Park safety production monitoring system and monitoring method

Abstract

The invention discloses a park safety production monitoring system and method, and the method comprises the steps: collecting video data through a camera in each monitoring region in a park, determining a safety event of each monitoring region through an edge terminal according to the number of the camera, inputting the video data into a safety event detection model, and obtaining a monitoring result, and when the monitoring result is that the security event occurs in the monitoring area, the alarm information including the video clip is pushed to the server through the switch and is stored in the server, so that the security event is identified and detected through the security event detection model in the edge terminal after the video data is collected through the camera. The alarm is given when the security event is identified and detected, the security event of each monitoring area in the video does not need to be watched manually, the workload of monitoring personnel is reduced, the problem that the security event is easy to mistake and omit due to manual subjective and technical level influence is avoided, and the security event of the monitoring area can be accurately identified and detected in real time; and the intelligent degree of park safety monitoring is improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of work safety, and in particular, to a work safety monitoring system and a monitoring method for a park. Background Art

[0002] In manufacturing enterprises, in order to ensure normal and orderly production and manufacturing activities in the park, many areas need to be monitored in the park. For example, the monitored areas may include production workshops, office areas, roads, parking lots, etc., so as to conduct safety monitoring on these monitored areas.

[0003] With the development of smart parks and network technologies, in the prior art, in order to save the increasing labor costs caused by manual inspections of each monitored area and the problem of inability to conduct real-time manual inspections, cameras are installed in each monitored area, and the videos of the monitored areas are collected by the cameras and sent to the monitoring room, and safety events are discovered by manually viewing the videos in the monitoring room.

[0004] Currently, the method of using manual video viewing for work safety monitoring requires manual real-time video viewing, which increases the workload of monitoring personnel and is prone to missing safety events, with low intelligence. Summary of the Invention

[0005] Embodiments of the present invention provide a work safety monitoring system and a monitoring method for a park to solve the problems of large workload, easy omission, and low intelligence in the existing manual work safety monitoring of the park.

[0006] In a first aspect, embodiments of the present invention provide a work safety monitoring system for a park, including cameras, edge terminals, switches, and servers. The cameras are connected to the edge terminals, and the edge terminals are connected to the servers through the switches;

[0007] The number of the cameras is at least one, and the cameras are arranged in each monitored area in the park for collecting video data of the monitored area and sending the video data to the edge terminals;

[0008] The edge terminals are used for receiving the video data collected by the cameras, determining safety events in each monitored area according to the numbers of the cameras, inputting the video data into a safety event detection model to obtain a monitoring result, and when the monitoring result indicates that a safety event occurs in the monitored area, pushing alarm information including a video clip to the server through the switch;

[0009] The servers are used for storing the alarm information.

[0010] Optionally, it further includes a management terminal, and the management terminal is connected to the edge terminal through the switch;

[0011] The management terminal is used to display a configuration page, respond to user operations on the configuration page, and generate configuration instructions to be sent to the edge terminal;

[0012] The edge terminal is used to set configuration items when receiving configuration instructions, and the configuration items include at least one of configuring security events, configuring monitoring time periods, configuring alarm information, and configuring security event recognition algorithms.

[0013] Optionally, the edge terminal includes:

[0014] A video preprocessing module, which is used to receive video data and extract video images from the video data;

[0015] A security event recognition module, which is used to input the video images into a pre-trained security event recognition model to obtain the categories of various security events;

[0016] A target security event category determination module, which is used to determine the target security event based on the monitored area to which the video data belongs;

[0017] A security event judgment module, which is used to judge whether the target security event occurs in the monitored area based on the category of the security event. If so, the alarm module is executed;

[0018] An alarm module, which is used to extract a video segment including when the target security event occurs from the video data and generate alarm information including the video segment.

