CN117710838A - Unmanned aerial vehicle high-voltage power network video inspection identification method and system based on deep learning - Google Patents

Abstract

The invention discloses a deep learning-based unmanned aerial vehicle high-voltage power network video inspection identification method and a deep learning-based unmanned aerial vehicle high-voltage power network video inspection identification system, which relate to the technical field of unmanned aerial vehicle inspection, and comprise the following steps: extracting outline features from video data, and marking the distribution positions of target components in a high-voltage power grid; controlling the virtual unmanned aerial vehicle to simulate flying according to the initial flying path; dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path; extracting defect characteristics of the electric wire from the inspection video data by using a texture recognition technology; based on the defect characteristics of different categories, respectively dispatching unmanned aerial vehicle groups of corresponding types to treat the cable defects; and recording and reporting the processing result, and making a maintenance plan based on the health condition of the high-voltage power grid. The method and the system can set the optimal initial flight path and correct the optimal initial flight path, so that the navigation distance and time of the unmanned aerial vehicle are reduced to the greatest extent.

Description

Unmanned aerial vehicle high-voltage power network video inspection identification method and system based on deep learning

Technical Field

The invention relates to the technical field of unmanned aerial vehicle inspection, in particular to an unmanned aerial vehicle high-voltage power grid video inspection identification method and system based on deep learning.

Background

The high-voltage power grid is a power system for transmitting and distributing high-voltage power, and is composed of a power transmission line, a transformer substation, a power distribution network and related equipment, wherein the power transmission line is used for transmitting the generated high-voltage power to various places, the power transmission line usually adopts a high-voltage direct current or high-voltage alternating current mode, and power transmission is carried out through a cable or overhead line, and the power transmission line usually consists of a wire, an insulator, a pole tower, a ground wire and the like.

The high-voltage power grid can transmit power from a power plant to a far place through the high-voltage power transmission line, so that long-distance transmission of the power is realized, and the current in the power transmission process is reduced by improving the voltage of the power transmission line, so that the power loss is reduced. In the inspection and identification of the high-voltage power grid, the unmanned aerial vehicle is generally provided with corresponding sensors and equipment so as to acquire images, videos and data of the power grid equipment, and further detect and identify defect characteristics in the high-voltage power grid.

However, when the power demand is great, need to build a large amount of and intensive electric power transmission lines and distribution lines, when carrying out video acquisition to the high-voltage power network among the prior art, can't carry out accurate detection to the electric wire intensive region, and then make can't catch the general view in the electric wire intensive region completely, lack the depth information of electric wire and can't confirm the relation of connection between the electric wire simultaneously, this can lead to splitting and discernment to the electric wire and produce the difficulty to unable accurate detection and analysis electric wire's overall structure in the electric wire intensive region, and then influence the accurate location to the electric wire of follow-up.

For the problems in the related art, no effective solution has been proposed at present.

Disclosure of Invention

The invention mainly aims to provide a deep learning-based unmanned aerial vehicle high-voltage power network video inspection identification method and system, so as to overcome the technical problems existing in the prior art.

For this purpose, the invention adopts the following specific technical scheme:

according to one aspect of the invention, a deep learning-based unmanned aerial vehicle high-voltage power network video inspection and identification method is provided, and the deep learning-based unmanned aerial vehicle high-voltage power network video inspection and identification method comprises the following steps:

s1, acquiring video data of a high-voltage power grid in real time, extracting outline features from the video data, and marking distribution positions of target components in the high-voltage power grid;

s2, setting an initial flight path by combining the distribution positions of the target components, and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path;

s3, dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path, and collecting inspection video data in real time;

s4, extracting defect characteristics of the electric wire from the inspection video data by using a texture recognition technology, and performing defect classification and defect position positioning on the defect characteristics by using a deep learning model;

S5, respectively dispatching unmanned aerial vehicle groups of corresponding types based on the defect characteristics of different types to treat the cable defects;

and S6, recording and reporting the processing result, and making an overhaul plan based on the health condition of the high-voltage power grid.

