Icing thickness monitoring method based on AI image

Abstract

The invention discloses an icing thickness monitoring method based on an AI image, which can improve the monitoring precision and save the labor cost. The method comprises an AI image analysis mechanism, a weighing method analysis mechanism, a temperature analysis mechanism, a comprehensive analysis mechanism and a fault alarm mechanism. The comprehensive analysis mechanism comprises an AI image module used for receiving analysis result information of the AI image analysis mechanism, the weighing method analysis mechanism and the temperature analysis mechanism, a weighing module, an environment module and a fault verification module, the effectiveness of weighing method data is preferentially judged, and then whether a calculation result of the weighing method analysis mechanism is used or not is determined. Secondly, according to the result of the weighing module, the application function of the AI image module is selected, the fault verification module is used for verifying the validity of the calculation result again, the validity of each module is judged from the perspective of fault information, and when one module is invalid, the analysis result of the module is overtopped. The method is applied to the technical field of power transmission automation.

Description

Technical Field

[0001] The present invention is applied to the technical field of power transmission automation, and in particular relates to an ice thickness monitoring method based on AI images. Background Art

[0002] China has invested heavily in the research of icing devices, especially in cold regions such as the Northeast and Northwest. Research focuses include icing process simulation, ice load testing, anti-icing design, deicing technology, etc. A variety of simulated icing devices have been developed in China and applied to the design of transmission lines, material selection, structural optimization and deicing technology research and development. Foreign countries, especially Canada, the United States and some European countries, are at the international advanced level in the research of transmission line simulated icing devices, with relatively mature technology and wide application. Their research direction focuses more on interdisciplinary cooperation, such as the combination of meteorology, fluid mechanics, materials science and power engineering, and improves the accuracy and practicality of icing simulation through comprehensive methods. There are many innovations in deicing technology, material research and development, monitoring technology, etc., such as the development of more efficient electric thermal deicing technology and new anti-icing coatings.

[0003] With the development of artificial intelligence and automation technology, the demand for de-icing devices will become more intelligent and automated, and they will be able to automatically adjust the experimental conditions and accurately simulate the de-icing process under different meteorological conditions. Integrated de-icing monitoring and early warning systems will become the focus of research. If an AI image-based de-icing thickness monitoring method can be developed, it will be possible to monitor the de-icing condition of transmission lines in real time, issue early warnings, reduce the impact of ice damage, and meet the needs of intelligence. Summary of the invention

[0004] The present invention solves the defects or shortcomings of the prior art and provides an ice thickness monitoring method based on AI images, which improves monitoring accuracy and saves labor costs.

[0005] The technical solution adopted by the present invention is: the monitoring system used in the method includes a camera assembly, a micro-meteorological sensor, a tension sensor, an analog wire and an ice detection structure, the camera assembly is installed on the ice detection structure, the analog wire is located below the ice detection structure, and both ends of the analog wire are fixed to the two ends of the ice detection structure through the tension sensor, and the micro-meteorological sensor is installed at the upper end of the camera assembly; the method includes an AI image analysis mechanism, a weighing analysis mechanism, a temperature analysis mechanism, a comprehensive analysis mechanism and a fault alarm mechanism;

[0006] The AI ​​image analysis mechanism includes the following steps: a1. Acquire the simulated wire image through the camera component, and judge the validity of the simulated wire image data through the fault alarm mechanism; a2. If the simulated wire image is invalid, output the abnormal mark code; a3. If the simulated wire image is valid, obtain the mask image of the simulated wire through the yolov8-seg segmentation algorithm, and calculate the height of the pixels in the mask image to obtain the ice thickness of the simulated wire;

[0007] The weighing analysis mechanism includes the following steps: b1. The tension of the simulated conductor is obtained through a tension sensor, the wind speed and wind direction are obtained through a micro-meteorological sensor, and the validity of the tension, wind speed and wind direction data is determined through a fault alarm mechanism; b2. If the tension value is invalid, the weighing analysis result is determined to be invalid, the current calculation is stopped, and the fault description and invalid status code are returned; b3. When the tension value is valid, the validity of the wind speed and wind direction data will be determined first. When the wind speed and wind direction are invalid, the weighing analysis mechanism will call a preset value to replace the calculation result of the wind speed and wind direction for subsequent calculations. When the wind speed and wind direction data are valid, the wind pressure on the simulated conductor is calculated. The force direction of the calculated wind pressure is in the direction of gravity. Secondly, the horizontal pressure should be perpendicular to the center of gravity and will not be affected. Therefore, the ice weight of the simulated conductor in the ice-covered state can be obtained, and then the ice thickness of the simulated conductor can be analyzed by the ice weight;

