Edge computing gateway applied to target detection and video recording
Through the edge computing gateway, video and gas data are processed in real time in the oil tank and uploaded to the cloud, the problems of real-time monitoring and data management in oil tank cleaning are solved, and efficient video and gas data monitoring and management are achieved.
Patent Information
- Application Number
- CN202420069017.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Utility models(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-11
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2034-01-11
AI Technical Summary
During the cleaning of the oil tank, it is difficult for the existing technology to realize real-time monitoring and recording of gas concentration and video data in the tank, and the equipment is cumbersome to install and cannot be uploaded to the server in time, resulting in data being unable to be obtained and managed in a timely manner.
Design an edge computing gateway for object detection and video recording, use 4G/5G communication module and WiFi module to achieve connection with the cloud platform, integrate AI object detection module and solid-state drive, and process video and gas data in real time through edge computing box. And upload it to the cloud, supporting real-time video viewing and gas concentration monitoring.
Realize real-time object detection and video recording functions, supports 6-channel video data inference, video delay is less than 2 seconds, can view gas detection instrument data in real time, data is saved in the edge computing box, saves traffic costs, and uploads videos at important moments to the server.
Smart Images

Figure CN222107955U_ABST
Abstract
Description
Technical Field
[0001] The utility model relates to the technical field of oil tank monitoring, and in particular to an edge computing gateway applied to target detection and video recording. Background Art
[0002] Oil storage tanks are a type of enclosed space, referred to as oil tanks or storage tanks, used to store oil (crude oil or other petroleum products). These large containers are mostly made of steel and are widely used in refineries, oil fields, oil depots and other industries. Oil tanks are usually large in size, ranging from 1,000 cubic meters to 10,000 cubic meters.
[0003] During the use of oil tanks, they are cleaned regularly. However, due to the special properties of oil tanks, toxic and harmful gases may be generated during the cleaning process, posing a risk of accidental injury to cleaning personnel. In order to ensure safety, in addition to requiring cleaning personnel to be equipped with protective equipment, close monitoring is also required inside and outside the tank. The concentration of toxic and harmful gases around the cleaning personnel needs to be monitored in real time. Once the concentration exceeds the safety threshold, the cleaning personnel should be notified immediately to evacuate. If personal injury occurs, the monitoring personnel outside the tank should be notified immediately so that rescue can be carried out in time.
[0004] In addition, since the oil tank is a metal enclosed space that will shield wireless signals such as 4G and WiFi, communication inside and outside the tank needs to rely on wired methods. However, since it is impossible to introduce wired power, the instruments and equipment inside the tank usually rely on batteries for power supply. At the same time, the residual gas in the oil tank may be flammable and explosive, so the instruments and equipment inside the tank must have explosion-proof functions.
[0005] In view of the above situation, the current solution usually includes the following steps: the cleaning personnel wear gas detectors to monitor the surrounding gas concentration. However, the gas data can only be saved on the detector, and the personnel outside the tank cannot obtain this data. A network camera needs to be installed in the tank to record real-time video. The personnel outside the tank can understand the status of the cleaning personnel in the tank by viewing the real-time video. This series of equipment is relatively large, and it takes a certain amount of time to install and deploy. Gas data and video data cannot be uploaded to the server in time. In addition, gas concentration data and video data are usually stored separately, which is difficult to match one to one.
[0006] To address these issues, the following improvements can be considered: Use Bluetooth or WiFi signals to transmit data from the gas detector to a mobile phone or tablet so that personnel outside the tank can obtain gas concentration data in real time. Use solar-powered automatic cameras to replace cameras that require wired power, which can reduce the time and cost of installation and deployment. Use 5G or more advanced network technology to upload real-time video and gas data to a cloud server for unified management and analysis. Develop an integrated system that associates and stores gas concentration data with video data for easier retrieval and analysis.
