An AI edge computing method and system for machine vision

Abstract

This invention relates to an AI edge computing method and system for machine vision, comprising multiple edge computers, an AI control terminal, and multiple vision acquisition terminals. Each vision acquisition terminal includes a camera and a housing. An insulating tube passes through one end of the housing, and an inner tube passes through the insulating tube. The end of the insulating tube inside the housing is closed. A magnetically attachable mounting sleeve is fitted onto the end of the insulating tube inside the housing to mount and fix the camera. The mounting sleeve is rotatable around the axial direction of the insulating tube. An annular piston is fitted onto the inner tube, and a magnet assembly for magnetically attaching the mounting sleeve is disposed within the annular piston. A one-way air inlet valve is disposed on the surface of the portion of the insulating tube inside the housing, and a detection unit for detecting gas composition and pressure is disposed at the outlet end of the insulating tube away from the housing. This significantly improves safety performance. The acquired video data and detection data are first processed by the edge computers and then sent to the AI ​​control terminal for global analysis, which can greatly reduce the data processing burden.

Description

Technical Field

[0001] This invention relates to the fields of edge computing and machine vision technology, and more specifically, to an AI edge computing method for machine vision. Background Technology

[0002] Machine vision is a rapidly developing branch of artificial intelligence. Machine vision products (i.e., image acquisition devices, which are divided into CMOS and CCD types) convert the captured target into image signals and transmit them to a dedicated image processing system to obtain the shape information of the captured target. Based on pixel distribution and information such as brightness and color, the image system converts the target into digital signals. The image system performs various operations on these signals to extract the features of the target and then controls the action of the equipment on site based on the judgment results.

[0003] In scenarios with high concentrations of flammable gases and dust (such as oil depots), the image acquisition process requires additional explosion-proof design and rapid intelligent handling of existing risks. Existing machine vision methods are inadequate for this task, requiring significant human intervention and offering limited protection. Furthermore, they cannot provide a comprehensive safety assessment. Therefore, a safer and more efficient AI edge computing method is needed for such scenarios. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an AI edge computing method for machine vision and an AI edge computing system for machine vision, in order to address the above-mentioned deficiencies of the prior art.

[0005] The technical solution adopted by this invention to solve its technical problem is:

[0006] An AI edge computing system for machine vision is constructed, comprising multiple edge computers, an AI control terminal, and multiple vision acquisition terminals. Each vision acquisition terminal includes a camera and a strip-shaped housing. One end of the housing is sealed and fixedly fitted with an insulating tube, within which a rotatable inner tube is inserted. The end of the insulating tube located inside the housing is closed. A magnetically attachable mounting sleeve is fitted onto the end of the insulating tube inside the housing to mount and fix the camera. The mounting sleeve is rotatable about the axial direction of the insulating tube. An annular piston is fitted onto the inner tube, and a magnet assembly for magnetically attracting the mounting sleeve is disposed within the annular piston. A transparent protective cover is embedded in the housing. One end of the inner tube is connected to a gas extraction unit for extracting inert gas, a pulling unit for pulling the inner tube, and a rotation unit for rotating the inner tube. A switching plug is rotatably connected to the other end of the inner tube; the switching plug is movably inserted through the end of the insulating tube, and a switching channel communicating with the inner hole of the inner tube is provided on the switching plug. The switching plug is pulled and moved by the inner tube and has a first state and a second state. In the first state, the switching channel communicates the inner hole of the insulating tube and the inner hole of the inner tube. In the second state, the switching channel communicates the inner hole of the insulating tube and the internal space of the outer shell. A one-way air inlet valve is provided on the surface of the insulating tube located inside the outer shell. A detection unit for detecting gas composition and pressure is provided at the outlet of the insulating tube away from the outer shell. The edge computer receives video data from the camera and detection data from the detection unit, performs corresponding data processing, generates a safety status index, and uploads it to the AI ​​control terminal.

[0007] The AI ​​edge computing system for machine vision described in this invention includes a gas extraction and degassing unit comprising a mounting plate, on which an inert gas filling module and a gas extraction module, both connected to the inner tube, are disposed.

[0008] The AI ​​edge computing system for machine vision described in this invention includes an inner tube rotatably connected to a mounting plate; a pulling unit comprising a cylinder, the movable end of which is fixedly connected to the mounting plate; a rotating unit comprising a servo motor mounted on the mounting plate, the movable end of which is provided with a spline; and a spline sleeve coaxially fitted on the inner tube to cooperate with the spline.