[0008] The temperature analysis mechanism includes the following steps: c1. Acquire temperature data of a set time interval through a micro-meteorological sensor, and determine the validity of the temperature data through a fault alarm mechanism; c2. If the temperature data is invalid, determine that the temperature analysis result is invalid, stop the current calculation, and return a fault description and an invalid status code; c3. When the temperature data is valid, calculate the temperature value by a dynamic average value, open the initial conditions, and set the initial icing temperature condition threshold. If the calculated temperature value is less than the icing threshold, the temperature analysis mechanism determines that the icing condition is met, otherwise it will be determined that the icing condition is not met;

[0009] The comprehensive analysis mechanism includes an AI image module, a weighing module, an environmental module, and a fault verification module. The AI ​​image module, the weighing module, and the environmental module are respectively used to receive the analysis result information of the AI ​​image analysis mechanism, the weighing analysis mechanism, and the temperature analysis mechanism. The comprehensive analysis mechanism includes the following steps: d1. Start comprehensive analysis and calculation of the information received by the three modules; d2. Detect the effectiveness of the weighing analysis mechanism. When the weighing analysis mechanism is effective, the ice thickness analyzed by the weighing analysis mechanism is given to the ice thickness of the comprehensive analysis, and then the AI ​​image module is called to determine whether the simulated wire exists. Ice covering; d3. If there is no ice covering, the ice covering thickness of the comprehensive analysis is assigned a value of 0. Secondly, if the AI ​​image analysis shows that there is ice covering, it is prioritized to determine whether the weighing module is valid. If it is invalid, the thickness value of the AI ​​image analysis is assigned to the ice covering thickness of the comprehensive analysis; d4. Call the environment module for judgment. If the temperature is invalid, skip the judgment of the environment module. If the temperature analysis result is valid, verify the correctness of the ice covering state through the temperature analysis result. Finally, call the fault verification module to verify the validity of the analysis results from the weighing mechanism, AI image analysis mechanism, and temperature analysis mechanism in turn, and output the current ice covering thickness and state.

[0010] Furthermore, the camera assembly includes a pan-tilt camera, a night-light camera and a daylight camera. The monitoring system also includes a battery system and a solar panel. The battery system provides power for the monitoring system and is externally connected to the solar panel. The tension sensor, micro-meteorological sensor, tension sensor and the battery system are all electrically connected to the camera assembly. The pan-tilt camera is equipped with a controller and is connected to the back-end platform for data communication via a wireless transceiver module.

[0011] Furthermore, the fault alarm mechanism in step a1 is used to determine whether a picture can be taken and whether the simulated wire in the picture can be identified; the ice thickness of the simulated wire in step a3 is calculated by the ice calculation formula: k=(Ll)*d / 2l, where k is the ice thickness of the simulated wire calculated by the AI ​​image analysis mechanism, uppercase L is the average pixel height after ice coating, lowercase l is the average pixel height when the simulated wire is not iced, and d is the diameter of the simulated wire.

[0012] Furthermore, when the wind speed and direction data in step b3 are valid, the formula:

[0013] F_P=W_O*S*COS(θ), where: F_P: is the wind pressure in the direction of gravity, W_O: is the wind pressure, S: is the cross-sectional area of ​​the simulated wire under the wind pressure, θ: is the angle between the wind pressure and the gravity direction of the simulated wire; the wind pressure on the simulated wire is calculated, and the force direction of the calculated wind pressure is in the direction of gravity. Secondly, the horizontal pressure should be perpendicular to the center of gravity and will not be affected. Therefore, the ice weight of the simulated wire in the ice-covered state is M_P=F-F_P-F_N, where: M_P is the ice weight of the simulated wire, F is the value of the tension sensor, F_P: is the wind pressure, and F_N is the initial tension value of the tension sensor in the un-iced state; after the ice weight is calculated, it will be calculated through the formula:

[0014]

[0015] Wherein: d is the diameter of the simulated conductor, L_P is the horizontal span of the simulated conductor, M_P is the weight of ice covering the simulated conductor, B_P is the ice thickness analyzed by the weighing analysis mechanism, f is the density of ice, and the ice thickness value and the ice status analyzed by the weighing method are calculated. When the ice thickness = 0, the simulated conductor is not covered with ice. When the ice thickness is greater than 0, the simulated conductor is in an ice-covered state. After the weighing analysis mechanism is completed, the ice thickness analyzed by the weighing method, the ice status analyzed by the weighing method, and the fault information status code will be output.