[0007] The utility model patent with application number 202222236926.3 discloses an edge computing intelligent visual gateway, including: a main board, which includes: a network communication module for receiving monitoring data sent by an external perception device that is communicatively connected to the edge computing intelligent visual gateway; a main processor, connected to the network communication module, for performing clarity enhancement, AI analysis, encoding and decoding, storage and forwarding on the monitoring data; a data interface module, connected to the main processor, for outputting the monitoring data processed by the main processor to a terminal; a slave device, connected to the main board through an interface, including: an onboard processor, installed on the slave device, equipped with a memory for temporarily storing data; an AI computing power module, including multiple AI computing power chips, all connected to the onboard processor, for providing an AI reasoning model, performing scene segmentation, target detection, behavior recognition and anomaly recognition on the monitoring data; a data processing module, connected to the onboard processor, for data processing and data encoding and decoding; a memory, connected to the onboard processor, for long-term data storage. The above patent provides a rich network communication mode and I / O interface, making the application scenarios of the edge computing intelligent visual gateway wider; there are multiple interfaces on the mainboard, which can connect multiple slave devices. The data calculation of the edge computing intelligent visual gateway is mainly concentrated in the slave device. The slave device uses the SOC architecture to integrate the hardware modules, which takes up little space and is cost-effective; different functional hardware can be equipped for the slave device according to the usage scenario, with high integration and strong versatility; by setting multiple AI computing chips on the slave device, higher AI computing power is provided to accelerate the calculation. However, the above patent is complex in calculation and cannot be applied to oil tank cleaning. Utility Model Content
[0008] In view of the technical problems that the existing oil tank cleaning requires a lot of equipment and the installation is complicated, the utility model proposes an edge computing gateway for target detection and video recording, which can realize the functions of helmet wearing detection and tank body video recording. Through the WiFi module and 4G / 5G communication module, it can achieve connection with peripheral devices and mobile terminals, reducing the installation steps.
[0009] In order to achieve the above-mentioned purpose, the technical solution of the utility model is implemented as follows: an edge computing gateway applied to target detection and video recording, including an edge computing box, the edge computing box is connected to a router, the router is respectively connected to a camera and a gas detector, the camera and the gas detector are both arranged in a tank, the edge computing box is connected to a cloud platform through a base station, and the edge computing box is connected to a mobile terminal; the edge computing box includes a CPU, and the CPU is respectively connected to an AI target detection module and a solid-state hard disk.
[0010] Preferably, the edge computing box is provided with a first network port and a second network port, both of which are connected to the CPU, the first network port is connected to the camera through a router, and the second network port is connected to the gas detector through a router.
[0011] Preferably, the edge computing box is provided with a 4G / 5G communication module and a WIFI module, both the 4G / 5G communication module and the WIFI module are connected to the CPU, the 4G / 5G communication module is connected to the base station, the WIFI module is connected to the antenna, and the antenna is connected to the mobile terminal.
[0012] Preferably, the 4G / 5G communication module is connected to the 4G amplifier, the 4G amplifier is connected to the base station, and the 4G amplifier is installed in the tank.
[0013] Preferably, a WiFi AP hotspot is provided on the WIFI module.
[0014] Preferably, the CPU is equipped with an eight-core ARM processor, and the ARM processor is integrated with a GPU and an NPU.
[0015] Preferably, the CPU is integrated with an NPU with a computing power of 6 TOPS, and the NPU is provided with an AI target detection module.
[0016] Preferably, the GPU is used to decode the video data collected by the camera into a single-frame image, and encode the single-frame image with the target result processed by the AI target detection module into a video stream.
[0017] Preferably, the AI target detection module is a Yolov5 target detection model, and the video stream is pushed to a streaming media server using the RTMP streaming media protocol.
[0018] Preferably, the CPU is connected to a rechargeable lithium battery, and the rechargeable lithium battery is installed in an iron box, which is sealed with sealing glue.
[0019] Compared with the prior art, the utility model has the following beneficial effects: it can perform target detection reasoning of 6-channel video data in real time, with a single-frame reasoning speed of 20ms, and has AI detection effect; it supports watching 6-channel real-time videos at the same time, and the video delay is less than 2s, with real-time video viewing effect; the video data of the network camera can be saved in the edge computing box, and a 1T solid-state hard drive can save about 400 hours of video, achieving real-time recording effect; it allows real-time viewing of the data of the gas detection instrument, so as to timely grasp the gas concentration data of the closed environment. The utility model is powered by batteries to meet the user's mobile needs and meet the user's use scenarios where the mains cannot be accessed; the video data can be saved, and the usual videos can be saved in the edge computing box, and only the important moment videos can be uploaded to the upper-level video server, which can save the traffic cost caused by video transmission; the gas concentration data is superimposed on the video screen, and the gas concentration can be monitored while monitoring the video. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the embodiments of the utility model or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the utility model. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 It is a structural schematic diagram of the utility model.
[0022] Figure 2 for Figure 1 Schematic diagram of the edge computing box shown.
