Tunnel bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system

Abstract

The invention belongs to the technical field of urban lifeline underground space infrastructure operation and maintenance, and particularly relates to an urban lifeline operation and maintenance data and video edge computing gateway which comprises an edge computing gateway and a Web system interface. The edge computing gateway is used for data acquisition of on-site water level, smoke, temperature, humidity and the like, protocol conversion and on-site computational analysis, video acquisition, AI operation, AI model online training, background data communication and other functions; the Web system interface is used for achieving the functions of tunnel and pipe gallery on-line operation state and operation condition detection, remote control and the like. According to the method, protocol conversion, AI operation and data storage sink to the edge sides of the tunnel and the pipe gallery, so that calculation is closer to a data source, the communication data volume is reduced, the time delay of network communication and service delivery is reduced, the pressure of a monitoring background server is reduced, and the working efficiency is improved. And the key requirements in the aspects of real-time monitoring, intelligent application, reconstruction and extension, safety, privacy protection and the like of urban lifeline underground space infrastructures are met.

Description

Technical Field

[0001] The present invention relates to the field of operation and maintenance technology of urban lifeline underground space infrastructure, and specifically to a distributed storage and video edge computing system for tunnel, bridge and pipe gallery operation and maintenance data. Background Art

[0002] With the rapid development of new infrastructure, visible cities on the ground have sprung up, and invisible tunnels and pipe galleries have been quietly built. Tunnels and pipe galleries allow complex traffic routes and various pipelines to "settle down" in a unified manner, and tunnels and various municipal public pipelines such as electricity, tap water, heat, gas, and drainage are centrally laid and managed. As the infrastructure of new smart cities, tunnels and pipe galleries concentrate a large number of important communications and livelihood facilities, and their safety is of paramount importance. The resulting emerging risks of urban safety, traditional industry risks, regional risks and other problems have accumulated and multiplied, are complex and changeable, and are prone to occur frequently. Some cities have successively experienced major production safety accidents / disasters, exposing the current problems of some cities in my country where the number of underground infrastructure is unclear, overall coordination is insufficient, the early warning and disposal mechanism is not sound, and operation management is not in place.

[0003] To address these issues, the current approach is to build servers in the city construction operation and maintenance department, run monitoring platforms, and access countless sensors and terminal devices in various tunnels and pipe corridors. The monitoring platform needs to parse various communication protocols with different protocols, which will increase the amount of data and computing requirements, and require the configuration of high-performance and expensive servers. In addition, due to the limitations of network latency, bandwidth, and server AI computing power, the monitoring platform cannot perform AI operations on multiple real-time videos.

[0004] At the same time, a common solution is that the monitoring background has very high secondary development and maintenance costs for the newly connected tunnels and corridors in the later stage. When the server performance and computing power are insufficient, it is necessary to upgrade and replace new hardware facilities. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a distributed storage and video edge computing system for tunnel, bridge and pipeline corridor operation and maintenance data, so as to realize the localized processing and lightweight operation and maintenance of the operating status of urban tunnels and pipeline corridors.

[0006] Technical solution: A distributed storage and video edge computing system for tunnel, bridge and pipe gallery operation and maintenance data, including an edge computing gateway deployed in the tunnel and pipe gallery. The edge computing gateway is used for on-site data collection, protocol conversion, and video collection, and the collected data can be communicated with the background terminal and stored locally. The edge computing gateway includes an AI recognition algorithm to identify abnormal events in the tunnel or pipe gallery.

[0007] The edge computing gateway also includes the following components:

[0008] Southbound protocol library, used by the gateway to collect data from various front-end devices, including Lora, Modbus, Restful, information board, ONVIF / SDK, private protocol, and secondary development interface;

[0009] Northbound protocol library, used for gateways to forward collected data, including MQTT, 104, 61850, and GB28181;

[0010] Support components include model management, configuration center, dbapi, netdaemon, device linkage module, alarm management module, warehousing management module, video AI processing module and Web system visualization module; through model management, the basic information of the front-end equipment and its point table information that the gateway needs to collect are configured, and the equipment is added according to the template when the collection task is subsequently configured. The configuration center includes collection task configuration, linkage configuration, alarm configuration, deployment and withdrawal configuration, and warehousing configuration, including query interfaces with various configurations. When the program starts, each functional module requests the corresponding interface to query the required configuration and executes the program according to the configuration. The dbapi is used to store real-time data of the device; netdaemon is a lightweight communication protocol developed based on the publish / subscribe mode; the video AI processing module is used for the AI ​​alarm triggering of the gateway machine access camera. When water accumulation, fire, foreign objects, pedestrians, and non-motor vehicles appear in the video surveillance screen, the AI ​​alarm will be triggered, the current screen will be accessed and the AI ​​alarm event will be published to the Web system visualization module through netdaemon, and the device linkage module and alarm management will be executed; the warehousing management is used to manage the storage of monitoring volume, alarm, and video AI events.