[0016] Furthermore, the logical steps of the fault verification module in step d4 are: determine whether the data of the two tension sensors are missing or out of bounds. If both are normal, the ice thickness is calculated through a comprehensive analysis mechanism. If there is a missing or out-of-bounds data, further determine whether the image can be captured and whether the captured image can identify the simulated wire. If the image can be captured and the simulated wire can be identified, output the ice thickness calculated by the AI ​​image segmentation algorithm; if the image cannot be captured or the simulated wire in the image cannot be identified, output the value of the last ice thickness.

[0017] Furthermore, a data mutation prevention logic is also set after the fault verification module in step d4, and the steps are: determine whether the difference between the simulated conductor ice coating thickness this time and the simulated conductor ice coating thickness obtained last time is greater than 2mm; if it is greater than 2mm, change the simulated conductor ice coating thickness this time to the value of the simulated conductor ice coating thickness obtained last time plus 2mm.

[0018] The beneficial effects of the present invention are: 1. The ice thickness monitoring method based on AI images of the present invention utilizes a camera component to collect multi-scene image information, and combines the micro-meteorological sensor data and the tension sensor data for comprehensive analysis to obtain the ice status and ice results of the simulated conductor, thereby solving the problems of false alarms, missed alarms, insufficient accuracy and inability to self-check and self-verify due to environmental influences of the existing single sensor. The monitoring data can not only be used for real-time early warning, but also provide theoretical data guidance for subsequent engineering planning and programs; 2. Remote monitoring and remote operation can be realized, reducing dependence on personnel, and can reduce the damage and maintenance frequency of transmission lines caused by ice and snow weather, reduce maintenance costs, and can also provide important technical support for the design, maintenance and anti-icing of transmission lines, which helps to improve the stability and reliability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic diagram of the monitoring system;

[0020] Figure 2 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0021] like Figure 1 and Figure 2 As shown, in this embodiment, a method for monitoring ice thickness based on AI images is provided, wherein the monitoring system used in the method comprises a camera assembly 1, a micro-meteorological sensor 2, a tension sensor 3, a simulated wire 4 and an ice detection structure 5, wherein the ice detection structure 5 is a test frame, the camera assembly 1 is installed on the middle part of the ice detection structure 5, the simulated wire 4 is located below the ice detection structure 5, and both ends of the simulated wire 4 are fixed to the two ends of the ice detection structure 5 through the tension sensor 3, gravity data on the simulated wire 4 is collected through the tension sensor 3, and the micro-meteorological sensor 3 is used to collect gravity data on the simulated wire 4. The sensor 2 is installed at the upper end of the camera assembly 1, and is used to collect real-time data on temperature, relative humidity, wind speed, wind direction, and atmospheric pressure of the environment. The camera assembly 1 includes a pan-tilt camera, a night-light camera, and a daylight camera. The monitoring system also includes a battery system 10 and a solar panel 11. The battery system 10 provides power for the monitoring system and is externally connected to the solar panel 11. The two tension sensors 3, the micro-meteorological sensor 2, and the battery system 10 are all electrically connected to the camera assembly 1. The pan-tilt camera is equipped with a controller, and is connected to the back-end platform 12 for data communication through a wireless transceiver module;

[0022] After the whole monitoring system is installed, the camera component 1 is powered on and the self-test starts successfully. The camera shooting points, image analysis and processing parameters, and the original tension value of the tension sensor 3 are initialized. After the initialization is completed, it automatically connects to the factory preset back-end platform 12. The camera component 1 periodically turns on the power of the micro-meteorological sensor 2. The micro-meteorological sensor 2 autonomously collects the temperature, relative humidity, wind speed, wind direction, and atmospheric pressure data of the surrounding environment of the sensor. When the data collection of the micro-meteorological sensor 2 is ready, the camera component 1 will collect the micro-meteorological data through the RS485 communication method. After the micro-meteorological data collection is completed, the power of the micro-meteorological sensor 2 will be turned off; and the power of the tension sensor 3 will be turned on. The tension sensor 3 autonomously collects the weight data of the simulated wire and transmits it to the camera component 1 through RS485. After the camera component 1 collects the data, the power of the tension sensor 3 is turned off. After the sensor data is collected, the camera component 1 starts to collect image data according to the preset setting points. The image data includes pictures of the simulated wire and pictures of the actual transmission wire.