[0019] Optionally, the edge terminal further includes:

[0020] A first frame rate adjustment module, which is used to send a frame rate adjustment instruction to reduce the frame rate to the camera when it is determined that the target security event does not occur in the monitored area within a preset duration;

[0021] A second frame rate adjustment module, which is used to send a frame rate adjustment instruction to increase the frame rate to the camera when it is determined that the target security event occurs in the monitored area;

[0022] The camera is used to increase or decrease the acquisition frame rate of video data when receiving a frame rate adjustment instruction.

[0023] Optionally, a mobile client is further included, and the mobile client is connected to the server;

[0024] The mobile client is used to send subscription information to the server, and the subscription information is used to subscribe to the alarm information of the security event from the server;

[0025] The server is used to, after receiving the subscription information, when receiving the alarm information of the subscribed security event, push the alarm information of the security event to the mobile client;

[0026] The mobile client is further used to, in response to a user's video playback operation on the security event, generate a video playback request and send it to the server;

[0027] The server is used to respond to the video playback request, judge the user's permission, and when the user has the video playback right item, send the video clip of the security event to the mobile client.

[0028] Optionally, the server includes:

[0029] A statistical request receiving module, configured to receive a security event statistical request, where the security event statistical request includes at least one of a monitoring area, a statistical time period, a security event, a camera status, and a preset display method;

[0030] A statistical module, configured to generate a statistical result of the security event in response to the security event statistical request;

[0031] A statistical result display module, configured to display the statistical result in the display method.

[0032] Optionally, the server further includes:

[0033] An event information acquisition module, configured to acquire the event information of the security event, where the event information includes at least one of the type of the security event, the target monitoring area where it occurs, the time when the security event occurs, and the camera change;

[0034] An event description module, configured to generate a description text of the security event according to the event information;

[0035] A storage module, configured to associate and record the description text and the video clip of the security event in a database.

[0036] Optionally, the server further includes:

[0037] An event search request receiving module, configured to receive a search text of a user for a target security event;

[0038] A text matching module, configured to calculate the similarity between the search text and the description text, and determine the target security event corresponding to the description text with the highest similarity to the search text;

[0039] A target security event data sending module, configured to send the event information and video clip of the target security event to the user.

[0040] Optionally, the security event detection model is trained by the following modules:

[0041] A video data sample acquisition module, configured to acquire video data samples, where the video data samples include data when security events occur in each monitoring area, and each video data sample is labeled with at least one first event type of a security event;

[0042] A model construction module, configured to construct a security event detection model;

[0043] A video data input module, configured to randomly extract video data samples and input them into the security event detection model to obtain a second event type of a security event;

[0044] A loss value calculation module, configured to calculate a loss value by using the first event type and the second event type;

[0045] A training condition judgment module, configured to judge whether the stop training condition is satisfied;

[0046] A stop training module, configured to stop training the security event detection model;

[0047] A model parameter adjustment module, configured to adjust the model parameters of the security event detection model by using the loss value, and return to execute the video data input module.

[0048] In a second aspect, an embodiment of the present invention provides a method for monitoring the safe production of a park, which is applied to the park safe production monitoring system according to any one of the first aspects, and includes:

[0049] Collect video data of the monitoring area;

[0050] Determine the security events in each monitoring area according to the numbers of the cameras;

[0051] Input the video data into a security event detection model to obtain a monitoring result;

[0052] When the monitoring result is that a security event occurs in the monitoring area, generate an alarm message including a video clip, and save the alarm message.

[0053] The park production safety monitoring system of this embodiment includes cameras, edge terminals, switches and servers. The cameras are set in various monitoring areas in the park and are used to collect video data of the monitoring areas and send them to the edge terminals. The security events in each monitoring area are determined according to the camera number, and the video data is input into the security event detection model to obtain the monitoring result. When the monitoring result is that a security event has occurred in the monitoring area, the alarm information including the video clip is pushed to the server through the switch, and the alarm is triggered according to the alarm information and stored in the server. This achieves that after the camera collects video data, the edge terminal can determine the security events that need to be monitored in each monitoring area, and alarm when a security event is detected. There is no need to manually watch the video to detect the security events in each monitoring area, which reduces the workload of the monitoring personnel and avoids the problem of security events that are easily missed due to human subjectivity and technical influences. It can accurately identify and detect security events in each monitoring area in real time, thereby improving the intelligence level of park security monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of a park production safety monitoring system provided by an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of a park production safety monitoring system according to another embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the configuration page;