[0023] Figure 3 The figure is a flow chart of helmet detection performed by the AI target detection module of the present invention.
[0024] In the figure, 1 is an edge computing box, 2 is a camera, 3 is a router, 4 is a gas detector, 5 is a mobile terminal, 6 is a cloud platform, 7 is a 4G amplifier, 8 is a base station, 11 is a CPU, 12 is a solid-state drive, 13 is an AI target detection module, 14 is a 4G / 5G communication module, 15 is a first network port, 16 is a second network port, and 17 is a WIFI module. DETAILED DESCRIPTION
[0025] The following will be combined with the drawings in the embodiments of the utility model to clearly and completely describe the technical solutions in the embodiments of the utility model. Obviously, the described embodiments are only part of the embodiments of the utility model, not all of the embodiments. Based on the embodiments in the utility model, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the utility model.
[0026] like Figure 1 As shown, an edge computing gateway for target detection and video recording includes an edge computing box 1, which is connected to a router 3, which is respectively connected to a camera 2 and a gas detector 4, which are both arranged in a tank body, and the edge computing box 1 is connected to a cloud platform 6 through a base station 8, and the edge computing box 1 is connected to a mobile terminal 5; the edge computing box 1 includes a housing, a mainboard is arranged in the housing, a CPU 11 is arranged on the mainboard, and the CPU 11 is respectively connected to an AI target detection module 13 and a solid state drive 12. The camera 2 is a network camera, and the AI target detection module 13 realizes automatic detection of helmet wearing problems through artificial intelligence technology. The camera 2 deployed in the tank collects videos in the tank body, and the AI target detection module 13 can monitor in real time whether the employee wears a helmet. When it is detected that the helmet is not worn, the CPU 11 converts the information into a control instruction and transmits it to the alarm module, the mobile terminal 5 and the cloud platform 6, thereby automatically alarming and notifying relevant personnel to handle it in time. The utility model takes the edge computing box as the center, and other peripheral devices are connected to the edge computing box through a shared Wi-Fi hotspot; up to 6 network cameras can be connected; up to 32 gas detection instruments can be connected; a laptop computer can be used to connect to the edge computing box; the application server, namely the cloud platform 6, is connected through a 4G network to save gas data and alarm data, and upload video data to the superior video server.
[0027] The utility model transmits the video screen of the network camera in real time, and the staff can access the live video through the Internet at any time and any place to understand the situation inside the tank. In addition, the manager can also remotely monitor the situation inside the tank through a mobile device or computer to achieve real-time management and supervision. The video screen of the camera 2 is transmitted to the outside of the tank in time and saved in the edge computing box 1. The video storage time can be set as needed to ensure the integrity and traceability of the video data. In addition, by playing back the video, the cause of the accident can be traced, the evidence can be verified, and other operations can be performed. The edge computing box 1 integrates a gateway function, which can transmit the gas concentration data of the gas detector 4 configured by the cleaning personnel in the tank to the outside of the tank in time, and upload it to the server of the cloud platform 6 for unified management and analysis. Through the gateway function, the manager can grasp the gas concentration in the tank in real time and take corresponding measures in time. The gas concentration data can be displayed in real time on the video screen, so that the observer can view the video and gas concentration information at the same time. In this way, the gas concentration in the tank can be more intuitively understood and countermeasures can be taken in time.
[0028] like Figure 2 As shown, the edge computing box 1 is provided with a first network port 15 and a second network port 16, both of which are connected to the CPU 11, the first network port 5 is connected to the camera 2 through the router 3, and the second network port 16 is connected to the gas detector 4 through the router 3. The utility model is designed to connect camera equipment, provide customers with video sources, and access gas detection instrument equipment through two 1Gb Ethernet ports. The camera 2 only needs to support the RTSP (Real Time Streaming Protocol) video transmission protocol, which is currently supported by all network cameras on the market. The gas detector is Hanwei portable four-gas detector E4000, E6000, and can support other models of gas detectors through docking protocols.