[0023] The method includes an AI image analysis mechanism 6, a weighing analysis mechanism 7, a temperature analysis mechanism 8, a comprehensive analysis mechanism 9 and a fault alarm mechanism; when data collection is completed, the camera component 1 transmits all data to the algorithm library through the interface, and the algorithm first calls the picture data to call the AI ​​image analysis mechanism 6, and the AI ​​image analysis mechanism 6 includes the following steps: a1. The image of the simulated wire 4 is obtained through the camera component 1, and the validity of the image data of the simulated wire 4 is judged through the fault alarm mechanism; a2. If the image of the simulated wire 4 is invalid, an abnormal mark code is output; a3. If the image of the simulated wire 4 is valid, a mask image of the simulated wire 4 is obtained through the yolov8-seg segmentation algorithm, and the height of the pixels in the mask image is calculated to obtain the ice thickness of the simulated wire 4; through The mask image of the simulated wire 4 is obtained by the yolov8-seg segmentation algorithm, and the image is sent to the neural network inference model for inference; the general process of inference is: the convolutional neural network can extract features from the image, classify and regress the extracted image features, so as to identify the position of the simulated wire 4 in the image and the distribution of pixels when the simulated wire 4 is covered with ice. According to the pixel distribution, the pixel height of the wire when it is covered with ice can be calculated, that is, the ice thickness of the simulated wire 4 is calculated through the ice calculation formula: k=(Ll)*d / 2l, wherein k is the ice thickness of the simulated wire 4 calculated by the AI ​​image analysis mechanism 6, the uppercase L is the average pixel height after ice covering, the lowercase l is the average pixel height when the simulated wire 4 is not covered with ice, and d is the diameter of the simulated wire 4.

[0024] The weighing method analysis mechanism 7 includes the following steps: b1. The tension of the simulated wire 4 is obtained through the tension sensor 3, the wind speed and wind direction are obtained through the micro-meteorological sensor 2, and the validity of the tension, wind speed and wind direction data is determined through the fault alarm mechanism; b2. If the tension value is invalid, the weighing method analysis result is determined to be invalid, the current calculation is stopped, and the fault description and invalid status code are returned; b3. When the tension value is valid, the validity of the wind speed and wind direction data will be determined first. When the wind speed and wind direction are invalid, the weighing method analysis mechanism 7 will call a preset value to replace the calculation result of the wind speed and wind direction for subsequent calculations. This preset value will be installed The wind speed and wind direction trends of the long-term stable operation are calculated. When the wind speed and wind direction data are valid, the wind pressure on the simulated conductor 4 is calculated. The force direction of the calculated wind pressure is in the direction of gravity. Secondly, the pressure in the horizontal direction will not be affected according to the verticality of the center of gravity. That is, the mechanical model is used to decompose the comprehensive force on the simulated conductor 4 into the vertical load in the direction of the center of gravity and the horizontal load in the horizontal direction. Then, the parallelogram law of force is applied to remove the wind speed and wind direction data to calculate the wind load of the simulated conductor 4. Therefore, the ice weight of the simulated conductor 4 in the ice-covered state can be obtained, and then the ice thickness of the simulated conductor 4 can be analyzed by the ice weight.

[0025] The temperature analysis mechanism 8 is used for auxiliary verification, and is used to determine whether the current state meets the icing condition. If and only if the condition is met, the output of the calculated ice thickness value is allowed, otherwise the previous calculation result is rejected. The temperature analysis mechanism 8 includes the following steps: c1. The temperature data of the set time interval is obtained through the micro-meteorological sensor 2, and the validity of the temperature data is determined through the fault alarm mechanism; c2. If the temperature data is invalid, the temperature analysis result is determined to be invalid, the current calculation is stopped, and the fault description and invalid status code are returned; c3. When the temperature data is valid, the temperature value is calculated by the dynamic average value, and the initial icing temperature condition threshold set under the initial condition is opened. If the calculated temperature value is less than the icing threshold, the temperature analysis mechanism 8 determines that the icing condition is met, otherwise it will be determined that the icing condition is not met;

