding the alarm level identifier to the alarm message.

[0013] In one implementation of the present application, the alarm message includes: alarm time, alarm area, alarm type, alarm level identifier, alarm video and alarm picture.

[0014] In a second aspect, an embodiment of the present application further provides an alarm push device for AI identification and drowning prevention, characterized in that the device comprises: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: identify and analyze the real-time video stream based on a preset drowning prevention alarm push model to detect whether there is a drowning prevention event; when a drowning prevention event is detected, continuously generate an alarm message corresponding to a preset number of consecutive video frames, and determine the alarm message. The alarm types corresponding to the information include: alarm for wandering persons in dangerous waters, alarm for entering water by persons in dangerous waters, and alarm for fishing in dangerous waters; in the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein, the message merging logic is that the time interval between adjacent alarm messages is not greater than the preset merging time threshold and the alarm types of adjacent alarm messages are the same; in the process of continuously generating alarm messages, if the number of alarm messages that meet the message merging logic reaches the merging upper limit, the alarm messages with the merging upper limit will be merged into an upper limit alarm package, and the upper limit alarm package will be pushed to the alarm platform.

[0015] In a third aspect, an embodiment of the present application further provides a non-volatile computer storage medium for AI-based anti-drowning alarm push, storing computer executable instructions, characterized in that the computer executable instructions are configured to: identify and analyze a real-time video stream based on a preset anti-drowning alarm push model to detect whether there is an anti-drowning event; when an anti-drowning event is detected, continuously generate an alarm message corresponding to a preset number of consecutive video frames, and determine the alarm type corresponding to the alarm message; wherein the alarm types include: alarms for wandering in dangerous waters, alarms for entering water in dangerous waters, and alarms for fishing in dangerous waters; in the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the upper limit of the merge, the alarm messages merging the upper limit are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

[0016] The embodiments of the present application provide a method, device and medium for AI-based drowning prevention warning push, which have the following beneficial effects: 1. Reduce the duplication and frequent push of alarm messages: By combining adjacent alarm messages that meet certain conditions into an alarm package for push, the duplication and frequent push of alarm messages caused by the continuous triggering of the same event are effectively reduced. This not only reduces the processing burden of the responsible person, but also avoids the risk of important alarms being ignored due to information overload.

[0017] 2. Improve the efficiency of message push: The alarm package is pushed as a whole, which retains the consistency and integrity of the alarm message, allowing the responsible person to obtain complete alarm event information at one time, thereby improving the efficiency of message push. At the same time, this also enables relevant departments and units to respond quickly and take effective measures, such as remote shouting and expulsion, rapid rescue, etc., to deal with drowning prevention incidents.

[0018] 3. Save resources and operating costs: By reducing unnecessary alert message pushes, this application saves network resources and operator resources and reduces system operating costs. This is especially important for large-scale deployment and long-term operation of drowning prevention monitoring systems.

[0019] 4. Enhanced intelligence and automation: This application combines advanced technologies such as AI image recognition technology and big data analysis to build an innovative alarm push mechanism. This mechanism can automatically identify drowning prevention incidents, automatically merge alarm messages and push alarm packages, thus improving the intelligence and automation level of drowning prevention monitoring.

[0020] 5. Improve safety and emergency response capabilities: By pushing alarm packages in a timely and accurate manner, this application helps relevant departments and units quickly grasp the situation of drowning prevention incidents, improves emergency response capabilities and safety assurance levels. This is of great significance for reducing economic losses and casualties caused by drowning incidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flow chart of an alarm push method for AI recognition and drowning prevention provided in an embodiment of the present application; Figure 2 A schematic diagram of the internal structure of an alarm push device for AI recognition and drowning prevention provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0023] The embodiments of the present application provide a method, device and medium for pushing alarms for AI identification and drowning prevention, which are used to solve the following technical problems: In water conservancy safety management, how to reduce repeated push and improve the accuracy of alarm messages for AI identification and drowning prevention has become a technical problem that needs to be solved urgently.

[0024] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0025] Figure 1 This is a flow chart of a method for pushing an alarm for AI-based drowning prevention. Figure 1 As shown, an alarm push method for AI recognition and drowning prevention provided in an embodiment of the present application specifically includes the following steps: Step 101: Based on a preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention event; In one embodiment of the present application, in order to implement AI-based anti-drowning alarm push, the present application first identifies and analyzes the real-time video stream based on a preset anti-drowning alarm push model to detect whether there is a drowning prevention incident.