Optionally, video data of the high-voltage power grid is acquired in real time, contour features are extracted from the video data, and the marking of the distribution positions of target components in the high-voltage power grid comprises the following steps:

s11, acquiring video data of a high-voltage power grid, and preprocessing the video data;

s12, extracting an image sequence from the preprocessed video data, extracting a contour line of the image sequence and matching the contour line, and identifying a target component of the high-voltage power grid, wherein the target component at least comprises an electric wire, a pole tower and an insulator;

s13, extracting contour points from the matched contour lines, and detecting and dividing an electric wire dense region in the image sequence;

s14, acquiring depth information of each wire in the wire dense area by utilizing a vision principle recovery algorithm, deducing a front-back connection relation of the wires, and describing the actual trend of the wires;

s15, marking the distribution positions of target components in the high-voltage power grid by respectively combining the actual trend of the electric wires.

Optionally, extracting contour points from the matched contour lines, and detecting and dividing the wire-dense region in the image sequence includes the following steps:

s131, extracting contour points from the matched contour lines, and judging whether an electric wire dense area exists in the image sequence;

s132, acquiring a chain code value of a boundary point of an electric wire dense region in an image sequence, and carrying out boundary tracking on the electric wire dense region in a clockwise direction;

s133, analyzing the chain code value to judge a chain code pair of the boundary point of the electric wire dense region, and judging whether the boundary point of the electric wire dense region is a characteristic point or not according to the chain code pair;

s134, detecting characteristic points and minimum circumscribed rectangles in the wire dense area, and detecting dense wires;

s135, performing bump detection on the dense electric wires to obtain the concave span and depth of the concave area, and if the preset condition is met, taking the concave area as the concave area of the dense electric wires and marking the direction of the concave area;

s136, calculating the distance from the feature point in the concave area to the top points on two sides of the convex point, and taking the shortest distance as the distance from the current feature point to the top points of the convex point;

s137, comparing the characteristic points in the concave areas, taking the maximum characteristic point as a dividing point of the dense wire area, judging whether the concave areas meeting the preset conditions exist, if so, jumping to the step S136, otherwise, continuing to execute the step S138;

S138, connecting the division points with different directions, analyzing whether the divided wire dense areas meet the dense wire judging principle, and judging whether stranded wire characteristics exist in the dense wires.

Optionally, the depth information of each wire in the wire-intensive area is obtained by using a vision principle recovery algorithm, the front-back connection relation of the wires is deduced, and the actual trend of the wires is described, which comprises the following steps:

s141, dividing the contour lines in the wire dense area into segments with equal lengths, and extracting the maximum tangent point of the current segment from each segment as a key point;

s142, establishing a corresponding relation between the current image and the key frame by utilizing homography transformation, and selecting a key point in the current image;

s143, drawing a polar line corresponding to the selected key point in the adjacent image, and taking the key point with the minimum distance with the polar line as a corresponding point in the adjacent image;

s144, calculating depth information of a corresponding point on the contour line by using a visual principle recovery algorithm, and using an average value of horizontal and vertical interpolation results as a depth value of pixels in the contour line;

s145, deducing the front-back connection relation of the electric wires through adjacent depth values, and describing the actual trend of the electric wires according to the front-back connection relation.

Optionally, setting an initial flight path in combination with the distribution position of the target member, and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path includes the following steps:

s21, constructing a three-dimensional digital model based on a high-voltage power grid, and modeling each component of the high-voltage power grid in the three-dimensional digital model to obtain a three-dimensional component body;

s22, setting an initial flight path based on the distribution position of the three-dimensional component body in the three-dimensional digital model, and carrying out simulated flight on the virtual unmanned aerial vehicle according to the initial flight path;

s23, adjusting control instructions of the virtual aircrafts in real time according to flight positions, flight speeds and flight postures of the virtual aircrafts in simulated flight;

s24, acquiring environment information, obstacle positions and height changes of the virtual unmanned aerial vehicle in simulated flight, and correcting an initial flight path by combining an environment sensing algorithm;

s25, integrating the corrected flight path and the distribution positions of the three-dimensional components to ensure that the virtual unmanned aerial vehicle can cover all the three-dimensional components in simulated flight.