[0029] The edge computing box 1 is provided with a 4G / 5G communication module 14 and a WIFI module 17, both of which are connected to the CPU, the 4G / 5G communication module 14 is connected to the base station 8, the WIFI module 17 is connected to the antenna, and the antenna is connected to the mobile terminal 5. The 4G / 5G communication module 14 uses EC20 of Quectel Communications, and the model of the WIFI module 17 is AP6212. Through the interface, the CPU 11 is connected to the 4G / 5G network module to ensure that the device can access the mobile Internet and connect to the upper-level cloud platform 6, i.e., the cloud server. The WIFI module 17 is provided with a WiFi AP hotspot. The WIFI module 17 carried on the CPU 11 provides a WiFi AP hotspot, which facilitates the access of surrounding gas meter equipment and network camera equipment to the edge computing box. The interfaces on the CPU 11 mainboard include network ports, WiFi, and 4G interfaces, which facilitate the access and connection of network camera equipment, gas detectors, and upper-level platforms.
[0030] The utility model adopts Rockchip's rk3588 CPU as the core processor, and is equipped with 2 1000M Ethernet ports, 1 72M bandwidth WIFI module 17, 4G / 5G communication module 14, 1T solid-state hard disk 12 and other peripheral devices. The CPU runs the Ubuntu operating system and is powered by a battery. The CPU is equipped with an eight-core ARM 64-bit processor, an integrated ARM GPU and an NPU with 6 TOPS computing power, and its memory can be expanded to 32GB at most. The device motherboard is equipped with a CPU responsible for collecting, processing, storing and transmitting video data and gas concentration data. A solid-state hard disk 12 is installed on the CPU through an interface to store video data.
[0031] The CPU 11 of the utility model is equipped with an eight-core ARM processor, and the ARM processor is integrated with a GPU and an NPU. The CPU 11 is integrated with an NPU with a computing power of 6 TOPS, and the NPU is provided with an AI target detection module 13. The GPU is used to decode the video data collected by the camera 2 into a single-frame image, and encode the single-frame image with the target result processed by the AI target detection module 13 into a video stream. The AI target detection module 13 is a Yolov5 target detection model, and the video stream is pushed to the streaming media server using the RTMP streaming media protocol. The NPU module inside the CPU is endowed with powerful AI reasoning capabilities, which can realize functions such as target detection, behavior recognition and anomaly detection.
[0032] The edge computing box of this utility model has the following functions:
[0033] 1. Integrate AI functions to detect helmet wearing: The network camera is connected to the router 3 via a network cable, and the edge computing box is connected to the same router 3 via wifi, so that the edge computing box 1 can connect to the network camera 2 via wifi. Collect pictures through camera 2 and train the model on the Yolov5 framework, convert the trained model into rknn format, obtain the video stream from the network camera 2, obtain the video data of each frame after decoding, use the video data as input for target detection reasoning, obtain the location coordinates, type and confidence of the target, and draw this information on the video screen, and then re-encode it into a video stream and upload it to the streaming media server, such as Figure 3 During the helmet detection process, if someone is not wearing a helmet, it will be reported to the CPU immediately. The management personnel can make corrections and management in time based on this information.
[0034] The CPU 11 of the edge computing box uses the rtsp protocol to extract video data from the network camera, and then sends the video data to the GPU to decode it into a single-frame image, and then passes the image to the yolov5 target detection model. The yolov5 target detection model will perform reasoning on the NPU, output the position coordinates of the helmet in the image represented by a box, and then draw the target result on the image. Then the single-frame image is sent to the GPU module and encoded into a video stream, and then the video stream is pushed to the streaming server using the RTMP streaming protocol. Yolov5 is a ready-made open source target detection algorithm, and a target detection model file can be trained using yolov5. The NPU reasoning function calls the rockchip rknpu sdk interface to load the target detection model file before reasoning.
[0035] The specific implementation steps of helmet detection are as follows: Collect image or video data containing whether workers are wearing helmets in various scenarios. These data should cover different lighting conditions, backgrounds, head postures, and helmet types. Annotate the data, including cases of wearing helmets and not wearing helmets. Annotations should include bounding boxes and corresponding category labels. Clean and preprocess the collected data, such as resizing, normalizing, and removing noise. Data augmentation can be used to increase the diversity of training samples, such as randomly rotating, flipping, scaling, and cropping images. Choose a deep learning model suitable for object detection, such as YOLO (YOLOv3, YOLOv4, or YOLOv5) or SSD. Consider the accuracy, speed, and resource requirements of the model, and make trade-offs based on the actual application scenario. Use the annotated dataset to train the selected object detection model. Adjust the model's hyperparameters, such as learning rate, batch size, optimizer, etc., to obtain the best performance. Multiple iterations and tuning may be required, including adjusting the network structure and training strategy. Use an independent validation set to evaluate the performance of the model. Common evaluation indicators include mean average precision (mAP), recall, and precision. Adjust the model and training strategy based on the evaluation results to ensure high detection accuracy when wearing and not wearing a helmet. Integrate the trained model into the CPU to process the image or video stream input by the camera in real time. Implement an efficient inference engine that can quickly detect helmets for each frame of video with the help of NPU inference acceleration.