[0026] It should be noted that in order to realize the recognition and analysis of real-time video streams, it is also necessary to train an anti-drowning alarm push model.

[0027] Specifically, sample video data to be labeled are obtained, and the sample video data to be labeled are labeled to extract scene feature values ​​under corresponding scenes; wherein the sample video data to be labeled include: scene data of people wandering in dangerous waters, scene data of people entering water in dangerous waters, and scene data of fishing in dangerous waters; the scene feature values ​​are input into a preset neural network model for training until a converged anti-drowning alarm push model is obtained.

[0028] In one embodiment of the present application, sample video data to be labeled are labeled to extract scene feature values ​​in corresponding scenes, specifically including: labeling each frame of sample video data to be labeled frame by frame to determine whether there is a drowning prevention event in each frame; for video frames with drowning prevention events, identifying and labeling event subjects, event behaviors and event scenes; wherein event subjects include people, objects or waters, event behaviors include wandering, entering water or fishing, and event scenes are dangerous waters; based on the labeling results, using a feature extraction algorithm to extract scene feature values ​​in the corresponding scene; wherein scene feature values ​​include but are not limited to personnel features, behavior features, water features and object features.

[0029] In one embodiment, a large amount of video data is collected from multiple actual monitoring scenes. These data should cover scenes such as people wandering in dangerous waters, people entering the water, fishing, etc., to ensure the diversity and representativeness of the data and improve the generalization ability of the model. The collected video data is finely annotated. The annotation work includes checking each frame of the video frame by frame to determine whether there is a drowning prevention event. For video frames with drowning prevention events, the event subject (such as people, objects), event behavior (such as wandering, entering the water, fishing) and event scene (such as dangerous waters) are further identified and annotated. The annotation results will serve as an important input for model training. Based on the annotation results, advanced feature extraction algorithms are used to extract key scene feature values ​​from the video data. These feature values ​​include but are not limited to personnel features (such as facial features, body features), behavior features (such as motion trajectories, action patterns), water features (such as water shape, water depth information) and object features (such as lifebuoys, fishing rods, etc.). The extracted scene feature values ​​are input into a preset neural network model (such as a convolutional neural network CNN) for training. During the training process, the model parameters are continuously adjusted to optimize the model performance until the model reaches a convergence state on the validation set.

[0030] Step 102: When a drowning prevention event is detected, continuously generate an alarm message corresponding to a preset number of consecutive video frames, and determine the alarm type corresponding to the alarm message.

[0031] It is understandable that the types of warnings include: warning of people wandering in dangerous waters, warning of people entering the water in dangerous waters, and warning of fishing in dangerous waters.

[0032] In one embodiment of the present application, after determining the alarm type corresponding to the alarm message, the method also includes: determining the alarm level according to the alarm type, and assigning a corresponding alarm level identifier to the alarm message; wherein the alarm level includes three levels: high, medium, and low; and adding the alarm level identifier to the alarm message.

[0033] It can also be understood that the alarm message includes at least: alarm time, alarm area, alarm type, alarm level identification, alarm video and alarm picture.

[0034] In one embodiment, the system continuously monitors the real-time video stream and uses the trained anti-drowning alarm push model to perform intelligent analysis on each frame in the video stream. When the model detects the presence of an anti-drowning event (such as a person wandering, a person entering the water, fishing, etc.), the alarm mechanism is triggered immediately. The system automatically extracts a corresponding number of coherent video frames from the video stream in which the anti-drowning event is detected according to the preset number of alarm message video frames. While extracting the video frames, the system determines the alarm type corresponding to the alarm message according to the recognition result of the anti-drowning alarm push model. The alarm type usually includes a person wandering, a person entering the water, fishing, etc., depending on the type of anti-drowning event actually monitored. Combining the video frames extracted in the above steps and the determined alarm type, the system generates a complete alarm message, which includes at least: alarm time, alarm area, alarm type, alarm level identifier, alarm video and alarm picture. As the real-time video stream is generated, the corresponding alarm message is continuously generated.

[0035] Step 103: In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic.

[0036] The message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same.