[0009] The AI ​​edge computing system for machine vision described in this invention includes a switching channel comprising a first channel disposed along the length direction of the switching plug, a second channel communicating with the first channel disposed radially on the switching plug, and a third channel communicating with the first channel.

[0010] In the first state, the inner hole of the insulating tube and the inner hole of the inner tube are connected through the first and second channels, and the third channel is blocked and closed.

[0011] In the second state, the first channel and the third channel connect the inner hole of the insulating tube and the internal space of the outer shell, at which time the second channel is blocked and closed.

[0012] The AI ​​edge computing system for machine vision described in this invention includes a magnet assembly comprising a plurality of strong magnets uniformly distributed circumferentially on an annular piston, and a plurality of iron pieces corresponding one-to-one with the strong magnets on the inner wall of the mounting sleeve.

[0013] In the AI ​​edge computing system for machine vision described in this invention, the one-way air intake valve and the switching plug are respectively located at both ends of the housing.

[0014] The AI ​​edge computing system for machine vision described in this invention includes an annular detection chamber connected to the end of an insulating tube. A sealed bearing assembly connected to an inner tube is located at the center of the detection chamber. The sealed bearing assembly includes a rotating sealed bearing and a linear sealed bearing. The linear sealed bearing passes through the rotating sealed bearing, and its outer ring is fixedly connected to the inner ring of the rotating sealed bearing. The inner tube passes through the linear sealed bearing. The detection unit includes a gas sensor for detecting gas composition and a pressure sensor located within the detection chamber. An exhaust unit is also provided on the detection chamber.

[0015] An AI edge computing method for machine vision, applied to the AI ​​edge computing system for machine vision as described above, wherein the method includes the following steps:

[0016] According to the established inspection procedure, internal gas detection is performed when the set time arrives:

[0017] The pulling unit moves the inner tube, which in turn moves the switching plug to the second state, and the switching channel connects the inner hole of the insulating tube and the internal space of the outer shell;

[0018] The gas extraction and degassing unit operates to fill the inner tube with inert gas. The inert gas enters the internal space of the outer shell through the switching channel and compresses the gas in the internal space of the outer shell into the inner hole of the insulating tube through the one-way inlet valve.

[0019] The detection unit detects the gas composition and pressure inside the insulating tube. If the inert gas content and pressure are both within the corresponding set normal threshold range, it is determined that there is no inert gas leakage; otherwise, it is determined that there is an inert gas leakage.

[0020] The edge computer receives video data from the camera and detection data from the detection unit, processes the data accordingly, generates a security status index, and uploads it to the AI ​​control terminal.

[0021] The AI ​​control terminal assesses overall security performance based on the security status index of multiple edge computers.

[0022] The AI ​​edge computing method for machine vision described in this invention includes the following steps:

[0023] Receive position adjustment commands from external sources and adjust the camera position accordingly.

[0024] The pulling unit moves the inner tube, which in turn moves the switching plug to the first state, and the switching channel connects the inner hole of the insulating tube and the inner hole of the inner tube.

[0025] The gas extraction and degassing unit operates by filling or extracting inert gas into the inner tube, thereby changing the position of the annular piston by relying on gas pressure.

[0026] The annular piston moves the mounting sleeve along the insulating tube due to magnetic attraction, thereby moving the camera.

[0027] The AI ​​edge computing method for machine vision described in this invention includes the following steps:

[0028] Receive rotation angle adjustment commands from external sources and adjust the camera's rotation angle accordingly.

[0029] The rotating unit drives the inner tube to rotate, which in turn drives the annular piston to rotate. The annular piston, based on magnetic attraction, drives the mounting sleeve to rotate around the insulating tube.

[0030] The beneficial effects of this invention are as follows: According to the set inspection procedure, internal gas detection is performed when the set time arrives: the pulling unit moves the inner tube, thereby moving the switching plug to the second state, and the switching channel connects the inner hole of the insulating tube and the internal space of the outer shell; the gas extraction unit operates to fill the inner tube with inert gas, which enters the internal space of the outer shell through the switching channel, and compresses the gas in the internal space of the outer shell into the inner hole of the insulating tube through the one-way inlet valve; the detection unit detects the gas composition and pressure inside the insulating tube; if the inert gas content and pressure are both within the corresponding set normal threshold range, it is determined that there is no inert gas leakage; otherwise, it is determined that there is inert gas leakage; the edge computer receives video data from the camera and detection data from the detection unit, performs corresponding data processing to generate a safety status index, and uploads it to the AI ​​control terminal; the AI ​​control terminal evaluates the overall safety performance based on the safety status indices of multiple edge computers.