[0011] Furthermore, the edge computing gateway is also configured with middleware for secondary development interfaces, including support for RS-485, Ethernet and LoRa interfaces. The middleware is used to convert private protocols into standard protocols, thereby satisfying the requirements for marking the data collected by existing equipment and the service years, thereby forming maintenance warnings for system architecture nodes and equipment life cycles.

[0012] Furthermore, the edge computing gateway includes an AI model online training process, so that the model can accept new data after background confirmation in actual applications and learn and optimize, so as to adapt to the ever-changing environment and needs, and maintain the timeliness and accuracy of the model. The implementation steps are as follows:

[0013] 1) When a terminal edge computing gateway on site sends an abnormal event alarm and abnormal event video to the background, the background duty personnel confirm whether the abnormal event actually occurred based on the video information;

[0014] 2) If an event is confirmed to have occurred, the video will be marked as a positive training sample. The monitoring background will send the positive training sample to all subordinate gateways in the same scene. Each node gateway receives the positive training sample, extracts the sample feature data, and uses it as the model training input data. According to the model training algorithm, the event data model is updated layer by layer from the input layer to the hidden layer and then to the output layer to optimize the AI ​​model of the event.

[0015] 3) All gateway machines in the same scenario at each node can continuously adjust the AI ​​algorithm model during operation, continuously improving the accuracy of their own AI recognition and applicability to the application environment;

[0016] The AI ​​model training algorithm receives a positive training sample, and its calculation formula is as follows:

[0017]

[0018] Among them, x i is the input value;

[0019] n is the number of feature data in the current calculation layer;

[0020] ω ij is the weight corresponding to the input value;

[0021] b ij It is the offset phasor of the direction of the input value corresponding to the output value;

[0022] y j It is the output value of the next layer of data;

[0023] If a negative training sample is received, the calculation formula is as follows:

[0024]

[0025] Because the received training samples are all from real video data of the same scene confirmed by the background, the algorithm chooses to increase or decrease the offset phasor according to the exact positive training samples and negative samples, and does not require background transfer learning, eliminating the back propagation process in the model's online learning, thereby reducing the AI ​​computing power pressure of the terminal gateway.

[0026] Furthermore, the edge computing gateway temporarily stores each sensor data in a scrollable data, adopts an algorithm for interpolating and averaging connected data, pulls the optimized data along the timeline, and uploads it to the monitoring background, thereby reducing the amount of background data and computing tasks, thereby optimizing the background storage space.

[0027] Furthermore, the edge gateway computing system configuration process includes:

[0028] (1) Create a collection task, configure basic information of the collection task, add the device corresponding to the collection task according to the device linkage module, and fill in basic information including the device address;

[0029] (2) The collection task is started, and the interface of the configuration center is called to obtain the configuration information of the current collection task;

[0030] (3) Establish communication with the front-end device, collect device data in the device linkage module, publish monitoring quantity change events through netdaemon after obtaining the device data, and store the real-time data of the device in dbapi;

[0031] (4) The alarm management module, equipment linkage module and inventory management module subscribe to monitoring quantity change events through netdaemon and execute programs according to their configurations.

[0032] Furthermore, the system controls the device linkage module / alarm management module as follows:

[0033] (1) Call the interface provided by the configuration center to obtain and save all linkage configuration information / alarm information in the current gateway system;

[0034] (2) Go to netdaemon to subscribe to telemetry and telesignaling quantity change events;

[0035] (3) After receiving the telemetry / telecommunication quantity change event, check whether there is a linkage configuration for the currently changed telemetry / telecommunication quantity. If not, ignore it directly. If yes, continue to execute step (4);

[0036] (4) Go to dbapi to query the current value of the telemetry / telecommunication quantity, compare it with the linkage configuration, and determine whether the monitoring quantity has triggered the linkage. If the current value does not trigger the linkage, ignore it directly, otherwise continue to execute step (5);

[0037] (5) Execute according to the linkage configuration, publish remote control / remote adjustment events to netdaemon, and the collection task executes the remote control / remote adjustment operation after receiving the event.