[0057] Figure 4 This is a schematic diagram of the interaction between the edge terminal and the camera when the camera frame rate is adjusted;

[0058] Figure 5 This is a schematic diagram of the interaction between the edge terminal, server, and WeChat client when the system alarm occurs;

[0059] Figure 6 A schematic diagram showing the statistical results of the security events displayed;

[0060] Figure 7 It is a flow chart of a park production safety monitoring method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0062] Figure 1Schematic diagram of the structure of a park safety production monitoring system provided by an embodiment of the present invention. Among them, the park safety production monitoring system is a system used to monitor each monitoring area in the park to detect and identify safety events. For example, Figure 1 As shown, the park safety production monitoring system includes a camera 1, an edge terminal 2, a switch 4, and a server 3. The camera 1 is connected to the edge terminal 2, and the edge terminal 2 is connected to the server 3 through the switch 4. The camera 1 can be various visible light cameras, and the number of cameras 1 is at least 1. The monitoring areas in the park can refer to the areas that need to be managed for safety events. Taking a manufacturing enterprise as an example, the monitoring areas can include the production workshop, the road areas outside the production workshop, the no-smoking areas, the office areas, the elevator interiors, the parking lots, etc. Those skilled in the art can set different monitoring areas according to the park with different functions. One or more cameras 1 can be set in one monitoring area, and at least one edge terminal 2 is set in the park. One edge terminal 2 can be connected to one or more cameras 1. A pre-trained safety event detection model can be deployed in the edge terminal 2, and this safety event detection model can detect various types of safety events through video data.

[0063] The server 3 can be a collection of one or more functional type servers. Various services or protocols can be integrated in the server 3. For example, MQTT (Message Queuing Telemetry Transport), MySQL database service, MinIO (object storage service), etc. can be integrated. Multiple edge terminals 2 are connected to the server 3 through the switch 4 for communication.

[0064] In this embodiment, the camera 1 is used to collect video data of each monitoring area in the park and send the video data to the edge terminal 2. The edge terminal 2 receives the video data collected by the camera 1, determines the safety events of each monitoring area according to the number of the camera 1, inputs the video data into the safety event detection model to obtain the monitoring result, and when the monitoring result is that a safety event occurs in the monitoring area, pushes an alarm message including a video clip to the server 3 through the switch 4. The server 3 is used to alarm and save the alarm message.

[0065] Taking the monitored area as a no-smoking area as an example, if the ID of the camera in the no-smoking area is Camera1, and the camera ID Camera1 is associated with the type of safety event "no-smoking", when the edge terminal receives video data, it can determine the camera ID Camera1 through the IP address of the video data source, and further determine the type of safety event as "no-smoking" through the camera ID Camera1. When the video data is input into the safety event detection model in the edge terminal, if there are people smoking in the no-smoking area, a safety event of the smoking type is output; if people in the no-smoking area are not smoking but people squatting on the ground are detected, a safety event of people squatting on the ground is output. However, since the type of safety event in the no-smoking area is "no-smoking", the safety event of people squatting on the ground will not trigger the determination that a safety event has occurred in the no-smoking area. When a corresponding safety event occurs in the monitored area, the edge terminal can intercept video segments of a preset duration before and after the occurrence of the safety event from the video data and send them to the server to generate an alarm message, so that the server can send the video segment to the monitoring personnel's terminal when alarming, to remind the monitoring personnel to promptly prevent the safety event behavior in the monitored area.