Optionally, the step of dispatching the unmanned aerial vehicle groups of the corresponding types to treat the cable defects based on the defect characteristics of different categories respectively includes the following steps:

S51, obtaining defect characteristics of different types, wherein the defect characteristics at least comprise surface cracks, surface abrasion, floaters on the surface of the electric wire and wire strands;

s52, constructing a collaborative game model, taking an unmanned aerial vehicle as a game participant, and synchronously executing the flight direction and the flight step length of the task decision of each game participant according to the step length constraint condition;

s53, combining the cooperative game model with a task priority scheduling algorithm according to the importance and the emergency degree of the defect characteristics;

s54, solving the cooperative game model, determining an optimal task strategy of each game participant, and dispatching an unmanned aerial vehicle group of a corresponding type to a corresponding task for defect processing;

s55, each unmanned aerial vehicle unit executes processing tasks of corresponding types of defect features according to the dispatching instructions, the processing tasks at least comprise cleaning tasks by using the cleaning unmanned aerial vehicle unit, inspection tasks by using the inspection unmanned aerial vehicle unit, and cruising processing tasks by using the cruising unmanned aerial vehicle unit;

s56, information sharing and resource coordination among the unmanned aerial vehicle units are achieved by utilizing a wireless communication technology, and the positions and defect processing progress of the unmanned aerial vehicle units are monitored in real time.

Optionally, constructing a collaborative game model, taking the unmanned aerial vehicle as a game participant, and synchronously executing the flight direction and the flight step length of the self task decision of each game participant according to the step length constraint condition, wherein the flight direction and the flight step length comprise the following steps:

s521, acquiring a state variable and a control variable of the unmanned aerial vehicle unit at a preset moment, and constructing a cooperative game model of the unmanned aerial vehicle unit;

s522, randomly generating initial positions of the unmanned aerial vehicle units, and setting initial step sizes of the unmanned aerial vehicle units by combining the perceived radius and the communication radius of the unmanned aerial vehicle units;

s523, detecting all flight environment information of each unmanned aerial vehicle in a sensing range at the current position, and carrying out local information exchange with the adjacent unmanned aerial vehicle in a communication range;

s524, each unmanned aerial vehicle unit decides an optimal flight direction with the maximum profit function at the current moment according to the acquired local information;

s525, determining the maximum feasible step length of the current moment according to the determined optimal flight direction of the current moment, and taking the maximum feasible step length as the movement step length of the current moment;

s526, each unmanned aerial vehicle unit simultaneously executes the optimal flight direction and the motion step length which are independently determined, and moves to a new position, and finally returns to the step S522.

Optionally, determining the maximum feasible step length of the current moment according to the determined optimal flight direction of the current moment, and taking the maximum feasible step length as the motion step length of the current moment comprises the following steps:

s5251, when the number of adjacent unmanned aerial vehicles is greater than a preset threshold, taking the maximum step length as the current movement step length;

and S5252, calculating the maximum feasible step length capable of keeping all the adjacent unmanned aerial vehicles according to the flight distance between the adjacent unmanned aerial vehicles and the optimal flight direction at the current moment when the number of the adjacent unmanned aerial vehicles is smaller than a preset threshold value, and taking the maximum feasible step length as the movement step length at the current moment.

Optionally, the expression of the benefit function is:

in the method, in the process of the invention,respectively indicate that the predicted current moment i is at the selection theta _-i (k) The distance sum and the distance standard deviation between the next moment and all adjacent unmanned aerial vehicles;

n represents the number of adjacent unmanned aerial vehicles;

θ (k) and θ _-i (k) Respectively representing the flight direction and the optimal flight direction;

i represents the number of game participants;

j represents the j-th adjacent unmanned aerial vehicle;

θ represents a decision variable;

k represents the kth time;

u represents an effect function value;

f represents a benefit function value;

d represents the distance.