[0031] The method described in this application not only solves the safety protection problem of visual acquisition, but also has the function of adjusting the position and angle of the camera, enabling the camera to complete monitoring and patrol in explosion-proof conditions. Moreover, all control components for the actions are located at a remote location away from the outer casing, and the transmission of actions is accomplished by using inert gas as a carrier, which can significantly improve safety performance. The acquired video data and detection data are first processed by the edge computer and then sent to the AI ​​control terminal for global analysis, which can greatly reduce the data processing pressure on the control terminal. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort:

[0033] Figure 1 This is a block diagram illustrating the principle of an AI edge computing system for machine vision according to a preferred embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of the visual acquisition end structure of an AI edge computing system for machine vision according to a preferred embodiment of the present invention.

[0035] Figure 3 This is a cross-sectional view of the vision acquisition end of an AI edge computing system for machine vision according to a preferred embodiment of the present invention.

[0036] Figure 4 This is a cross-sectional view of the end of an insulating tube in a preferred embodiment of an AI edge computing system for machine vision according to the present invention.

[0037] Figure 5 This is a schematic diagram of the external insulating tube of an AI edge computing system for machine vision according to a preferred embodiment of the present invention.

[0038] Figure 6 This is a flowchart of an AI edge computing method for machine vision according to a preferred embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, a clear and complete description will be provided below in conjunction with the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0040] A preferred embodiment of the present invention is an AI edge computing system for machine vision, such as... Figure 1 As shown, see also Figures 2-5 The system includes multiple edge computers 10, AI control terminals 11, and multiple visual acquisition terminals 12. Each visual acquisition terminal 12 includes a camera 1 and a strip-shaped housing 2 (preferably an explosion-proof housing). One end of the housing 2 is sealed and fixedly fitted with an insulating tube 3. A rotatable inner tube 4 passes through the insulating tube 3, and the end of the insulating tube 3 inside the housing 2 is closed. A magnetically attachable mounting sleeve 5 is fitted onto the end of the insulating tube 3 inside the housing 2 to mount and fix the camera 1. The mounting sleeve 5 can rotate axially around the insulating tube 3. An annular piston 6 is fitted onto the inner tube 4, and a magnet assembly 7 for magnetically attaching the mounting sleeve 5 is located inside the annular piston 6. A transparent protective cover 20 for shooting is embedded in the housing 2. One end of the inner tube 4 is connected to a gas extraction unit for extracting inert gas, a pulling unit 41 for pulling the inner tube, and a rotating unit 42 for rotating the inner tube. The other end is rotatably connected to a switching plug 43; the switching plug 43 is movably inserted through the end of the insulating tube 3, and the switching plug 43 is provided with a switching channel 430 communicating with the inner hole of the inner tube 4. The switching plug 43 is pulled and moved by the inner tube 4 and has a first state and a second state. In the first state, the switching channel 430 communicates the inner hole of the insulating tube 3 and the inner hole of the inner tube 4. In the second state, the switching channel 430 communicates the inner hole of the insulating tube 3 and the internal space of the outer shell 2. A one-way air inlet valve 30 is provided on the surface of the insulating tube 3 located inside the outer shell 2. A detection unit 31 for detecting gas composition and pressure is provided at the outlet of the insulating tube 3 away from the outer shell 2. The edge computer 10 receives the video data from the camera 1 and the detection data from the detection unit 31, performs corresponding data processing, generates a safety status index, and uploads it to the AI ​​control terminal 11.