[0037] Step 104: In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the merging upper limit, the alarm messages of the merging upper limit are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

[0038] In one embodiment, the system sets the logical conditions for message merging, which mainly include two aspects: one is that the time interval between adjacent alarm messages is not greater than the preset merging time threshold; the other is that the alarm types of adjacent alarm messages must be the same. These two conditions together constitute the logical basis for message merging. In the process of continuously generating alarm messages, the system will check in real time whether the newly generated alarm message and the last stored alarm message meet the above merging logic. If so, the two messages are regarded as candidates for merging. In order to avoid the timeliness of the alarm, the system also sets the upper limit of the number of merges, which means that even if the adjacent alarm messages meet the merging logic, when the preset upper limit of the number of merges is reached, the system will stop further merging and pack the currently merged alarm messages into an upper limit alarm package. When the alarm messages that meet the merging logic reach the upper limit of the number of merges, the system will automatically merge these alarm messages into an upper limit alarm package. The upper limit alarm package contains all the merged alarm data, including the alarm time, alarm type, merged alarm video (consisting of a collection of multiple coherent video frames) and possible alarm pictures. After the upper limit alarm package is generated, the system will immediately push the alarm package to the alarm platform according to the preset push rules and methods. The alarm platform will be responsible for receiving, processing and displaying these alarm information so that the relevant responsible persons can understand and respond in time.

[0039] In one embodiment of the present application, in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and the message type of the current alarm message generated is different from the message type of several previous alarm messages, the previous alarm messages are merged into a segmented alarm package, and the segmented alarm package is pushed to the alarm platform; and, with the current alarm message as the initial alarm message that is continuously generated, the upper limit of the number of alarm messages that can be merged is monitored.

[0040] In one embodiment of the present application, in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and the time interval between the current alarm message generated and the previous alarm message is greater than the merging time threshold, several previous alarm messages are merged into a segmented alarm package, and the segmented alarm package is pushed to the alarm platform; and, with the current alarm message as the initial alarm message that is continuously generated, the upper limit of the number of alarm messages to be merged is monitored.

[0041] In one embodiment of the present application, in the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merges, and there are no subsequent alarm messages, several previous alarm messages are merged into a termination alarm package, and the termination alarm package is pushed to the alarm platform.

[0042] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, this application embodiment also provides an alarm push device for AI recognition and drowning prevention, and its structure is as follows: Figure 2 shown.

[0043] Figure 2 This is a schematic diagram of the internal structure of an alarm push device for AI recognition and drowning prevention provided in an embodiment of the present application. Figure 2 As shown, the device includes: at least one processor 201; and, a memory 202 communicatively connected to the at least one processor; The memory 202 stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor 201 to enable at least one processor 201 to: Based on the preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention incident; When a drowning prevention event is detected, an alarm message corresponding to a preset number of consecutive video frames is continuously generated, and the alarm type corresponding to the alarm message is determined; the alarm types include: alarm for people wandering in dangerous waters, alarm for people entering water in dangerous waters, and alarm for fishing in dangerous waters; In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; In the process of continuously generating alarm messages, if the number of alarm messages that meet the message merging logic reaches the merging upper limit, the alarm messages that meet the merging upper limit will be merged into an upper limit alarm package, and the upper limit alarm package will be pushed to the alarm platform.

[0044] Some embodiments of the present application provide corresponding Figure 1 A non-volatile computer storage medium for AI-based drowning prevention warning push, storing computer executable instructions, wherein the computer executable instructions are set as follows: Based on the preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention incident; When a drowning prevention event is detected, an alarm message corresponding to a preset number of consecutive video frames is continuously generated, and the alarm type corresponding to the alarm message is determined; the alarm types include: alarm for people wandering in dangerous waters, alarm for people entering water in dangerous waters, and alarm for fishing in dangerous waters; In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; In the process of continuously generating alarm messages, if the number of alarm messages that meet the message merging logic reaches the merging upper limit, the alarm messages that meet the merging upper limit will be merged into an upper limit alarm package, and the upper limit alarm package will be pushed to the alarm platform.

[0045] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the IoT device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0046] The system and medium provided in the embodiments of the present application correspond one-to-one to the method. Therefore, the system and medium also have similar beneficial technical effects to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the system and medium will not be repeated here.

[0047] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0048] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0049] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0050] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0051] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0052] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0053] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0054] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0055] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for pushing alarms for AI-based drowning prevention, characterized in that: The method comprises: Based on the preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention incident; When a drowning prevention event is detected, an alarm message corresponding to a preset number of consecutive video frames is continuously generated, and the alarm type corresponding to the alarm message is determined; wherein the alarm types include: alarm for wandering in dangerous waters, alarm for entering water in dangerous waters, and alarm for fishing in dangerous waters; In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the merging upper limit, the alarm messages of the merging upper limit are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

2. According to claim 1, a method for pushing alarms for AI recognition and drowning prevention, characterized in that: The method further comprises: In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merging, and the message type of the current alarm message generated is different from the message type of several previous alarm messages, the previous alarm messages are merged into a segmented alarm package, and the segmented alarm package is pushed to the alarm platform; and, The current alarm message is used as the continuously generated initial alarm message to monitor the upper limit number of alarm message merging.