[0036] Design an intuitive and easy-to-use user interface to display real-time video streams and gas concentration detection results. When a helmet is not worn, the user can be reminded through visual cues (such as highlight boxes, warning icons, etc.) or sound alarms. Integrate the yolov5 target detection model with other necessary functions (such as video storage, alarm notification, statistical analysis, etc.) into a complete software system, and deploy the system on the edge computing box 1 for performance optimization and resource management.
[0037] The utility model can integrate AI detection function into the network camera, and detect the wearing of the helmet while the camera generates video. If it is found that the helmet is not worn, an alarm message will be issued immediately. The utility model can also adopt other target detection model technology frameworks, including YOLO, SSD, Fast R-CNN and MobileNet target detection models. These technical frameworks have efficient target detection capabilities and can meet the needs of different application scenarios.
[0038] 2. Integrated streaming media server function to meet the needs of users outside the tank to view the video inside the tank in real time.
[0039] The network camera is connected to the edge computing box through a router. The CPU in the edge computing box extracts the video stream from the network camera using the RTSP protocol, and then pushes the stream directly to the SRS streaming server without encoding or decoding to ensure the real-time nature of the video. SRS is a simple and efficient real-time video server (Simple Realtime Server) that is open source. The official website is https: / / ossrs.net / lts / zh-cn / . The user's laptop is connected to the edge computing box through a wifi network, and the streaming live page is opened on the browser. Then, you can click on the page to view the live video of the camera.
[0040] Integrate SRS streaming media server in the edge computing box system to realize live video broadcast. SRS streaming media server is an open source streaming media server that supports multiple streaming protocols such as RTMP, HLS, HTTP-FLV, etc. It has rich functions such as push streaming, pull streaming, transcoding, recording, forwarding, etc., and has the characteristics of high performance and low latency. Using SRS streaming media server, you can build a streaming media server to realize real-time transmission and playback of audio and video.
[0041] The architecture of the SRS (Simple Realtime Server) streaming media server includes:
[0042] Main Thread: Responsible for initializing server configuration, listening to network ports, processing semaphores, and managing child threads. Worker Threads Pool: Use the worker thread pool to handle client connections and data streams. Each worker thread can handle multiple connections, and improve concurrency performance through efficient event loops and non-blocking I / O. RTMP Handler: One of the main protocols supported by SRS, responsible for parsing and processing RTMP protocol messages, including handshake, connection, publishing, playback, heartbeat and other operations. HLS / HTTP-FLV Handler: Responsible for converting live streams to HLS or HTTP-FLV formats for playback on different clients and network environments. DVR Module: Provides recording and playback functions, can save live content as files, and supports retrieval and playback by timestamp. FFMPEG Transcoder: The optional transcoding module uses the FFMPEG library to transcode, filter, and otherwise process video and audio to accommodate different encoding formats and bit rate requirements. HTTP API Handler: Provides HTTP interface, allowing external systems to query server status, control streaming services, or obtain statistical data through API. Logging and Statistics: SRS includes detailed logging and performance statistics functions, which can help administrators monitor server operation status and diagnose problems.
[0043] 3. Realize gateway function: The edge computing box has three network interfaces: Ethernet interface, with 2 1G network interfaces, which can easily access the local area network; with wifi network interface, it can access the existing wifi network and provide wifi hotspots for other devices; 4G and 5G networks can access the mobile Internet. With the support of these interfaces, the edge computing box 1 can be used as an edge computing gateway to conveniently forward and transmit the data of gas detection instruments and network camera monitoring data to the superior cloud platform, which can save and display gas data to users. The gas detector configured by the cleaning personnel in the tank can timely transmit the data outside the tank and upload it to the cloud platform. The gas detector worn by the personnel can detect the concentration of toxic and harmful gases such as carbon monoxide and hydrogen sulfide, and access the edge computing gateway through the wifi network. The gateway parses the data and forwards it to the cloud platform through the 4G network.