[0041] According to the set inspection procedure, internal gas detection is performed when the set time arrives: the pulling unit 41 moves the inner tube 4, thereby moving the switching plug 43 to the second state, and the switching channel 430 connects the inner hole of the insulating tube and the internal space of the outer shell; the gas extraction unit operates to fill the inner tube 4 with inert gas, and the filled inert gas enters the internal space of the outer shell 2 through the switching channel 430, and squeezes the gas in the internal space of the outer shell 2 into the inner hole of the insulating tube 3 through the one-way air inlet valve 30; the detection unit 31 detects the gas composition and pressure in the insulating tube 3. If the inert gas content and pressure in the gas composition are within the corresponding set normal threshold range, it is determined that there is no inert gas leakage; otherwise, it is determined that there is inert gas leakage; the edge computer 10 receives the video data from the camera 1 and the detection data from the detection unit 31, performs corresponding data processing, generates a safety status index, and uploads it to the AI ​​control terminal 11; the AI ​​control terminal 11 evaluates the overall safety performance based on the safety status indices of multiple edge computers 10.

[0042] The method described in this application not only solves the safety protection problem of visual acquisition, but also has the function of adjusting the position and angle of the camera, enabling the camera to complete monitoring and patrol in explosion-proof conditions. Moreover, all control components for the actions are located at a remote end away from the outer casing (extended to a relatively safe area by means of insulating tube 3 and inner tube 4). The transmission of actions is accomplished by using inert gas as a carrier, which can greatly improve safety performance. The acquired video data and detection data are first processed by the edge computer and then sent to the AI ​​control terminal for global analysis, which can greatly reduce the data processing pressure on the control terminal.

[0043] Preferably, the gas extraction and degassing unit includes a mounting plate 400, on which an inert gas inflation module 401 and a gas extraction module 402, both communicating with the inner tube, are mounted. The inner tube 4 is rotatably connected to the mounting plate 400 (the rotation angle can be limited by two limiting blocks). The pulling unit 41 includes a cylinder, the movable end of which is fixedly connected to the mounting plate 400. The rotating unit 42 includes a servo motor 420, which is mounted on the mounting plate 400. The movable end of the servo motor 420 is provided with a spline 421, and a spline sleeve 422 that mates with the spline is coaxially sleeved on the inner tube 4. This structural layout design allows for a smaller overall size, rich functionality, and a very compact structure.

[0044] Preferably, the switching channel 430 includes a first channel 4300 disposed along the length direction of the switching plug 43, a second channel 4301 disposed radially on the switching plug and communicating with the first channel 4300, and a third channel 4302 communicating with the first channel 4300.

[0045] In the first state, the inner hole of the insulating tube 3 and the inner hole of the inner tube 4 are connected through the first channel 4300 and the second channel 4301, while the third channel 4302 is blocked and closed.

[0046] In the second state, the first channel 4300 and the third channel 4302 connect the inner hole of the insulating tube 3 and the internal space of the outer shell 2, while the second channel 4301 is blocked and closed.

[0047] It has a simple structure, good switching reliability, is very convenient to operate, and has low cost.

[0048] Preferably, the magnet assembly 7 includes a plurality of strong magnets 70 evenly distributed circumferentially on the annular piston 6, and the inner wall of the mounting sleeve 5 is provided with a plurality of iron parts 50 corresponding one-to-one with the strong magnets 70.

[0049] Preferably, the one-way air intake valve 30 and the switching plug 43 are located at both ends of the housing 2 to ensure the detection effect of inert gas leakage detection.

[0050] Preferably, the end of the insulating tube 3 is connected to an annular detection chamber 8, and a sealed bearing assembly 80 connected to the inner tube 4 is provided at the center of the detection chamber 8; the sealed bearing assembly 80 includes a rotating sealed bearing and a linear sealed bearing, the linear sealed bearing is inserted inside the rotating sealed bearing, and the outer ring of the linear sealed bearing is fixedly connected to the inner ring of the rotating sealed bearing, and the inner tube is inserted inside the linear sealed bearing; the detection unit 31 includes a gas sensor 310 for detecting gas composition and a pressure sensor 311 located in the detection chamber 8; the structure is reasonable and compact, and easy to install;

[0051] Preferably, the detection chamber 8 is also equipped with an exhaust unit 81; when detecting inert gas leakage, the exhaust unit should be in a closed state to detect the gas pressure and gas composition inside the insulating tube; when adjusting the position of the camera 1, the exhaust unit 81 should be opened to exhaust gas in an appropriate amount to avoid the formation of negative pressure inside the insulating tube, which would interfere with the smooth adjustment.

