Power transmission line unmanned aerial vehicle autonomous inspection adaptive optimization method based on AI assistance

Abstract

The invention discloses a power transmission line unmanned aerial vehicle autonomous inspection adaptive optimization method based on AI assistance, which is used in the field of artificial intelligence, and comprises the following steps: collecting environmental parameter data of a power transmission line, and carrying out data labeling; training a power line distance and light sensing and identifying module by using the marked data, and identifying the distance and light change of the power transmission line; adjusting the position and the flight path of the unmanned aerial vehicle according to the environment data collected in real time and the distance of the power transmission line; and generating an optimal inspection route based on the environment data and the position and the flight path of the unmanned aerial vehicle, and performing autonomous inspection according to the optimal inspection route in the inspection process. According to the method, heuristic path planning and statistical optimization are combined, and the accuracy and adaptability of path planning are realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and more specifically, to an adaptive optimization method for autonomous inspection of power transmission lines by unmanned aerial vehicles (UAVs) based on AI assistance. Background Art

[0002] In modern power systems, the safe and stable operation of transmission lines is of vital importance. In order to ensure the reliability of transmission lines and reduce the occurrence of faults, regular inspections are essential. Traditional manual inspection methods have problems such as low efficiency, high cost and poor safety. Therefore, more and more power companies have begun to introduce drone technology to inspect transmission lines.

[0003] Drones can cover a large area of ​​power transmission lines in a short period of time, significantly improving inspection efficiency. Compared with manual inspections, drone inspections can greatly reduce manpower and time costs.

[0004] Since there is no effective mechanism to adjust camera parameters according to environmental changes, the quality of images collected during the inspection process may fluctuate significantly due to changes in light, affecting subsequent data processing and analysis; the inspection path generated by the path planning algorithm in the existing technology is not optimized enough, resulting in low inspection efficiency, long flight time, and high battery consumption, resulting in poor inspection results in complex or changing environments, increasing safety risks.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In order to overcome the above problems, the present invention aims to propose an adaptive optimization method for autonomous inspection of transmission lines by UAVs based on AI assistance. The purpose is to solve the problem that the path planning algorithm in the prior art may not be able to comprehensively consider the path distance, flight energy consumption and safety, and the generated inspection path is not optimized enough, resulting in low inspection efficiency, long flight time, and high battery consumption. During the inspection process, the flight path and camera parameters of the UAV cannot be dynamically adjusted according to the real-time collected data, resulting in poor inspection results in complex or changing environments, which increases safety risks.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] The adaptive optimization method for autonomous inspection of power transmission lines by unmanned aerial vehicles (UAVs) assisted by AI comprises the following steps:

[0009] S 1. A pre-configured drone equipped with a camera collects environmental parameter data of the power transmission line and annotates the data;

[0010] S2. Use the labeled data to train the power line distance and light perception recognition module to identify the distance and light changes of the transmission line;

[0011] S3. During the inspection process, the environmental data and the distance of the power transmission line are collected in real time, and the exposure parameters and focal length of the camera are adjusted based on the distance of the power line and the light changes. At the same time, the position and flight path of the drone are adjusted according to the environmental data collected in real time and the distance of the power transmission line;

[0012] S4. Generate the best inspection route based on environmental data and the location and flight path of the UAV, and perform autonomous inspections according to the best inspection route during the inspection process.

[0013] Optionally, using the labeled data to train a power line distance and light perception recognition module to recognize the distance and light changes of the power transmission line includes the following steps:

[0014] S21, obtaining labeled environmental parameter data, and dividing the environmental parameter data into a training set and a validation set;

[0015] S22, constructing the architecture of the power line distance and light perception recognition module, and training the power line distance and light perception recognition module using the training set;

[0016] S23, using the trained power line distance and light perception recognition module to extract transmission line image data, identify the location of the transmission line, and calculate the distance between the transmission line and the camera;

[0017] S24, integrating light change annotation data into model training, identifying light change features in the image, and determining the light conditions of the current environment;

[0018] S25. Use the validation set to evaluate the model performance and adjust the architecture of the power line distance and light perception recognition module based on the validation results.