[0038] Furthermore, the system publishes the alarm event through netdaemon, including the following steps:

[0039] 1) Call the interface provided by the configuration center to obtain and save all storage configuration information in the current system;

[0040] 2) Go to netdaemon to subscribe to monitoring quantity change events, alarm events, and video AI events.

[0041] 3) After receiving the subscribed message, it is necessary to determine whether there is a storage configuration for monitoring quantity change events and alarm events. Video AI events have a storage configuration by default. If there is no storage configuration for monitoring quantity change or alarm events, it will be ignored directly. If there is, continue to step 4);

[0042] 4) According to the corresponding storage configuration, determine whether the current event needs to be stored. If not, ignore it directly. Otherwise, continue to step 5);

[0043] 5) Store the corresponding information into the database according to the storage configuration.

[0044] Furthermore, the edge computing gateway collects data including on-site water level, smoke, temperature, and humidity, and then completes AI recognition operations and online AI model training based on video acquisition through protocol conversion and on-site calculation and analysis, and communicates the processing structure with the background terminal. The background terminal realizes the online operation status and operating condition detection and remote control of the tunnel and pipeline corridor through the Web system visualization module, and stores the historical data of each node in a local distributed manner, and the feature data is stored in the terminal, which reduces the background storage pressure and optimizes network security.

[0045] In this system, the Web system visualization module adopts a front-end and back-end separation architecture design, with the front-end using the Vue framework and the back-end using the Gin Web framework developed based on Golang.

[0046] Beneficial effects: Compared with the existing technology, the edge computing system described in the present invention adopts a distributed computing architecture, which sinks protocol conversion, AI computing and data storage to the edge side of tunnels and pipe galleries, so that the computing is closer to the data source, reducing the amount of communication data, reducing the delay of network communication and service delivery, reducing the pressure on the monitoring background server, and generating faster network service response, meeting the key needs of real-time monitoring, intelligent application, renovation and expansion, security and privacy protection of the city's lifeline underground space infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a schematic diagram of the overall connection of the structure computing gateway of the present invention;

[0048] Figure 2 This is a functional module diagram of the edge computing gateway of the present invention;

[0049] Figure 3 This is a flowchart of the edge computing gateway program execution of the present invention;

[0050] Figure 4 This is a flow chart of online training of the AI ​​model during operation of the edge computing gateway of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The present invention discloses a distributed storage and video edge computing system for tunnel, bridge and pipe gallery operation and maintenance data: Figure 1 The deployment structure referred to mainly includes the edge computing gateway and the Web system for backend terminal monitoring.

[0053] The edge computing gateway is installed in underground spaces such as tunnels and pipe corridors, which are the lifelines of the city. It is responsible for on-site data collection, protocol conversion, video collection, AI computing, background data communication and other functions. The edge computing gateway collects on-site sensor data, converts the sensor data into protocols and sends it to the monitoring management background. For the collected on-site video, AI computing is performed on the real-time video, and the AI ​​alarm information and the 30-second video before and after the AI ​​abnormal event are sent to the monitoring background. The abnormal event identification and monitoring can be realized based on the camera equipped with AI recognition. The edge computing gateway is also responsible for receiving background control commands to realize remote control functions of on-site equipment such as on-site lighting, signal lights, information boards, water pumps, fans, etc. The Web system realizes the online operation status of tunnels and pipe corridors, operating environment detection and remote control functions, so as to realize visual display and dynamic management.

[0054] like Figure 2 As shown, the edge computing gateway software functions described in the present invention are mainly divided into three parts: southbound protocol library, northbound protocol library, and support functions. The southbound protocol library is mainly used for the gateway to collect data from various front-end devices; the northbound protocol library is mainly used for the gateway to forward the collected data; the support function is mainly used to implement some basic functions such as linkage and alarm inside the gateway;

[0055] The southbound protocol library mainly includes Lora, Modbus, Restful, information board, ONVIF / SDK, private protocol, and secondary development interface;

[0056] The northbound protocol library mainly includes MQTT, 104, 61850, and GB28181;

[0057] The southbound protocol communicates with front-end sensors and other devices to collect device data. After obtaining the data, it publishes data events to Netdaemon through the publish-subscribe mechanism, and decodes the data and writes it into the real-time database dbs. The northbound protocol program subscribes to data events from Netdaemon, reads the database from dbs after receiving the event, and then packages the event to the corresponding topic through the northbound protocol.