[0044] 4. Gas data superimposed on the video screen: Superimpose the gas concentration data on the video screen so that the gas concentration can be viewed while watching the video. After receiving the gas concentration data, the edge computing box will call the system interface to draw the characters of the gas concentration data on each frame of the video data and save the data as a video. At the same time, the data will also be transmitted to the SRS streaming server for real-time live broadcast.
[0045] The video footage from the camera inside the tank will be transmitted to the outside of the tank and saved in the gateway. The video storage time can be set according to actual needs. In addition, the video data will be saved according to the HLS protocol, specifically saving the M3U8 video list and TS video file for video playback and video download, which can provide more flexible and efficient video management and use. The user's laptop is connected to the edge computing gateway through the wifi network.
[0046] CPU 11 is connected to a rechargeable lithium battery, which is installed in an iron box and sealed with glue. In order to meet the needs of mobility and normal use in special environments without mains power supply, a lithium battery is used for power supply. The lithium battery has explosion-proof characteristics and can ensure that the device can work continuously for more than 8 hours. CPU 11 is equipped with a 12000mah rechargeable lithium battery with a voltage of 12V, which ensures that the device can run for 8 hours. In order to meet the explosion-proof requirements, the battery will be installed in an iron box and sealed with glue.
[0047] The 4G / 5G communication module 14 is connected to the 4G amplifier 7, the 4G amplifier 7 is connected to the base station 8, and the 4G amplifier 7 is installed in the tank. The utility model can also use a 4G signal amplifier to introduce a 4G signal outside the tank into the tank to achieve wireless communication. By installing a 4G signal amplifier in the tank, it can be ensured that the mobile devices carried by the personnel in the tank use the mobile Internet for network communication. In this way, even in a closed tank environment, communication with the outside can be maintained to meet various communication needs.
[0048] The core GPU of the utility model can also be replaced with a CPU similar to RK3588, thereby providing a more flexible hardware configuration solution. This design enables the device to maintain efficient computing speed and stable operation when processing large amounts of data. The utility model also supports multiple network protocols, such as RTMP, HLS, HTTP-FLV, etc., to adapt to different streaming media transmission requirements.
[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. An edge computing gateway for target detection and video recording, characterized in that: The invention comprises an edge computing box (1), wherein the edge computing box (1) is connected to a router (3), wherein the router (3) is respectively connected to a camera (2) and a gas detector (4), wherein the camera (2) and the gas detector (4) are both arranged in a tank body, wherein the edge computing box (1) is connected to a cloud platform (6) via a base station (8), and wherein the edge computing box (1) is connected to a mobile terminal (5); wherein the edge computing box (1) comprises a CPU (11), wherein the CPU (11) is respectively connected to an AI target detection module (13) and a solid state drive (12).
2. The edge computing gateway for target detection and video recording according to claim 1, characterized in that: The edge computing box (1) is provided with a first network port (15) and a second network port (16), both of which are connected to the CPU (11), the first network port (15) is connected to the camera (2) via a router (3), and the second network port (16) is connected to the gas detector (4) via the router (3).
3. The edge computing gateway for target detection and video recording according to claim 2, characterized in that: The edge computing box (1) is provided with a 4G / 5G communication module (14) and a WIFI module (17); the 4G / 5G communication module (14) and the WIFI module (17) are both connected to the CPU; the 4G / 5G communication module (14) is connected to the base station (8); the WIFI module (17) is connected to the antenna; and the antenna is connected to the mobile terminal (5).
4. The edge computing gateway for target detection and video recording according to claim 3, characterized in that: The 4G / 5G communication module (14) is connected to the 4G amplifier (7), the 4G amplifier (7) is connected to the base station (8), and the 4G amplifier (7) is installed in the tank.
5. The edge computing gateway for target detection and video recording according to claim 3, characterized in that: The WIFI module (17) is provided with a WiFi AP hotspot.
6. The edge computing gateway for target detection and video recording according to any one of claims 1 to 5, characterized in that: The CPU (11) is equipped with an eight-core ARM processor, and the ARM processor is integrated with a GPU and an NPU.
7. The edge computing gateway for target detection and video recording according to claim 6, characterized in that: The CPU (11) is integrated with an NPU with a computing power of 6 TOPS, and the NPU is provided with an AI target detection module (13).
8. The edge computing gateway for target detection and video recording according to claim 7, characterized in that: The CPU (11) is connected to a rechargeable lithium battery, and the rechargeable lithium battery is installed in an iron box, which is sealed with sealing glue.