[0019] Optionally, using the trained power line distance and light perception recognition module to extract the transmission line image data, identify the transmission line position, and calculate the distance between the transmission line and the camera includes the following steps:

[0020] S231, extracting transmission line image data related to the transmission line location from the annotated environmental parameter data, and performing preprocessing;

[0021] S232, inputting the pre-processed transmission line image data into the trained power line distance and light perception recognition module to identify the location information of the transmission line;

[0022] S233. Calculate the distance between the transmission line and the camera based on the identified transmission line position and known internal parameters of the camera.

[0023] Optionally, incorporating light change annotation data into model training, identifying light change features in an image, and determining the light conditions of the current environment include the following steps:

[0024] S241, obtaining light change feature annotation data, and extracting light change feature data;

[0025] S242, during the training process of the power line distance and light perception recognition module, using the light change feature data as an additional input feature, so that the power line distance and light perception recognition module learns the image features under various light conditions;

[0026] S243, using the trained power line distance and light perception recognition module to extract light change features from new power transmission line image data, and recognize light conditions in the power transmission line image data;

[0027] S244: Based on the identified light change characteristics, the light condition of the current environment is evaluated by setting a threshold.

[0028] Optionally, during the inspection process, real-time collection of environmental data and the distance of the power line, and adjustment of the exposure parameters and focal length of the camera based on the distance of the power line and the light change, and adjustment of the position and flight path of the drone based on the real-time collected environmental data and the distance of the power line include the following steps:

[0029] S31. During the inspection process, various parameter data of the current environment are collected in real time, and the distance between the transmission line and the drone is measured;

[0030] S32, inputting the real-time collected environmental data and transmission line distance data into the trained power line distance and light perception recognition module;

[0031] S33, adjusting the exposure parameters and focal length of the camera according to the output results of the power line distance and light perception recognition module;

[0032] S34. Adjust the position and flight path of the UAV based on the real-time collected data and the output of the model.

[0033] Optionally, adjusting the exposure parameters and focal length of the camera according to the output result of the power line distance and light perception recognition module includes the following steps:

[0034] S331, analyzing the output of the power line distance and light perception recognition module to obtain recommended camera exposure parameters and focal length adjustment values;

[0035] S332, obtaining current camera settings, including exposure parameters and focal length values, and calculating required exposure parameters and focal length adjustment amounts;

[0036] S333, applying the calculated new exposure parameter and focal length value to the camera setting, and monitoring the adjusted effect in real time;

[0037] S334. Record the camera parameters before and after adjustment.

[0038] Optionally, the camera exposure parameters include shutter speed, aperture size and sensitivity.

[0039] Optionally, generating an optimal inspection route based on environmental data and the position and flight path of the UAV, and performing autonomous inspection according to the optimal inspection route during the inspection process includes the following steps:

[0040] S41, using the A* algorithm to initialize the inspection path of the drone as the starting point of the Bayesian optimization model;

[0041] S42. define the objective function of the Bayesian optimization model;

[0042] S43, initializing the Bayesian optimization model, and evaluating the quality of the sampling path based on the objective function to obtain a path score;

[0043] S44, updating the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function;

[0044] S45, repeating the steps of S43-S45, iteratively optimizing the path parameters until the optimal path parameters are obtained, and generating the final optimal route for the drone inspection based on the optimal path parameters;

[0045] S46. Perform real-time inspections based on the optimal path parameters. During the inspection process, collect environmental data in real time and input it into the Bayesian optimization model, and autonomously adjust the path based on the latest environmental data and model suggestions.