[0026] The comprehensive analysis mechanism 9 includes an AI image module 91, a weighing module 92, an environmental module 93, and a fault verification module 94. The AI ​​image module 91, the weighing module 92, and the environmental module 93 are respectively used to receive the analysis result information of the AI ​​image analysis mechanism 6, the weighing method analysis mechanism 7, and the temperature analysis mechanism 8, and first determine the validity of the weighing method data, and then decide whether to use the calculation result of the weighing method analysis mechanism 7, and then select the application function of the AI ​​image module 91 according to the result of the weighing module 92; its comprehensive analysis mechanism 9 includes the following steps: d1. Start comprehensive analysis and calculation of the information received by the three modules; d2. Detect the validity of the weighing method analysis mechanism 7. When the weighing method analysis mechanism 7 is valid, the weighing method analysis mechanism 7 will be analyzed. The ice thickness is given to the ice thickness of the comprehensive analysis, and then the AI ​​image module 91 is called to determine whether the simulated wire 4 is iced; d3. If there is no ice, the ice thickness of the comprehensive analysis is assigned a value of 0. Secondly, if the AI ​​image analysis shows ice, it is determined again whether the weighing module 92 is valid. If it is invalid, the thickness value of the AI ​​image analysis is assigned to the ice thickness of the comprehensive analysis; d4. The environment module 93 is called for judgment. If the temperature is invalid, the judgment of the environment module 93 is skipped. If the temperature analysis result is valid, the correctness of the ice state is verified by the temperature analysis result. Finally, the fault verification module 94 is called to verify the validity of the analysis results from the weighing method mechanism, the AI ​​image analysis mechanism 6, and the temperature analysis mechanism 8 in turn, and the current ice thickness and state are output.

[0027] In this embodiment, the fault alarm mechanism in step a1 is used to determine whether a picture can be taken and whether the simulated wire 4 in the picture can be identified.

[0028] In this embodiment, when the wind speed and direction data in step b3 are valid, the formula is: F_P = W_O*S*COS(θ), wherein: F_P: is the wind pressure in the direction of gravity, W_O: is the wind pressure, S: is the cross-sectional area of ​​the simulated conductor 4 under the wind pressure, θ: is the angle between the wind pressure and the gravity direction of the simulated conductor 4; the wind pressure on the simulated conductor 4 is calculated, and the force direction of the calculated wind pressure is in the direction of gravity. Secondly, the horizontal pressure should be perpendicular to the center of gravity and will not be affected. Therefore, the ice weight of the conductor in the ice-covered state is M_P = F-F_P-F_N, wherein: M_P is the ice weight of the simulated conductor 4, F is the value of the tension sensor 3, F_P: is the wind pressure, and F_N is the initial tension value of the tension sensor 3 in the un-iced state; after the ice weight is calculated, the formula is used:

[0029]

[0030] Wherein: d: is the diameter of the simulated conductor 4, L_P: is the horizontal span of the simulated conductor 4, M_P: is the weight of ice covering the simulated conductor 4, B_P: is the ice thickness analyzed by the weighing method analysis mechanism 7, f: is the density of ice, the ice thickness value and the ice status analyzed by the weighing method are calculated, when the ice thickness = 0, and the device is not covered with ice, when the ice thickness > 0, the device is in an ice-covered state, after the weighing method analysis mechanism 7 is completed, the ice thickness analyzed by the weighing method, the ice status analyzed by the weighing method, and the fault information status code will be output.

[0031] In this embodiment, the fault verification module 94 in step d4 is used to verify the validity of the calculation results again, and to judge the validity of each module from the perspective of fault information. When one of the modules is invalid, the analysis result of this module will be rejected; the logical steps of the fault verification module 94 are: judging whether the data of the two tension sensors 3 are missing or out of bounds, if both are normal, the ice thickness is calculated by the comprehensive analysis mechanism 9; if the data of the tension sensor 3 is missing or out of bounds, it is further judged whether the image can be captured and whether the captured image can identify the simulated wire 4; if the image can be captured and the simulated wire 4 can be identified, the ice thickness calculated by the AI ​​image segmentation algorithm is output; if the image cannot be captured or the simulated wire 4 in the image cannot be identified, the value of the last ice thickness is output.

[0032] In this embodiment, a data mutation prevention logic is also set after the fault verification module 94 in step d4, and the steps are: determine whether the difference between the simulated conductor 4 ice coating thickness this time and the simulated conductor 4 ice coating thickness obtained last time is greater than 2 mm. If it is greater than 2 mm, change the simulated conductor 4 ice coating thickness this time to the value of the simulated conductor 4 ice coating thickness obtained last time plus 2 mm.