[0058] Support functions mainly include model management, unified configuration center, dbapi, netdaemon, device linkage, alarm management, warehousing management, video AI, and Web system;

[0059] The video AI module uses Python programming language, based on the YOLO series model, and uses a 6*4090 graphics card to train the AI ​​algorithm, obtain the AI ​​model of the application scenario, and use the post method to send messages. The application scenarios include water accumulation, fire, foreign objects, pedestrians, non-motor vehicles, etc.

[0060] The model management function is used to configure the basic information of the front-end devices and their point table information that the gateway needs to collect. For example, the wireless temperature and humidity sensor needs to be configured with a telesignal point (communication status) and four telemetry points (temperature, humidity, battery voltage, and remaining battery power). When configuring the collection task later, you can add devices according to the template and configure the address and other information of each device. The same information such as the device point table does not need to be added repeatedly. The configuration center includes collection task configuration, linkage configuration, alarm configuration, arming and disarming configuration, and storage configuration. The configuration center provides query interfaces for various configurations. When the program starts, each functional module requests the corresponding interface to query the required configuration and executes the program according to the configuration. For example, for the collection task, it is necessary to connect the corresponding serial port or port according to the configured channel information and communicate with the configured front-end device through this channel; dbapi is mainly used to store real-time data of the device; netdaemon is a lightweight communication protocol developed based on the publish / subscribe model; video AI is mainly used for AI alarm triggering of the gateway machine access camera. When water accumulation, fire, foreign objects, pedestrians, and non-motor vehicles appear in the video surveillance screen, the AI ​​alarm will be triggered, the current screen will be connected, and the AI ​​alarm event will be released through netdaemon;

[0061] The device linkage module is mainly used for device linkage control. The execution sequence of this module program is as follows:

[0062] 1) When the program starts, call the interface provided by the configuration center to obtain and save all linkage configuration information in the current gateway system;

[0063] 2) Go to netdaemon to subscribe to telemetry and telesignaling quantity change events;

[0064] 3) After receiving the telemetry / telecommunication quantity change event, check whether there is linkage configuration for the currently changed telemetry / telecommunication quantity. If not, ignore it directly. If yes, continue to step 4;

[0065] 4) Go to dbapi to query the current value of this telemetry / telecommunication quantity, compare it with the linkage configuration, and determine whether the monitoring quantity has triggered the linkage (whether the current time is within the deployment period associated with the linkage configuration, whether the current value matches the linkage configuration, etc.). If the current value does not trigger the linkage, ignore it directly, otherwise continue to step 5;

[0066] 5) Execute according to the linkage configuration, publish remote control / remote adjustment events to netdaemon, and the collection task executes the remote control / remote adjustment operation after receiving the event.

[0067] See also Figure 2 The alarm management module is mainly used to manage device alarm events. The execution order of this module is as follows:

[0068] 1) When the program starts, call the interface provided by the configuration center to obtain and save all the alarm configuration information in the current gateway system.

[0069] 2) Go to netdaemon to subscribe to telemetry and telesignaling quantity change events.

[0070] 3) After receiving the telemetry / telecommunication quantity change event, check whether there is an alarm configuration for the currently changed telemetry / telecommunication quantity. If not, ignore it directly. If yes, continue to step 4.

[0071] 4) Go to dbapi to query the current value of this telemetry / telecommunication quantity, compare it with the alarm configuration, and determine whether this monitoring quantity has triggered an alarm (whether the current time is within the deployment period associated with the alarm configuration, whether the current value matches the alarm configuration, etc.). If the current value does not trigger an alarm, ignore it directly, otherwise continue to step 5.

[0072] 5) Publish warning events through netdaemon;

[0073] See also Figure 2 , warehousing management is mainly used to manage the storage of monitoring volume, alarms, video AI events and other information. The execution process of this module program is as follows:

[0074] 1) When the program starts, call the interface provided by the configuration center to obtain and save all storage configuration information in the current gateway system;

[0075] 2) Go to netdaemon to subscribe to monitoring quantity change events, alarm events, and video AI events;

[0076] 3) After receiving the subscribed message, it is necessary to determine whether there is a storage configuration for monitoring quantity change events and alarm events. Video AI events have a storage configuration by default. If there is no storage configuration for monitoring quantity change or alarm events, ignore it directly. If there is, continue to step 4;

[0077] 4) According to the corresponding storage configuration, determine whether the current event needs to be stored. If not, ignore it directly. Otherwise, continue to step 5;