[0046] Optionally, initializing the Bayesian optimization model and evaluating the quality of the sampled path based on the objective function to obtain the path score includes the following steps:

[0047] S431, setting initial parameters of Bayesian optimization, creating and initializing a Gaussian process model;

[0048] S432, adding the initial inspection path and the objective function value generated by the A* algorithm as data points to the Gaussian process model;

[0049] S433, sampling new path parameters from the path parameter space using a Bayesian optimization sampling strategy;

[0050] S434, evaluating the sampled path, calculating the objective function value, and obtaining the path score;

[0051] S435, adding the new path and its scoring data points to the Gaussian process model, and updating the training data of the model;

[0052] The formula for calculating the objective function value is:

[0053] S=ω1·L(x)+ω2·E(x)+ω3·C(x);

[0054] In the formula, S represents the path score;

[0055] L(x) represents the total distance of path x;

[0056] E(x) represents the total energy consumption of the UAV on path x;

[0057] C(x) represents the risk assessment value of potential danger on path x;

[0058] ω1 represents the weight of the adjusted path distance stored in the database;

[0059] ω2 represents the weight of energy consumption stored in the database;

[0060] ω3 represents the weight of the risk assessment value stored in the database.

[0061] Optionally, updating the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function comprises the following steps:

[0062] S441. Obtain path parameters and scores from the latest evaluation;

[0063] S442, adding the new path parameters and scores as data points to the training data set of the Gaussian process model;

[0064] S443, retraining the Gaussian process model based on the newly added data points;

[0065] S444, calculating a new posterior distribution of the objective function according to the retrained Gaussian process model;

[0066] S445. Using the new posterior distribution, adjust the sampling strategy of the Bayesian optimization model;

[0067] S446, checking whether the latest path score meets the preset convergence condition;

[0068] If the preset convergence condition is met, the iteration ends;

[0069] If the preset convergence condition is not met, the process returns to step S442.

[0070] Compared with the prior art, this application has the following beneficial effects:

[0071] 1. The present invention uses a pre-configured drone equipped with a camera to collect environmental parameter data of the transmission line, and performs data annotation, and trains the power line distance and light perception recognition module to identify the distance and light changes of the transmission line. During the inspection process, environmental data and the distance of the transmission line are collected in real time, and the trained recognition module is used to adjust the exposure parameters and focal length of the camera. At the same time, the position and flight path of the drone are adjusted according to the real-time data. This method ensures that accurate environmental information is obtained during the inspection process, improves image quality and inspection safety, and optimizes the flight path of the drone.

[0072] 2. The A* algorithm in the present invention generates an initial inspection path by considering path distance, flight energy consumption and safety, providing a good foundation for subsequent optimization. Then, the Bayesian optimization model iteratively updates and evaluates path quality, continuously adjusts and optimizes path parameters, and finally generates the optimal inspection route. In the actual inspection process, environmental data is collected in real time and the flight path is adjusted autonomously to ensure the efficiency and safety of the inspection process. The overall method combines heuristic path planning and statistical optimization to achieve the accuracy and adaptability of path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The above characteristics, features and advantages of the present invention and their implementation methods and methods will become more clearly understood in conjunction with the following description of the embodiments, which are described in detail in conjunction with the accompanying drawings. Herein, a schematic diagram is shown:

[0074] Figure 1 It is a flow chart of an adaptive optimization method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to an embodiment of the present invention. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0076] According to an embodiment of the present invention, an AI-assisted adaptive optimization method for autonomous inspection of power transmission lines by drones is provided.

[0077] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to an AI-assisted transmission line drone autonomous inspection adaptive optimization method according to an embodiment of the present invention, the transmission line drone autonomous inspection adaptive optimization method includes the following steps:

[0078] S 1. A pre-configured drone equipped with a camera collects environmental parameter data of the power transmission line and annotates the data.

[0079] It needs to be explained that first, the pre-configured drone-mounted camera needs to collect environmental parameter data. The collected data includes various detailed information of the transmission line, such as tower location, conductor height, etc.; then, the collected data is labeled for subsequent data processing and analysis to ensure that the drone can obtain accurate environmental information during the inspection process.

[0080] S2. Use the labeled data to train the power line distance and light perception recognition module to identify the distance and light changes of the transmission lines.