[0066] The park safety production monitoring system of this embodiment includes cameras, edge terminals, switches, and servers. The cameras are set in various monitored areas in the park and are used to collect video data of the monitored areas and send it to the edge terminals. The safety events of each monitored area are determined according to the camera numbers, the video data is input into the safety event detection model to obtain the monitoring result, and when the monitoring result is that a safety event has occurred in the monitored area, an alarm message including the video segment is pushed to the server through the switch, and an alarm is made according to the alarm message and stored in the server. It realizes that after the edge terminal collects video data through the camera, it can determine the safety events that need to be monitored in each monitored area, identify and detect alarms when a safety event occurs, without the need for manual viewing of the video for safety events in each monitored area, reducing the workload of the monitoring personnel, and avoiding the problem of easily missing safety events due to subjective and technical influences of manual work. It can identify and detect safety events in each monitored area in real time and accurately, improving the intelligent level of park safety monitoring.

[0067] As Figure 2 shown, in an optional embodiment, the park safety production monitoring system further includes a management terminal 5. The management terminal 5 can be connected to the edge terminal 2 through the switch 4. Among them, the management terminal 5 can be a personal computer used by the monitoring personnel. The management terminal 5 can display a configuration page, and the configuration page can be used to configure the edge terminal 2. For example, in response to the operations on the configuration page, configuration instructions are generated and sent to the edge terminal 2. The edge terminal 2 is used to set configuration items when receiving the configuration instructions. Exemplarily, the configuration items include at least one of configuring the type of safety event, configuring the monitoring time period, configuring the alarm message, and configuring the safety event recognition algorithm.

[0068] As Figure 3 shown in the schematic diagram of a configuration page, in this configuration page, the algorithms that the edge terminal needs to load can be selected. For example, a security event detection model can be selected, and the detection range can be further selected. For example, the cameras applicable to the algorithm can be further configured with the algorithm configuration details. For example, the sending quality of the detection result message, the sending interval, the screenshot interval, etc. The type of security event can also be set. For example, whether to enable the safety helmet detection. If enabled, it is determined that one of the security event types is "not wearing a safety helmet". If the electrostatic clothing detection is selected, it is determined that one of the security event types is "not wearing electrostatic clothing". For an edge terminal, when the edge terminal accesses multiple video data streams (when connecting multiple cameras), the types of security events that need to be detected and identified for the multiple video data streams can be configured separately. However, the same security event detection model can be used for the multiple video data streams. The security event detection model can detect multiple types of configured security events for all the multiple video data streams, and then determine whether there is a target security event that needs to be monitored in the monitoring area corresponding to each video data stream from the detected multiple types of security events. If so, it is determined that a target security event has occurred in the monitoring area.

[0069] Exemplarily, assume that an edge terminal accesses cameras Camera2 and Camera3. Among them, camera Camera2 collects video data 2 in the production workshop, and camera Camera3 collects video data 3 outside the production workshop (such as on the road). It can be determined that the security events in the production workshop are not wearing a safety helmet and not wearing electrostatic clothing, and the security events outside the production workshop are illegal parking and vehicle reverse. After inputting video data 3 into the security event detection model, even if it is recognized that the person is not wearing electrostatic clothing and a safety helmet, since the security events outside the production workshop (such as on the road) are illegal parking and vehicle reverse, it can be confirmed that no security event has occurred outside the production workshop (such as on the road). After inputting video data 2 into the security event detection model, if it is recognized that the person is not wearing electrostatic clothing and a safety helmet, it can be determined that a security event has occurred in the production workshop.

[0070] In this embodiment, the edge terminal is configured through the configuration page on the management terminal. Different types of security events can be configured according to different monitoring areas. On the one hand, different types of security events share a security event detection model, eliminating the need to train multiple security event detection models, reducing the model training cost and simplifying the deployment operation of the model on the edge terminal. On the other hand, corresponding security events can be flexibly set for different monitoring areas, making it applicable to different parks.

[0071] In one embodiment, the edge terminal 2 may include a video preprocessing module, a security event recognition module, a target security event category determination module, a security event judgment module, and an alarm module. The video preprocessing module is used to receive video data and extract video images from the video data. The security event recognition module is used to input the video images into a pre-trained security event recognition model to obtain the categories of various security events. The target security event category determination module is used to determine the target security event based on the monitored area to which the video data belongs. The security event judgment module is used to judge whether a target security event occurs in the monitored area based on the category of the security event. If so, the alarm module is executed. The alarm module is used to extract the video segment including the occurrence of the target security event from the video data and generate an alarm message including the video segment.