3. The method for pushing alarms for AI-based drowning prevention according to claim 1, characterized in that: The method further comprises: In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the upper limit of the number of merged messages, and the time interval between the current alarm message generated and the previous alarm message is greater than the merging time threshold, the previous alarm messages are merged into a segmented alarm package, and the segmented alarm package is pushed to the alarm platform; and, The current alarm message is used as the continuously generated initial alarm message to monitor the upper limit number of alarm message merging.

4. The method for pushing alarms for AI-based drowning prevention according to claim 1, characterized in that: The method further comprises: In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic have not reached the merging upper limit and there are no subsequent alarm messages, the previous alarm messages are merged into a termination alarm package, and the termination alarm package is pushed to the alarm platform.

5. The method for pushing alarms for AI-based drowning prevention according to claim 1, characterized in that: Before identifying and analyzing the real-time video stream based on the preset drowning prevention alarm push model, the method further includes: Obtain sample video data to be annotated, and annotate the sample video data to be annotated to extract scene feature values ​​under corresponding scenes; wherein the sample video data to be annotated includes: scene data of people wandering in dangerous waters, scene data of people entering water in dangerous waters, and scene data of fishing in dangerous waters; The scene feature values ​​are input into a preset neural network model for training until a converged anti-drowning alarm push model is obtained.

6. The method for pushing alarms for AI-based drowning prevention according to claim 5, characterized in that: The sample video data to be labeled is labeled to extract scene feature values ​​under corresponding scenes, specifically including: Label each frame of sample video data to be labeled frame by frame to determine whether there is a drowning prevention event in each frame; For video frames with drowning prevention events, identify and mark the event subject, event behavior, and event scene; wherein the event subject includes people, objects, or waters, the event behavior includes wandering, entering the water, or fishing, and the event scene is dangerous waters; Based on the labeling results, a feature extraction algorithm is used to extract scene feature values ​​in the corresponding scene; wherein the scene feature values ​​include but are not limited to personnel features, behavior features, water features and object features.

7. The method for pushing alarms for AI-based drowning prevention according to claim 1, characterized in that: After determining the alarm type corresponding to the alarm message, the method further includes: Determine the alarm level according to the alarm type, and assign a corresponding alarm level identifier to the alarm message; wherein the alarm level includes three levels: high, medium and low; The alarm level identifier is added to the alarm message.

8. The method for pushing alarms for AI-based drowning prevention according to claim 1, characterized in that: The alarm message includes: alarm time, alarm area, alarm type, alarm level identifier, alarm video and alarm picture.

9. An alarm push device for AI recognition and drowning prevention, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Based on the preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention incident; When a drowning prevention event is detected, an alarm message corresponding to a preset number of consecutive video frames is continuously generated, and the alarm type corresponding to the alarm message is determined; wherein the alarm types include: alarm for wandering in dangerous waters, alarm for entering water in dangerous waters, and alarm for fishing in dangerous waters; In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the merging upper limit, the alarm messages of the merging upper limit are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

10. A non-volatile computer storage medium for AI recognition and drowning prevention warning push, storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Based on the preset drowning prevention alarm push model, the real-time video stream is identified and analyzed to detect whether there is a drowning prevention incident; When a drowning prevention event is detected, an alarm message corresponding to a preset number of consecutive video frames is continuously generated, and the alarm type corresponding to the alarm message is determined; wherein the alarm types include: alarm for wandering in dangerous waters, alarm for entering water in dangerous waters, and alarm for fishing in dangerous waters; In the process of continuously generating alarm messages, continuously determine whether adjacent alarm messages meet the message merging logic; wherein the message merging logic is that the time interval between adjacent alarm messages is not greater than a preset merging time threshold and the alarm types of adjacent alarm messages are the same; In the process of continuously generating alarm messages, if the alarm messages that meet the message merging logic reach the merging upper limit, the alarm messages of the merging upper limit are merged into an upper limit alarm package, and the upper limit alarm package is pushed to the alarm platform.

Images