[0081] Preferably, using the labeled data to train the power line distance and light perception recognition module to recognize the distance and light changes of the power transmission line includes the following steps:

[0082] S21, obtaining labeled environmental parameter data, and dividing the environmental parameter data into a training set and a validation set;

[0083] S22, constructing the architecture of the power line distance and light perception recognition module, and training the power line distance and light perception recognition module using the training set;

[0084] S23, using the trained power line distance and light perception recognition module to extract transmission line image data, identify the location of the transmission line, and calculate the distance between the transmission line and the camera;

[0085] S24, integrating light change annotation data into model training, identifying light change features in the image, and determining the light conditions of the current environment;

[0086] S25. Use the validation set to evaluate the model performance and adjust the architecture of the power line distance and light perception recognition module based on the validation results.

[0087] Preferably, using the trained power line distance and light perception recognition module to extract the transmission line image data, identify the transmission line position, and calculate the distance between the transmission line and the camera includes the following steps:

[0088] S231, extracting transmission line image data related to the transmission line location from the annotated environmental parameter data, and performing preprocessing;

[0089] S232, inputting the pre-processed transmission line image data into the trained power line distance and light perception recognition module to identify the location information of the transmission line;

[0090] S233. Calculate the distance between the transmission line and the camera based on the identified transmission line position and known internal parameters of the camera.

[0091] Preferably, incorporating light change annotation data into model training, identifying light change features in an image, and determining the light conditions of the current environment include the following steps:

[0092] S241, obtaining light change feature annotation data, and extracting light change feature data;

[0093] S242, during the training process of the power line distance and light perception recognition module, using the light change feature data as an additional input feature, so that the power line distance and light perception recognition module learns the image features under various light conditions;

[0094] S243, using the trained power line distance and light perception recognition module to extract light change features from new power transmission line image data, and recognize light conditions in the power transmission line image data;

[0095] S244: Based on the identified light change characteristics, the light condition of the current environment is evaluated by setting a threshold.

[0096] The evaluation process is as follows:

[0097] First, the light change characteristic data are obtained, wherein the light change characteristic data include the average light intensity in the time window, the maximum light intensity in the observation period, the slope of the trend line of the light intensity change and the standard deviation of the light intensity.

[0098] The light change characteristic data is evaluated based on the light change characteristic evaluation function to obtain the light change evaluation index, wherein the light change characteristic evaluation function formula is as follows:

[0099]

[0100] Wherein, Lw is the average light intensity in the time window, which is calculated by averaging the light intensity values ​​measured in a certain time window (such as daily or hourly), Lm represents the maximum light intensity during the observation period, which is the maximum light intensity value recorded during the same observation period, α is the weight factor of the light intensity stored in the database, m is the slope of the trend line of light intensity change, which is obtained by linear regression analysis of the light intensity data in the time window, and represents the trend of light intensity change, β is the weight factor of the trend line slope stored in the database, and σ L is the standard deviation of light intensity, the standard deviation of light intensity value in the time window, which is used to measure the degree of change of light intensity, and γ is stored in the database The weight factor of , C(t) is the light change evaluation index.

[0101] For the weight factors α, β, and γ stored in the database, it is necessary to collect relevant historical data, such as the average light intensity within the time window, the maximum light intensity, the slope of the trend line of light intensity change, and the standard deviation of light intensity. Then, these variables are input into the regression model as independent variables, and the actual light intensity change or other related response variables are used as dependent variables. By fitting this model, the regression coefficient of each independent variable can be obtained, and these coefficients directly represent the weight of the corresponding independent variable in the prediction model. In this way, the size and sign of each coefficient will indicate the contribution of the corresponding input variable to the output variable, thereby effectively determining the value of each weight factor.