[0072] Exemplarily, after receiving the video data from camera Camera 0, the edge terminal decodes the video data to extract video images, inputs the video images into the security event recognition model, and obtains the categories of security events including Category 1, Category 2, and Category 3. Further, by determining the ID of camera Camera 0, it is determined that camera Camera 0 collects video data for monitored area A, and the pre-configured security event category for monitored area A is Category 2. Then, it can be determined that a security event occurs in the monitored area, and the video images or video segments when the security event of Category 2 occurs can be intercepted from the video data to generate an alarm message for alarm.

[0073] In one embodiment, the security event recognition model is trained by a video data sample acquisition module, a model construction module, a video data input module, a loss value calculation module, a training condition judgment module, a stop training module, and a model parameter adjustment module. Among them, the video data sample acquisition module is used to acquire video data samples, and the video data samples include the data when security events occur in each monitored area. Each video data sample is labeled with at least one first event type of security event. Exemplarily, video segments when corresponding security events occur in areas such as production workshops, offices, elevators, campus roads, parking lots, and stairwells in the park can be acquired. The model construction module is used to construct a security event detection model. The video data input module is used to randomly extract video data samples and input them into the security event detection model to obtain the second event type of security event. The loss value calculation module is used to calculate the loss value by using the first event type and the second event type (such as mean square error loss, cross-entropy loss, etc.). The training condition judgment module is used to judge whether the stop training condition is satisfied. The stop training module is used to stop training the security event detection model. The model parameter adjustment module is used to adjust the model parameters of the security event detection model by using the loss value (such as adjusting the model parameters by various gradient descent methods), and then return to execute the video data input module.

[0074] Among them, the security event detection model can adopt the YOLO series neural network model, and labels can be marked based on the types of security events. For example, labels such as human head, face, body, cigarette, smoking, smoke, fire, safety helmet, electric bicycle, car, backpack, electrostatic protective clothing, work clothes, etc. can be marked. Labels can be marked according to different types of security events, enabling the security event detection model to learn the capabilities of various marked security events through video data samples, and also making the security event detection model a general model for identifying and detecting various security events.

[0075] In one embodiment, the edge terminal 2 may further include a first frame rate adjustment module and a second frame rate adjustment module. The first frame rate adjustment module is used to send a frame rate adjustment instruction to reduce the frame rate to the camera when it is determined that no target security event occurs in the monitoring area within a preset duration. The second frame rate adjustment module is used to send a frame rate adjustment instruction to increase the frame rate to the camera when it is determined that a target security event occurs in the monitoring area. The camera is used to increase or decrease the acquisition frame rate of video data when receiving the frame rate adjustment instruction.

[0076] Exemplarily, as Figure 4 shown in the interaction schematic diagram between the camera and the edge terminal, the edge terminal pulls the video stream from the video input end (camera) for decoding and adds it to the target detection queue. Target detection video analysis is performed on the edge terminal to obtain the detection result, and the frame rate of the camera is controlled based on the detection result. For example, when no target security event configured for the monitoring area is detected within a preset duration, the camera of the monitoring area reduces the acquisition frame rate of video data to reduce the data volume of the acquired video data, save network bandwidth and the resource utilization rate of the edge terminal for processing video data. If a target security event configured for the monitoring area is detected, the camera of the monitoring area increases the acquisition frame rate of video data to track and detect the security event through high-frame-rate video data, improving the accuracy of security event detection.