[0052] An AI edge computing method for machine vision is applied to the aforementioned AI edge computing system for machine vision, such as... Figure 6 As shown, the method includes the following steps:

[0053] According to the established inspection procedure, internal gas detection is performed when the set time arrives:

[0054] S01: The pulling unit drives the inner tube to move, thereby causing the switching plug to be in the second state, and the switching channel connects the inner hole of the insulating tube and the internal space of the outer shell;

[0055] S02: The gas extraction and degassing unit operates to fill the inner tube with inert gas. The inert gas enters the internal space of the outer shell through the switching channel and compresses the gas in the internal space of the outer shell into the inner hole of the insulating tube through the one-way air inlet valve.

[0056] S03: The detection unit detects the gas composition and pressure inside the insulating tube. If the inert gas content and pressure are both within the corresponding set normal threshold range, it is determined that there is no inert gas leakage; otherwise, it is determined that there is inert gas leakage.

[0057] S04: The edge computer receives video data from the camera and detection data from the detection unit, processes the data accordingly, generates a security status index, and uploads it to the AI ​​control terminal.

[0058] The safety status index can be generated by a custom mapping table that maps video data to dangerous scenarios and detection data to leakage status. The safety status index is generated by combining dangerous scenarios and leakage status.

[0059] S05: The AI ​​control terminal assesses the overall security performance based on the security status index of multiple edge computers.

[0060] The assessment can be weighted by combining the location weight of the edge computer with the security status index to obtain the security performance level of each edge computer's location. Edge computers that are below the set security threshold will be subject to key monitoring or timely maintenance.

[0061] The method also includes the following steps:

[0062] Receive position adjustment commands from external sources and adjust the camera position accordingly.

[0063] The pulling unit moves the inner tube, which in turn moves the switching plug to the first state, and the switching channel connects the inner hole of the insulating tube and the inner hole of the inner tube.

[0064] The gas extraction and degassing unit operates by filling or extracting inert gas into the inner tube, thereby changing the position of the annular piston by relying on gas pressure.

[0065] The annular piston moves the mounting sleeve along the insulating tube due to magnetic attraction, thereby moving the camera.

[0066] The method also includes the following steps:

[0067] Receive rotation angle adjustment commands from external sources and adjust the camera's rotation angle accordingly.

[0068] The rotating unit drives the inner tube to rotate, which in turn drives the annular piston to rotate. The annular piston, based on magnetic attraction, drives the mounting sleeve to rotate around the insulating tube.

[0069] The method described in this application not only solves the safety protection problem of visual acquisition, but also has the function of adjusting the position and angle of the camera, enabling the camera to complete monitoring and patrol in explosion-proof conditions. Moreover, all control components for the actions are located at a remote location away from the outer casing, and the transmission of actions is accomplished by using inert gas as a carrier, which can significantly improve safety performance. The acquired video data and detection data are first processed by the edge computer and then sent to the AI ​​control terminal for global analysis, which can greatly reduce the data processing pressure on the control terminal.