[0102] The light change evaluation index is compared with the light change threshold stored in the database. If the light change evaluation index is greater than the light change threshold stored in the database, the light condition of the current environment is frequently changing. If the light change evaluation index is less than or equal to the light change threshold stored in the database, the light condition of the current environment is stable. The light change threshold stored in the database is obtained by a detailed analysis of the historical light data. The specific method is to first collect light intensity data under different environments over a period of time, and then use statistical methods such as percentiles and standard deviations to analyze the distribution of light intensity changes. By determining the common range and extreme change points of light changes, a threshold can be set to distinguish between the stability and frequent changes of light conditions. This threshold helps identify significant light changes so that corresponding measures can be taken.

[0103] It should be explained that the power line distance and light perception recognition module is a convolutional neural network model. First, the labeled environmental parameter data is obtained and divided into a training set and a validation set. Then, a CNN model architecture is constructed and trained using the training set. The trained model is used to extract transmission line image data, identify the location of the transmission line and calculate the distance between it and the camera. At the same time, the light change annotation data is incorporated into the model training so that it can recognize the light change characteristics in the image and judge the light conditions of the current environment. Finally, the model performance is evaluated through the validation set, and the model architecture is adjusted according to the verification results to improve the recognition accuracy and robustness.

[0104] S3. During the inspection process, environmental data and the distance of the power lines are collected in real time, and the exposure parameters and focal length of the camera are adjusted based on the distance of the power lines and light changes. At the same time, the position and flight path of the drone are adjusted according to the real-time collected environmental data and the distance of the power lines.

[0105] Preferably, during the inspection process, real-time collection of environmental data and the distance of the power line is performed, and the exposure parameters and focal length of the camera are adjusted based on the distance of the power line and the light change, and the position and flight path of the drone are adjusted according to the real-time collected environmental data and the distance of the power line, including the following steps:

[0106] S31. During the inspection process, various parameter data of the current environment are collected in real time, and the distance between the transmission line and the drone is measured;

[0107] S32, inputting the real-time collected environmental data and transmission line distance data into the trained power line distance and light perception recognition module;

[0108] S33, adjusting the exposure parameters and focal length of the camera according to the output results of the power line distance and light perception recognition module;

[0109] S34. Adjust the position and flight path of the UAV based on the real-time collected data and the output of the model.

[0110] Preferably, adjusting the exposure parameters and focal length of the camera according to the output result of the power line distance and light perception recognition module comprises the following steps:

[0111] S331, analyzing the output of the power line distance and light perception recognition module to obtain recommended camera exposure parameters and focal length adjustment values;

[0112] The process of obtaining the recommended camera exposure parameters and focus adjustment values ​​is as follows:

[0113] Obtain the output of the power line distance and light perception recognition module, namely, the power line distance, the average light intensity in the time window, the maximum light intensity during the observation period, the slope of the trend line of light intensity change, and the standard deviation of the light intensity;

[0114] The output of the power line distance and light perception recognition module is compared with the reference output data corresponding to the recommended combination of each exposure parameter and focal length adjustment value in the database to obtain a comparison index, and the reference output data includes a reference power line distance, a reference average light intensity in a time window, a reference maximum light intensity during an observation period, a reference trend line slope of light intensity change, and a reference standard deviation of light intensity;

[0115] The calculation formula of the comparison index is as follows:

[0116]

[0117] Where M is the comparison index, D is the distance between power lines, and D f is the reference power line distance, L wf is the reference average light intensity within the time window, L mf is the reference maximum value of light intensity during the observation period, m f is the reference maximum value of light intensity during the observation period, σ Lf is the reference standard deviation of light intensity.

[0118] The comparison indexes are arranged in ascending order, and the recommended combination of exposure parameter and focal length adjustment value corresponding to the smallest comparison index is the recommended camera exposure parameter and focal length adjustment value.

[0119] S332, obtaining current camera settings, including exposure parameters and focal length values, and calculating required exposure parameters and focal length adjustment amounts;

[0120] S333, applying the calculated new exposure parameter and focal length value to the camera setting, and monitoring the adjusted effect in real time;

[0121] S334. Record the camera parameters before and after adjustment.

[0122] Preferably, the camera exposure parameters include shutter speed, aperture size and sensitivity.