[0077] As Figure 2 shown, in one embodiment, the park safety production monitoring system further includes a mobile client 6. The mobile client 6 is connected to the server 3. The mobile client 6 is used to send subscription information to the server 3. The subscription information is used to subscribe to the alarm information of security events from the server 3. The server 3 is used to, after receiving the subscription information, when receiving the alarm information of the subscribed security event, push the alarm information of the security event to the mobile client 6. The mobile client 6 is further used to generate a video play request and send it to the server 3 in response to the user's video play operation for a security event. The server 3 is used to respond to the video play request, judge the user's permission, and when the user has the video play permission, send the video segment of the security event to the mobile client 6.

[0078] As Figure 5As shown, taking WeChat on the mobile client as an example, the server can integrate BIVALE (an integrated management subsystem) and the MQTT service. When the edge terminal detects a security event in the monitored area, it first generates an alarm and pushes it to the MQTT service. If the user subscribes to the alarm information through WeChat to BIVALE, the MQTT service will push the alarm information to BIVALE. After BIVALE saves the relevant pictures in the alarm information, it will push the alarm information to WeChat. When the user receives the alarm information on WeChat, if the user needs to play the video clip when the security event occurred, WeChat generates a play request and sends it to BIVALE. BIVALE determines whether the user has the permission. If so, it sends the play request to the MQTT service. After receiving the subscribed play request, the MQTT service pulls the video clip of 5 seconds before and after the security event occurred from the edge terminal, and distributes it to BIVALE through the MQTT service. BIVALE uploads the video clip to the MinIO storage and sends the address to WeChat. WeChat connects to play the video clip through the address, enabling the user to subscribe to receive alarm information and view the video clip in real time through the mobile client (including but not limited to WeChat), and being able to timely remind the user that a security event has occurred in the monitored area.

[0079] In one embodiment, the server may include a statistical request receiving module, a statistical module, and a statistical result display module. The statistical request receiving module is used to receive a security event statistical request (such as a request sent from a management terminal or a mobile client). The security event statistical request includes at least one of a monitored area, a statistical time period, a security event, and a camera status, as well as a preset display method. The statistical module is used to generate a statistical result of the security event in response to the security event statistical request. The statistical result display module is used to display the statistical result using the display method.

[0080] As Figure 6 shown is a schematic diagram of the statistical results of security events at multiple points (monitored areas) in the park from January to July 2024. The security events include safety helmets, static electricity protective clothing, backpack-wearing personnel, smoking behavior, vehicle illegal parking, vehicle reverse driving, etc. In Figure 6 it, the statistical results of each security event are displayed through a line graph, and the statistical results of the camera online status are displayed through a pie chart, realizing the visual display of the detection results of security events.

[0081] In one embodiment, the server further includes an event information acquisition module, an event description module, and a storage module. Among them, the event information acquisition module is used to acquire event information of a security event, and the event information includes at least one of the type of the security event, the target monitoring area where it occurs, the time when the security event occurs, and the camera change. The event description module is used to generate a description text of the security event according to the event information. Exemplarily, the description text can be obtained by filling in the pre-configured template with each item of event information. The description text is used to describe the location, event, type, etc. of the security event. Of course, each item of event information can also be input into a language generation model to generate the description text. The description text and the video clip of the security event are associated and recorded in the database through the storage module.

[0082] Furthermore, the server further includes an event search request receiving module, a text matching module, and a target security event data sending module. The event search request receiving module is used to receive the search text of the user for the target security event. The text matching module is used to calculate the similarity between the search text and the description text, and determine the target security event corresponding to the description text with the greatest similarity to the search text. The target security event data sending module is used to send the event information and video clip of the target security event to the user.

[0083] Exemplarily, the user can directly input the search text in the security event search bar, or generate the search text by selecting the monitoring area and the time when the security event occurs, or generate the search text after voice input, which is convenient for the user to view the security event record and improves the efficiency of finding the security event video.

[0084] Figure 7 The flowchart of a method for monitoring production safety in a park provided by an embodiment of the present invention. The method for monitoring production safety in a park according to the embodiment of the present invention is applicable to the situation of monitoring production safety in a park. This method can be executed by a production safety monitoring system in the park, such as Figure 7 As shown, the method for monitoring production safety in a park according to the embodiment of the present invention specifically may include the following steps:

[0085] S701. Collect video data of the monitoring area.