[0078] 5) Store the corresponding information in the database according to the storage configuration;

[0079] See also Figure 3 ,The execution process of the gateway machine software system program is as follows:

[0080] 1) Create a collection task and configure basic information of the collection task, such as channel information, etc. Add the device corresponding to the collection task according to the device template and fill in the basic information of the device (device address, etc.);

[0081] 2) The collection task is started, and the interface of the configuration center is called to obtain the configuration information of the current collection task;

[0082] 3) Establish communication with the front-end device and collect device data. After obtaining the device data, publish the monitoring quantity change event through netdaemon and store the real-time data of the device in dbapi;

[0083] 4) The alarm, linkage, and storage modules subscribe to monitoring quantity change events through netdaemon and execute programs according to the configuration.

[0084] See also Figure 4 , the process of the AI ​​model online training module during the operation of the gateway machine is as follows:

[0085] 1) When a terminal edge computing gateway on site sends an abnormal event alarm and abnormal event video to the background, the background duty personnel can confirm whether the abnormal event actually occurred based on the video information;

[0086] 2) If an event is confirmed to have occurred, the video will be marked as a positive training sample. The monitoring background will send the positive training sample to all subordinate gateways in the same scene. Each node gateway receives the positive training sample, extracts the sample feature data, and uses it as the model training input data. According to the model training algorithm, the event data model is updated layer by layer from the input layer to the hidden layer and then to the output layer to optimize the AI ​​model of the event.

[0087] 3) Therefore, all gateway machines in the same scenario at each node can continuously adjust the AI ​​algorithm model during operation, and continuously improve the accuracy of their own AI recognition and applicability to the application environment.

[0088] In some embodiments, the model training algorithm receives a positive training sample, and the calculation formula is as follows:

[0090] is the offset phasor of the input value corresponding to the output value direction; y jIt is the output value of the next layer of data;

[0091] If a negative training sample is received, the calculation formula is as follows:

[0092]

[0093] 4) Update AI algorithm model data.

[0094] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A distributed storage and video edge computing system for tunnel, bridge and pipe gallery operation and maintenance data, characterized in that: It includes an edge computing gateway deployed in tunnels and pipe corridors. The edge computing gateway is used for on-site data collection, protocol conversion, and video collection, and the collected data can be communicated with the backend terminal and stored locally. The edge computing gateway includes an AI recognition algorithm to identify abnormal events in tunnels or pipe corridors. The edge computing gateway also includes the following components: Southbound protocol library, used to collect data from various front-end devices, including Lora, Modbus, Restful, information board, ONVIF / SDK, private protocol, and secondary development interface; Northbound protocol library, used to forward collected data upward, including MQTT, 104, 61850, and GB28181; Support components include model management, configuration center, dbapi, netdaemon, device linkage module, alarm management module, warehousing management module, video AI processing module and Web system visualization module; the basic information of the front-end equipment and its point table information collected through model management, and the equipment is added according to the template when the collection task is subsequently configured. The configuration center includes collection task configuration, linkage configuration, alarm configuration, deployment and withdrawal configuration, and warehousing configuration, including query interfaces with various configurations. When the program starts, each functional module requests the corresponding interface to query the required configuration and executes the program according to the configuration. The dbapi is used to store real-time data of the device; netdaemon is a lightweight communication protocol developed based on the publish / subscribe model; the video AI processing module is triggered based on the abnormal event of the camera. When water accumulation, fire, foreign objects, pedestrians, and non-motor vehicles appear in the video surveillance screen, the AI ​​alarm will be triggered, the current screen will be connected and the AI ​​alarm event will be published to the Web system visualization module through netdaemon, and the device linkage module and alarm management will be executed; the warehousing management is used to manage the storage of monitoring volume, alarm, and video AI events.

2. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The edge computing gateway is also equipped with middleware for secondary development interfaces, including support for RS-485, Ethernet and LoRa interfaces. The middleware is used to convert private protocols into standard protocols, thereby marking the data collected by existing equipment and the service life, thereby forming maintenance warnings for system architecture nodes and equipment life cycles.

3. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The edge computing gateway includes an AI model online training process, which enables the model to accept new data confirmed by the background and learn and optimize in actual applications, so as to adapt to the ever-changing environment and needs, and maintain the timeliness and accuracy of the model. The implementation steps are as follows: 1) When a terminal edge computing gateway on site sends an abnormal event alarm and abnormal event video to the background, the background duty personnel confirm whether the abnormal event actually occurred based on the video information; 2) If an event is confirmed to have occurred, the video will be marked as a positive training sample. The monitoring background will send the positive training sample to all subordinate gateways in the same scene. Each node gateway receives the positive training sample, extracts the sample feature data, and uses it as the model training input data. According to the model training algorithm, the event data model is updated layer by layer from the input layer to the hidden layer and then to the output layer to optimize the AI ​​model of the event. 3) All gateway machines in the same scenario at each node can continuously adjust the AI ​​algorithm model during operation, continuously improving the accuracy of their own AI recognition and applicability to the application environment; The AI ​​model training algorithm receives a positive training sample, and its calculation formula is as follows: Among them, x i is the input value; n is the number of feature data in the current calculation layer; ω ij is the weight corresponding to the input value; b ij It is the offset phasor of the direction of the input value corresponding to the output value; y j It is the output value of the next layer of data; If a negative training sample is received, the calculation formula is as follows: Because the received training samples are all from real video data of the same scene confirmed by the background, the algorithm chooses to increase or decrease the offset phasor according to the exact positive training samples and negative samples, and does not require background transfer learning, eliminating the back propagation process in the model's online learning, thereby reducing the AI ​​computing power pressure of the terminal gateway.

4. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The edge computing gateway temporarily stores each sensor data in a scrollable data, uses an algorithm for interpolating and averaging connected data, pulls the optimized data along the timeline, and uploads it to the monitoring background, reducing the amount of background data and computing tasks, thereby optimizing the background storage space.

5. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The edge gateway computing system configuration process includes: (1) Create a collection task, configure basic information of the collection task, add the device corresponding to the collection task according to the device linkage module, and fill in basic information including the device address; (2) The collection task is started, and the interface of the configuration center is called to obtain the configuration information of the current collection task; (3) Establish communication with the front-end device, collect device data in the device linkage module, publish monitoring quantity change events through netdaemon after obtaining the device data, and store the real-time data of the device in dbapi; (4) The alarm management module, equipment linkage module and inventory management module subscribe to monitoring quantity change events through netdaemon and execute programs according to their configurations.

6. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The control method for the device linkage module / alarm management module is as follows: (1) Call the interface provided by the configuration center to obtain and save all linkage configuration information / alarm information in the current gateway system; (2) Go to netdaemon to subscribe to telemetry and telesignaling quantity change events; (3) After receiving the telemetry / telecommunication quantity change event, check whether there is a linkage configuration for the currently changed telemetry / telecommunication quantity. If not, ignore it directly. If yes, continue to execute step (4); (4) Go to dbapi to query the current value of the telemetry / telecommunication quantity, compare it with the linkage configuration, and determine whether the monitoring quantity has triggered the linkage. If the current value does not trigger the linkage, ignore it directly, otherwise continue to execute step (5); (5) Execute according to the linkage configuration, publish remote control / remote adjustment events to netdaemon, and the collection task executes the remote control / remote adjustment operation after receiving the event.

7. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The steps to publish an alarm event through netdaemon include: 1) Call the interface provided by the configuration center to obtain and save all storage configuration information in the current system; 2) Go to netdaemon to subscribe to monitoring quantity change events, alarm events, and video AI events. 3) After receiving the subscribed message, it is necessary to determine whether there is a storage configuration for monitoring quantity change events and alarm events. Video AI events have a storage configuration by default. If there is no storage configuration for monitoring quantity change or alarm events, it will be ignored directly. If there is, continue to step 4); 4) According to the corresponding storage configuration, determine whether the current event needs to be stored. If not, ignore it directly. Otherwise, continue to step 5); 5) Store the corresponding information into the database according to the storage configuration.

8. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The edge computing gateway collects data including on-site water level, smoke, temperature, and humidity, and then completes AI recognition operations and online AI model training based on video acquisition through protocol conversion and on-site calculation and analysis, and communicates the processing structure with the background terminal. The background terminal realizes the online operation status and operating condition detection and remote control of the tunnel and pipeline corridor through the Web system visualization module, and stores the historical data of each node in a local distributed manner, and the feature data is stored in the terminal, which reduces the background storage pressure and optimizes network security.

9. The tunnel, bridge and pipe gallery operation and maintenance data distributed storage and video edge computing system according to claim 1 is characterized in that: The Web system visualization module adopts a front-end and back-end separation architecture design, with the front-end using the Vue framework and the back-end using the Gin Web framework developed based on Golang.

Images