[0123] It needs to be explained that during the inspection process, environmental data and the distance of the transmission lines are collected in real time, and the trained power line distance and light perception recognition module is used to adjust the camera's exposure parameters and focal length. At the same time, the position and flight path of the UAV are adjusted according to the real-time collected data. The specific steps include real-time collection of current environment and distance data, inputting them into the recognition module, adjusting the camera settings (such as shutter speed, aperture size and sensitivity) according to the module output results, and optimizing the UAV's flight path based on the adjustment results and real-time data. This can ensure image quality and flight safety during the UAV inspection process.

[0124] S4. Generate the best inspection route based on environmental data and the location and flight path of the UAV, and perform autonomous inspections according to the best inspection route during the inspection process.

[0125] Preferably, generating an optimal inspection route based on environmental data and the position and flight path of the drone, and performing autonomous inspection according to the optimal inspection route during the inspection process includes the following steps:

[0126] S41, using the A* algorithm to initialize the inspection path of the drone as the starting point of the Bayesian optimization model;

[0127] S42. define the objective function of the Bayesian optimization model;

[0128] S43, initializing the Bayesian optimization model, and evaluating the quality of the sampling path based on the objective function to obtain a path score;

[0129] S44, updating the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function;

[0130] S45, repeating the steps of S43-S45, iteratively optimizing the path parameters until the optimal path parameters are obtained, and generating the final optimal route for the drone inspection based on the optimal path parameters;

[0131] S46. Perform real-time inspections based on the optimal path parameters. During the inspection process, collect environmental data in real time and input it into the Bayesian optimization model, and autonomously adjust the path based on the latest environmental data and model suggestions.

[0132] Preferably, initializing the Bayesian optimization model and evaluating the quality of the sampled path based on the objective function to obtain the path score comprises the following steps:

[0133] S431, setting initial parameters of Bayesian optimization, creating and initializing a Gaussian process model;

[0134] S432, adding the initial inspection path and the objective function value generated by the A* algorithm as data points to the Gaussian process model;

[0135] S433, sampling new path parameters from the path parameter space using a Bayesian optimization sampling strategy;

[0136] S434, evaluating the sampled path, calculating the objective function value, and obtaining the path score;

[0137] S435, adding the new path and its scoring data points to the Gaussian process model, and updating the training data of the model;

[0138] The formula for calculating the objective function value is:

[0139] S=ω1·L(x)+ω2·E(x)+ω3·C(x);

[0140] In the formula, s represents the path score;

[0141] L(x) represents the total distance of path x;

[0142] E(x) represents the total energy consumption of the UAV on path x;

[0143] C(x) represents the risk assessment value of potential danger on path x;

[0144] ω1 represents the weight of the adjusted path distance stored in the database;

[0145] ω2 represents the weight of energy consumption stored in the database;

[0146] ω3 represents the weight of the risk assessment value stored in the database.

[0147] Preferably, updating the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function comprises the following steps:

[0148] S441. Obtain path parameters and scores from the latest evaluation;

[0149] S442, adding the new path parameters and scores as data points to the training data set of the Gaussian process model;

[0150] S443, retraining the Gaussian process model based on the newly added data points;

[0151] S444, calculating a new posterior distribution of the objective function according to the retrained Gaussian process model;

[0152] S445. Using the new posterior distribution, adjust the sampling strategy of the Bayesian optimization model;

[0153] S446, checking whether the latest path score meets the preset convergence condition;

[0154] If the preset convergence condition is met, the iteration ends;

[0155] If the preset convergence condition is not met, the process returns to step S442.

[0156] It should be explained that the A* algorithm is first used to initialize the inspection path of the drone as the starting point of the Bayesian optimization model, and then the objective function of the Bayesian optimization model is defined; then, the Bayesian optimization model is initialized, and the quality of the sampling path is evaluated based on the objective function to obtain a path score; then, the Bayesian optimization model is updated based on the path score, and the posterior distribution of the objective function is adjusted, and this process is repeated until the optimal path parameters are obtained to generate the final optimal inspection route; in actual inspections, environmental data is collected in real time and input into the Bayesian optimization model, and the flight path is autonomously adjusted according to the latest data and model recommendations to ensure the efficiency and safety of the inspection.