[0086] As Figure 1 shown, the production safety monitoring system in the park includes a camera 1, an edge terminal 2, a switch 4, and a server 3. One or more cameras 1 can be set in a monitoring area in the park, and at least one edge terminal 2 is set in the park. One edge terminal 2 can be connected to one or more cameras 1. The camera can collect video data of the monitoring area at a fixed or dynamic frame rate and send the video data to the edge terminal connected to the camera.

[0087] S702. Determine the security events in each monitoring area according to the camera number.

[0088] In one embodiment, each edge terminal can be pre-configured through a configuration page. The configuration items include at least one of configuring the type of security event, configuring the monitoring time period, configuring the alarm information, and configuring the security event recognition algorithm. For example, the type of security event to be monitored in the monitoring area corresponding to a camera can be associated with the camera number. When the edge terminal receives video data, the camera number can be determined through the IP address of the video data source, and the type of security event to be detected for each video data collected by the camera can be determined through the number, so that the target security event to be monitored in the monitoring area can be determined through the type.

[0089] S703. Input the video data into the security event detection model to obtain the monitoring result.

[0090] Specifically, in the edge terminal, after receiving the video data, extract the video image from the video data, input the video image into the pre-trained security event recognition model to obtain the categories of each security event, and based on the categories of the security events, determine whether a target security event has occurred in the monitoring area. If so, execute the alarm module, which is used to extract the video segment including the occurrence of the target security event from the video data and generate the alarm information including the video segment.

[0091] S704. When the monitoring result is that a security event has occurred in the monitoring area, generate the alarm information including the video segment and save the alarm information.

[0092] In an alternative embodiment, the event information of the security event can be obtained first. The event information includes at least one of the type of the security event, the target monitoring area where it occurs, the time when the security event occurs, and the camera change. Further, generate the description text of the security event according to the event information, and associate and record the description text and the video segment of the security event in the database. When receiving the search text for the target security event from the user, calculate the similarity between the search text and the description text, determine the target security event corresponding to the description text with the largest similarity to the search text, and send the event information and video segment of the target security event to the user.

[0093] It should be noted that for the method embodiment, since it is basically similar to the embodiment of the park safety production monitoring system, the description is relatively simple. For the related parts, refer to the partial description of the embodiment of the park safety production monitoring system.

[0094] The park safety production monitoring method provided by the embodiment of the present invention is applied to the park safety production monitoring system provided by the embodiment of the present invention, so that the park safety production monitoring method has corresponding beneficial effects.

[0095] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk, or optical disc of a computer, etc., and includes several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the park safety production monitoring method of each embodiment of the present invention.

[0096] It should be noted that in the above embodiments of the park safety production monitoring system, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0097] Note that the above is only the preferred embodiment of the present invention and the applied technical principle. Those skilled in the art will understand that the present invention is not limited to the specific embodiments here, and various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A park production safety monitoring system, characterized in that: It includes a camera, an edge terminal, a switch and a server, wherein the camera is connected to the edge terminal, and the edge terminal is connected to the server through the switch; The number of the camera is at least one, and the camera is set in each monitoring area in the park, and is used to collect video data of the monitoring area and send the video data to the edge terminal; The edge terminal is used to receive video data collected by the camera, determine the security events of each monitoring area according to the number of the camera, input the video data into the security event detection model to obtain the monitoring result, and when the monitoring result is that the security event occurs in the monitoring area, push the alarm information including the video clip to the server through the switch; The server is used to store the alarm information.

2. The park production safety monitoring system according to claim 1 is characterized in that: Also included is a management terminal, the management terminal is connected to the edge terminal through the switch; The management terminal is used to display a configuration page, respond to user operations on the configuration page, generate a configuration instruction and send it to the edge terminal; The edge terminal is used to set configuration items upon receiving a configuration instruction, wherein the configuration items include at least one of configuring security events, configuring monitoring time periods, configuring alarm information, and configuring security event identification algorithms.