[0033] Although the embodiments of the present invention are described with practical solutions, they do not constitute limitations on the meaning of the present invention. For those skilled in the art, it is obvious to modify the implementation scheme and combine it with other solutions based on this description.

Claims

1. A method for monitoring ice thickness based on AI images, wherein the monitoring system used in the method comprises a camera assembly (1), a micro-meteorological sensor (2), a tension sensor (3), an analog wire (4) and an ice detection structure (5), wherein the camera assembly (1) is mounted on the ice detection structure (5), the analog wire (4) is located below the ice detection structure (5), and both ends of the analog wire (4) are fixed to both ends of the ice detection structure (5) through the tension sensor (3), and the micro-meteorological sensor (2) is mounted on the upper end of the camera assembly (1); characterized in that: The method comprises an AI image analysis mechanism (6), a weighing method analysis mechanism (7), a temperature analysis mechanism (8), a comprehensive analysis mechanism (9) and a fault alarm mechanism; The AI ​​image analysis mechanism (6) comprises the following steps: a1. acquiring an image of the simulated conductor (4) through the camera component (1), and judging the validity of the image data of the simulated conductor (4) through the fault alarm mechanism; a2. outputting an abnormal marking code if the image of the simulated conductor (4) is invalid; a3. obtaining a mask image of the simulated conductor (4) through the yolov8-seg segmentation algorithm if the image of the simulated conductor (4) is valid, and calculating the height of the pixels in the mask image to obtain the ice thickness of the simulated conductor (4); The weighing method analysis mechanism (7) comprises the following steps: b1. obtaining the tension of the simulated conductor (4) through the tension sensor (3), obtaining the wind speed and wind direction through the micro-meteorological sensor (2), and judging the validity of the tension, wind speed and wind direction data through the fault alarm mechanism; b2. if the tension value is invalid, judging the result of the weighing method analysis as invalid, stopping the current calculation, and returning the fault description and invalid status code; b3. when the tension value is valid, first judging the validity of the wind speed and wind direction data; when the wind speed and wind direction are invalid, the weighing method analysis mechanism (7) will call a preset value to replace the calculation result of the wind speed and wind direction for subsequent calculation; when the wind speed and wind direction data are valid, calculating the wind pressure on the simulated conductor (4); the force direction of the calculated wind pressure is in the direction of gravity; secondly, the horizontal pressure should be perpendicular to the center of gravity and will not be affected, so the ice weight of the simulated conductor (4) in the ice-covered state can be obtained, and then the ice thickness of the simulated conductor (4) can be analyzed through the ice weight; The temperature analysis mechanism (8) comprises the following steps: c1. obtaining temperature data of a set time interval through a micro-meteorological sensor (2), and judging the validity of the temperature data through a fault alarm mechanism; c2. if the temperature data is invalid, judging that the temperature analysis result is invalid, stopping the current calculation, and returning a fault description and an invalid status code; c3. when the temperature data is valid, calculating the temperature value by a dynamic average value, and opening the initial icing temperature condition threshold set under the initial condition; if the calculated temperature value is less than the icing threshold, the temperature analysis mechanism (8) judges that the icing condition is met, otherwise it judges that the icing condition is not met; The comprehensive analysis mechanism (9) includes an AI image module (91), a weighing module (92), an environment module (93), and a fault verification module (94). The AI ​​image module (91), the weighing module (92), and the environment module (93) are respectively used to receive the analysis result information of the AI ​​image analysis mechanism (6), the weighing method analysis mechanism (7), and the temperature analysis mechanism (8). The comprehensive analysis mechanism (9) includes the following steps: d1. Starting comprehensive analysis and calculation of the information received by the three modules; d2. Detecting the validity of the weighing method analysis mechanism (7). When the weighing method analysis mechanism (7) is valid, the ice thickness analyzed by the weighing method analysis mechanism (7) is given to the ice thickness of the comprehensive analysis, and then calling the AI ​​image module (91), determine whether the simulated wire (4) is covered with ice; d3. If there is no ice, the ice thickness of the comprehensive analysis is assigned a value of 0. Secondly, if the AI ​​image analysis shows that there is ice, it is determined again whether the weighing module (92) is valid. If it is invalid, the thickness value of the AI ​​image analysis is assigned to the ice thickness of the comprehensive analysis; d4. Call the environment module (93) to make a judgment. If the temperature is invalid, skip the judgment of the environment module (93). If the temperature analysis result is valid, verify the correctness of the ice state through the temperature analysis result. Finally, call the fault verification module (94), verify the validity of the analysis results from the weighing method mechanism, the AI ​​image analysis mechanism (6), and the temperature analysis mechanism (8) in turn, and output the current ice thickness and state.