[0157] The steps of initializing the inspection path of the UAV using the A* algorithm include: first, determining the starting and ending positions within the inspection area, as well as the obstacles and no-fly zones that need to be avoided. Then, based on the environmental map and the flight parameters of the UAV, the A* algorithm is used for path planning. The A* algorithm estimates the cost from the starting point to the end point and selects the optimal path in combination with a heuristic function. During the path search process, the A* algorithm will evaluate each possible path node, select the path with the lowest total cost, and finally generate an initial inspection path. This path not only takes into account the shortest distance, but also takes into account flight energy consumption and safety, providing a good starting point for subsequent Bayesian optimization.

[0158] Although the present invention has been disclosed as above with preferred embodiments, the embodiments are merely examples for the convenience of description and are not intended to limit the present invention. Those skilled in the art may make several changes and modifications without departing from the spirit and scope of the present invention. The scope of protection claimed by the present invention shall be based on the claims.

Claims

1. An adaptive optimization method for autonomous inspection of power transmission lines by drones based on AI, characterized in that: The transmission line UAV autonomous inspection adaptive optimization method comprises the following steps: S 1. A pre-configured drone equipped with a camera collects environmental parameter data of the power transmission line and annotates the data; S2. Use the labeled data to train the power line distance and light perception recognition module to identify the distance and light changes of the transmission line; S3. During the inspection process, the environmental data and the distance of the power transmission line are collected in real time, and the exposure parameters and focal length of the camera are adjusted based on the distance of the power line and the light changes. At the same time, the position and flight path of the drone are adjusted according to the environmental data collected in real time and the distance of the power transmission line; S4. Generate the best inspection route based on environmental data and the location and flight path of the UAV, and perform autonomous inspections according to the best inspection route during the inspection process.

2. The method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 1 is characterized in that: The method of using the labeled data to train the power line distance and light perception recognition module to recognize the distance and light changes of the power transmission line includes the following steps: S21, obtaining labeled environmental parameter data, and dividing the environmental parameter data into a training set and a validation set; S22, constructing the architecture of the power line distance and light perception recognition module, and training the power line distance and light perception recognition module using the training set; S23, using the trained power line distance and light perception recognition module to extract transmission line image data, identify the location of the transmission line, and calculate the distance between the transmission line and the camera; S24, integrating light change annotation data into model training, identifying light change features in the image, and determining the light conditions of the current environment; S25. Use the validation set to evaluate the model performance and adjust the architecture of the power line distance and light perception recognition module based on the validation results.

3. The method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 2 is characterized in that: The method of extracting transmission line image data using the trained power line distance and light perception recognition module, identifying the position of the transmission line, and calculating the distance between the transmission line and the camera includes the following steps: S231, extracting transmission line image data related to the transmission line location from the annotated environmental parameter data, and performing preprocessing; S232, inputting the pre-processed transmission line image data into the trained power line distance and light perception recognition module to identify the location information of the transmission line; S233. Calculate the distance between the transmission line and the camera based on the identified transmission line position and known internal parameters of the camera.

4. The method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 3 is characterized in that: Incorporating light change annotation data into model training, identifying light change features in an image, and determining the light conditions of the current environment include the following steps: S241, obtaining light change feature annotation data, and extracting light change feature data; S242, during the training process of the power line distance and light perception recognition module, using the light change feature data as an additional input feature, so that the power line distance and light perception recognition module learns the image features under various light conditions; S243, using the trained power line distance and light perception recognition module to extract light change features from new power transmission line image data, and recognize light conditions in the power transmission line image data; S244: Based on the identified light change characteristics, the light condition of the current environment is evaluated by setting a threshold.