3. The park production safety monitoring system according to claim 1 is characterized in that: The edge terminal comprises: A video preprocessing module, used for receiving video data and extracting video images from the video data; A security event recognition module, used to input the video image into a pre-trained security event recognition model to obtain the category of each security event; A target security event category determination module, used to determine the target security event based on the monitoring area to which the video data belongs; A security event judgment module, used to judge whether the target security event occurs in the monitoring area based on the category of the security event, and if so, execute the alarm module; The alarm module is used to extract the video clip including the occurrence of the target security event from the video data, and generate alarm information including the video clip.

4. The park production safety monitoring system according to claim 3 is characterized in that: The edge terminal also includes: A first frame rate adjustment module, configured to send a frame rate adjustment instruction to reduce the frame rate to the camera when it is determined that the target security event has not occurred in the monitoring area within a preset time period; A second frame rate adjustment module is used to send a frame rate adjustment instruction to the camera to increase the frame rate when it is determined that the target security event occurs in the monitoring area; The camera is used to increase or decrease the acquisition frame rate of the video data when receiving a frame rate adjustment instruction.

5. The park production safety monitoring system according to any one of claims 1 to 4, characterized in that: Also included is a mobile client, the mobile client being connected to the server; The mobile client is used to send subscription information to the server, where the subscription information is used to subscribe to the alarm information of the security event from the server; The server is used for, after receiving the subscription information, pushing the alarm information of the subscribed security event to the mobile client when receiving the alarm information of the subscribed security event; The mobile client is also used to generate a video playback request and send it to the server in response to the user's video playback operation for the security event; The server is used to respond to the video playback request, determine the authority of the user, and send the video clip of the security event to the mobile client when the user has the video playback authority.

6. The park production safety monitoring system according to claim 5 is characterized in that: The server comprises: A statistical request receiving module, used to receive a security event statistical request, wherein the security event statistical request includes at least one of a monitoring area, a statistical time period, a security event, a camera status, and a preset display method; A statistics module, used to generate statistics results of security events in response to the security event statistics request; The statistical result display module is used to display the statistical result in the above-mentioned display mode.

7. The park production safety monitoring system according to any one of claims 1 to 4, characterized in that: The server also includes: An event information acquisition module, used to acquire event information of the security event, wherein the event information includes at least one of the type of security event, the target monitoring area where the security event occurred, the time when the security event occurred, and camera changes; An event description module, used to generate a description text of the security event according to the event information; The storage module is used to associate the description text with the video clip of the security event and record it in a database.

8. The park production safety monitoring system according to claim 7 is characterized in that: The server also includes: An event search request receiving module, used to receive a user's search text for a target security event; A text matching module, used to calculate the similarity between the search text and the description text, and determine the target security event corresponding to the description text with the greatest similarity to the search text; The target security event data sending module is used to send the event information and video clips of the target security event to the user.

9. The park production safety monitoring system according to any one of claims 1 to 4, characterized in that: The security incident detection model is trained through the following modules: A video data sample acquisition module, used to acquire video data samples, wherein the video data samples include data when security events occur in each monitoring area, and each video data sample is marked with a first event type of at least one security event; Model building module, used to build security incident detection models; A video data input module, used for randomly extracting video data samples and inputting them into the security event detection model to obtain a second event type of the security event; A loss value calculation module, configured to calculate a loss value using the first event type and the second event type; A training condition judgment module is used to judge whether the training stop condition is met; A stop training module, used to stop training the security event detection model; The model parameter adjustment module is used to adjust the model parameters of the security event detection model using the loss value and return to execute the video data input module.

10. A method for monitoring production safety in a park, characterized in that: The park production safety monitoring system according to any one of claims 1 to 9 comprises: Collecting video data of the monitoring area; Determine security events in each monitoring area based on the camera number; Inputting the video data into a security event detection model to obtain monitoring results; When the monitoring result is that the security incident occurs in the monitoring area, alarm information including a video clip is generated, and the alarm information is saved.

Images