2. The ice thickness monitoring method based on AI image according to claim 1 is characterized in that: The camera assembly (1) includes a pan-tilt camera, a night-light camera and a daylight camera. The monitoring system also includes a battery system (10) and a solar panel (11). The battery system (10) provides power for the monitoring system and is externally connected to the solar panel (11). The tension sensor (3), the micro-meteorological sensor (2), the tension sensor (3) and the battery system (10) are all electrically connected to the camera assembly (1). The pan-tilt camera is provided with a controller and is connected to a back-end platform (12) for data communication via a wireless transceiver module.

3. The ice thickness monitoring method based on AI image according to claim 2 is characterized in that: The fault alarm mechanism in step a1 is used to determine whether a picture can be taken and whether the simulated wire (4) in the picture can be identified; the ice thickness of the simulated wire (4) in step a3 is calculated by using the ice calculation formula: k=(Ll)*d / 2l, where k is the ice thickness of the simulated wire (4) calculated by the AI ​​image analysis mechanism (6), the uppercase L is the average pixel height after ice coating, the lowercase l is the average pixel height of the simulated wire (4) when not ice coated, and d is the diameter of the simulated wire (4).

4. The ice thickness monitoring method based on AI image according to claim 3 is characterized in that: When the wind speed and direction data in step b3 are valid, the formula: F_P = W_O*S*COS(θ) is used, wherein: F_P: is the wind pressure in the direction of gravity, W_O: is the wind pressure, S: is the cross-sectional area of ​​the simulated conductor (4) under the wind pressure, and θ: is the angle between the wind pressure and the gravity direction of the simulated conductor (4); the wind pressure on the simulated conductor (4) is calculated, and the force direction of the calculated wind pressure is in the direction of gravity. Secondly, the horizontal pressure should be perpendicular to the center of gravity and will not be affected. Therefore, the ice weight of the conductor in the ice-covered state is M_P = F-F_P-F_N, wherein: M_P is the ice weight of the simulated conductor (4), F is the value of the tension sensor (3), F_P: is the wind pressure, and F_N is the initial tension value of the tension sensor (3) in the un-iced state; after the ice weight is calculated, the formula: Wherein: d is the diameter of the simulated conductor (4), L_P is the horizontal span of the simulated conductor (4), M_P is the weight of ice covering the simulated conductor (4), B_P is the ice thickness analyzed by the weighing method analysis mechanism (7), and f is the density of ice. The ice thickness value and the ice status analyzed by the weighing method are calculated. When the ice thickness = 0, the simulated conductor (4) is not covered with ice. When the ice thickness is greater than 0, the simulated conductor (4) is in an ice-covered state. After the weighing method analysis mechanism (7) is completed, the ice thickness analyzed by the weighing method, the ice status analyzed by the weighing method, and the fault information status code are output.

5. The ice thickness monitoring method based on AI image according to claim 4 is characterized in that: The logical steps of the fault verification module (94) in step d4 are as follows: determine whether the data of the two tension sensors (3) are missing or out of bounds; if both are normal, calculate the ice thickness through the comprehensive analysis mechanism (9); if missing or out of bounds, further determine whether an image can be captured and whether the captured image can identify the simulated wire (4); if the image can be captured and the simulated wire (4) can be identified, output the ice thickness calculated by the AI ​​image segmentation algorithm; if the image cannot be captured or the simulated wire (4) in the image cannot be identified, output the value of the last ice thickness.

6. The ice thickness monitoring method based on AI image according to claim 5 is characterized in that: The fault verification module (94) in step d4 is also provided with a logic for preventing data mutation, wherein the steps are as follows: determining whether the difference between the ice coating thickness of the simulated conductor (4) obtained this time and the ice coating thickness of the simulated conductor (4) obtained last time is greater than 2 mm; if so, changing the ice coating thickness of the simulated conductor (4) obtained this time to the value of the ice coating thickness of the simulated conductor (4) obtained last time plus 2 mm.

Images