5. The method for adaptive optimization of autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 1 is characterized in that: During the inspection process, real-time collection of environmental data and the distance of the power line, and adjustment of the exposure parameters and focal length of the camera based on the distance of the power line and the light change, and adjustment of the position and flight path of the drone based on the real-time collected environmental data and the distance of the power line include the following steps: S31. During the inspection process, various parameter data of the current environment are collected in real time, and the distance between the transmission line and the drone is measured; S32, inputting the real-time collected environmental data and transmission line distance data into the trained power line distance and light perception recognition module; S33, adjusting the exposure parameters and focal length of the camera according to the output results of the power line distance and light perception recognition module; S34. Adjust the position and flight path of the UAV based on the real-time collected data and the output of the model.

6. The method for adaptive optimization of autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 5 is characterized in that: The step of adjusting the exposure parameters and focal length of the camera according to the output result of the power line distance and light perception recognition module comprises the following steps: S331, analyzing the output of the power line distance and light perception recognition module to obtain recommended camera exposure parameters and focal length adjustment values; S332, obtaining current camera settings, including exposure parameters and focal length values, and calculating required exposure parameters and focal length adjustment amounts; S333, applying the calculated new exposure parameter and focal length value to the camera setting, and monitoring the adjusted effect in real time; S334. Record the camera parameters before and after adjustment.

7. The method for adaptive optimization of autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 6 is characterized in that: The camera exposure parameters include shutter speed, aperture size and sensitivity.

8. The method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 1 is characterized in that: Generating the best inspection route based on environmental data and the position and flight path of the UAV, and performing autonomous inspection according to the best inspection route during the inspection process includes the following steps: S41, using the A* algorithm to initialize the inspection path of the drone as the starting point of the Bayesian optimization model; S42. define the objective function of the Bayesian optimization model; S43, initializing the Bayesian optimization model, and evaluating the quality of the sampling path based on the objective function to obtain a path score; S44, updating the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function; S45, repeating the steps of S43-S45, iteratively optimizing the path parameters until the optimal path parameters are obtained, and generating the final optimal route for the drone inspection based on the optimal path parameters; S46. Perform real-time inspections based on the optimal path parameters. During the inspection process, collect environmental data in real time and input it into the Bayesian optimization model, and autonomously adjust the path based on the latest environmental data and model suggestions.

9. The method for autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 8 is characterized in that: Initializing the Bayesian optimization model and evaluating the quality of the sampled path based on the objective function to obtain the path score includes the following steps: S431, setting initial parameters of Bayesian optimization, creating and initializing a Gaussian process model; S432, adding the initial inspection path and the objective function value generated by the A* algorithm as data points to the Gaussian process model; S433, sampling new path parameters from the path parameter space using a Bayesian optimization sampling strategy; S434, evaluating the sampled path, calculating the objective function value, and obtaining the path score; S435, adding the new path and its scoring data points to the Gaussian process model, and updating the training data of the model; The formula for calculating the objective function value is: S=ω1·L(x)+ω2·E(x)+ω3·C(x); In the formula, s represents the path score; L(x) represents the total distance of path x; E(x) represents the total energy consumption of the UAV on path x; C(x) represents the risk assessment value of potential danger on path x; ω1 represents the weight of the adjusted path distance stored in the database; ω2 represents the weight of energy consumption stored in the database; ω3 represents the weight of the risk assessment value stored in the database.

10. The method for adaptive optimization of autonomous inspection of power transmission lines by unmanned aerial vehicles based on AI assistance according to claim 9 is characterized in that: The updating of the Bayesian optimization model based on the path score and adjusting the posterior distribution of the objective function comprises the following steps: S441. Obtain path parameters and scores from the latest evaluation; S442, adding the new path parameters and scores as data points to the training data set of the Gaussian process model; S443, retraining the Gaussian process model based on the newly added data points; S444, calculating a new posterior distribution of the objective function according to the retrained Gaussian process model; S445. Using the new posterior distribution, adjust the sampling strategy of the Bayesian optimization model; S446, checking whether the latest path score meets the preset convergence condition; If the preset convergence condition is met, the iteration ends; If the preset convergence condition is not met, the process returns to step S442.

Images