[0070] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An AI edge computing system for machine vision, characterized by, The application relates to an edge computer, an AI control terminal and a plurality of visual collection terminals, wherein the visual collection terminal comprises a camera and a strip-shaped shell, one end of the shell is fixedly provided with an insulating pipe, a rotatable inner pipe is arranged in the insulating pipe, and the end of the insulating pipe in the shell is closed; a magnetically-attractive mounting sleeve for mounting the camera is arranged on one end of the insulating pipe in the shell, the mounting sleeve can rotate axially around the insulating pipe, a ring-shaped piston is arranged on the inner pipe, and a magnet assembly for magnetically attracting the mounting sleeve is arranged in the ring-shaped piston; a transparent protective cover is arranged on the shell; one end of the inner pipe is connected with a gas extraction unit for extracting inert gas, a pulling unit for pulling the inner pipe and a rotating unit for driving the inner pipe to rotate, the other end of the inner pipe is rotatably connected with a switching plug; the switching plug is movably arranged in the end of the insulating pipe, a switching channel is arranged on the switching plug and communicates with the inner hole of the inner pipe, the switching plug is pulled by the inner pipe and has a first state and a second state, when in the first state, the switching channel communicates the inner hole of the insulating pipe and the inner hole of the inner pipe, when in the second state, the switching channel communicates the inner hole of the insulating pipe and the inner space of the shell, a one-way air inlet valve is arranged on the surface of the part of the insulating pipe in the shell, a detection unit for detecting the gas composition and pressure is arranged at the outlet of the end of the insulating pipe away from the shell; the edge computer receives the video data of the camera and the detection data of the detection unit, generates a safety state index after corresponding data processing and uploads the safety state index to the AI control terminal. 2.The AI edge computing system for machine vision of claim 1, wherein, The gas extraction unit comprises a mounting plate, and the mounting plate is provided with an inert gas inflation module and an air extraction module which communicate with the inner pipe. 3.The AI edge computing system for machine vision of claim 2, wherein, The inner pipe is rotatably connected with the mounting plate, the pulling unit comprises a gas cylinder, the movable end of the gas cylinder is fixedly connected with the mounting plate, the rotating unit comprises a servo motor, the servo motor is mounted on the mounting plate, the movable end of the servo motor is provided with a spline, and the inner pipe is coaxially provided with a spline sleeve matched with the spline. 4.The AI edge computing system for machine vision of claim 1, wherein, The switching channel comprises a first channel arranged along the length direction of the switching plug, a second channel arranged along the radial direction of the switching plug and communicated with the first channel, and a third channel communicated with the first channel; When in the first state, the first channel and the second channel communicate the inner hole of the insulating pipe and the inner hole of the inner pipe, and the third channel is shielded and closed; When in the second state, the first channel and the third channel communicate the inner hole of the insulating pipe and the inner space of the shell, and the second channel is shielded and closed. 5.The AI edge computing system for machine vision of claim 1, wherein, The magnet assembly comprises a plurality of strong magnets uniformly distributed along the circumferential direction of the ring-shaped piston, and the inner wall of the mounting sleeve is provided with a plurality of iron pieces corresponding to the strong magnets. 6.The AI edge computing system for machine vision of claim 1, wherein, The one-way air inlet valve and the switching plug are respectively arranged at the two end positions of the shell. 7.The AI edge computing system for machine vision of claim 1, wherein, The end of the insulating tube is provided with an annular detection chamber, and a sealing bearing assembly connected to the inner tube is arranged at the center of the detection chamber; the sealing bearing assembly comprises a rotary sealing bearing and a linear sealing bearing, the linear sealing bearing is arranged in the rotary sealing bearing, and the outer ring of the linear sealing bearing is fixedly connected with the inner ring of the rotary sealing bearing, and the inner tube is arranged in the linear sealing bearing; the detection unit comprises a gas sensor for detecting the gas composition and a pressure sensor in the detection chamber; the detection chamber is also provided with an exhaust unit.

8. An AI edge computing method for machine vision, applied to the AI edge computing system for machine vision according to any one of claims 1-7, characterized in that, The method comprises the steps of: According to the set inspection program, internal gas detection is performed when the set time arrives: The pulling unit moves the inner tube, thereby driving the switching plug to the second state, and the switching channel communicates the inner hole of the insulating tube and the internal space of the shell; The inert gas charging unit fills inert gas into the inner tube, and the inert gas fills into the internal space of the shell through the switching channel, and extrudes the gas in the internal space of the shell into the inner hole of the insulating tube through the one-way air inlet valve; The detection unit detects the gas composition and pressure in the insulating tube, and if the inert gas content and pressure in the gas composition are within the corresponding normal threshold range, it is determined that there is no inert gas leakage, otherwise it is determined that there is inert gas leakage; The edge computer receives the video data of the camera and the detection data of the detection unit, generates a safety state index after corresponding data processing, and uploads it to the AI control terminal; The AI control terminal evaluates the overall safety performance according to the safety state index of multiple edge computers. 9.The AI edge computing method for machine vision of claim 8, wherein, The method comprises the steps of: Receive the position adjustment instruction sent by the outside, and adjust the position of the camera according to the adjustment instruction: The pulling unit moves the inner tube, thereby driving the switching plug to the first state, and the switching channel communicates the inner hole of the insulating tube and the inner hole of the inner tube; The inert gas charging unit fills or extracts inert gas into the inner tube, thereby changing the position of the annular piston by relying on air pressure; The annular piston drives the mounting sleeve to move along the insulating tube according to magnetic attraction, thereby driving the camera to move. 10.The AI edge computing method for machine vision of claim 8, wherein, The method comprises the steps of: Receive the rotation angle adjustment instruction sent by the outside, and adjust the rotation angle of the camera according to the adjustment instruction: The rotating unit drives the inner tube to rotate, thereby driving the annular piston to rotate, and the annular piston drives the mounting sleeve to rotate around the insulating tube according to magnetic attraction.

Images