According to another aspect of the invention, the invention also provides an unmanned aerial vehicle high-voltage power network video inspection recognition system based on deep learning, which comprises a component position labeling module, a simulated flight module, an inspection video acquisition module, a defect recognition module, a defect processing module and an inspection plan making module;

The component position labeling module is used for acquiring video data of the high-voltage power grid in real time, extracting contour features from the video data and labeling the distribution positions of target components in the high-voltage power grid;

the simulated flight module is used for setting an initial flight path in combination with the distribution position of the target component and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path;

the inspection video acquisition module is used for dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path and acquiring inspection video data in real time;

the defect recognition module is used for extracting defect characteristics of the electric wire from the inspection video data by utilizing a texture recognition technology, and carrying out defect classification and defect position positioning on the defect characteristics by utilizing a deep learning model;

the defect processing module is used for respectively dispatching unmanned aerial vehicle groups of corresponding types to process the cable defects based on the defect characteristics of different types;

and the overhaul plan making module is used for recording and reporting the processing result and making an overhaul plan based on the health condition of the high-voltage power grid.

The beneficial effects of the invention are as follows:

1. the invention can accurately identify the positions of the wires by extracting the contour points and detecting and dividing the wire dense areas, thereby improving the accuracy and precision of wire detection, being beneficial to positioning the positions and boundaries of the wires, providing more accurate power line information for unmanned aerial vehicle inspection, being capable of clearly displaying the areas of the wires, being beneficial to identifying and judging the existence of stranded wire characteristics and reducing the potential line fault risk.

2. According to the invention, the depth information of each wire in the wire dense area can be deduced according to the parallax information in the image through the vision principle recovery algorithm, so that the three-dimensional position and distance information of the wires are obtained, the spatial distribution of the wires can be more comprehensively understood, the wires can be accurately positioned, meanwhile, the topological relation among the wires can be known through deducing the front-rear connection relation of the wires, the structure and the characteristics of the power lines can be more comprehensively analyzed, and more accurate power line information can be further provided for unmanned aerial vehicle inspection.

3. According to the invention, by combining the distribution positions of the target components, the optimal initial flight path can be set and corrected, so that the navigation distance and time of the unmanned aerial vehicle are reduced to the greatest extent, the actual inspection efficiency and coverage rate are improved, the situation that the unmanned aerial vehicle repeatedly covers or does not fly in the actual inspection process is avoided, the risk that the unmanned aerial vehicle collides with other obstacles in the flight process is avoided, and the flight safety is improved.

4. According to the invention, the deep learning model is utilized to classify the defect characteristics, and the cooperative game model is combined with the task priority scheduling algorithm, so that the unmanned aerial vehicle unit of the corresponding type can be dispatched to the corresponding defect processing task according to the defect characteristics, thereby improving the defect processing efficiency, shortening the fault repairing period, simultaneously helping to reasonably allocate the resources of the unmanned aerial vehicle unit, allocating the defect processing task with higher priority to the unmanned aerial vehicle unit with higher experience and expertise, further helping to ensure the safety of defect processing, and reducing the potential risk and possibility of accidents.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.

In the drawings:

fig. 1 is a flowchart of a deep learning-based unmanned aerial vehicle high-voltage power network video inspection identification method according to an embodiment of the invention;

fig. 2 is a schematic block diagram of a deep learning-based unmanned aerial vehicle high-voltage power network video inspection recognition system according to an embodiment of the invention.

In the figure:

1. a component position marking module; 2. simulating a flight module; 3. a patrol video acquisition module; 4. a defect identification module; 5. a defect processing module; 6. and an overhaul plan making module.

Detailed Description

It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein.

As described in the background art, in the prior art, the connection relation between the wires cannot be determined due to the lack of depth information of the wires, which may cause difficulty in segmentation and recognition of the wires, and in order to solve the above problems, the invention provides an unmanned aerial vehicle high-voltage network video inspection recognition method and system based on deep learning.

Referring to the drawings and the specific embodiments, as shown in fig. 1, the method for identifying the video inspection of the high-voltage network of the unmanned aerial vehicle based on the deep learning according to the embodiment of the invention comprises the following steps:

s1, video data of a high-voltage power grid are obtained in real time, outline features are extracted from the video data, and distribution positions of target components in the high-voltage power grid are marked.

The method for acquiring the video data of the high-voltage power grid in real time, extracting outline features from the video data, and marking the distribution positions of target components in the high-voltage power grid comprises the following steps:

s11, acquiring video data of the high-voltage power grid, and preprocessing the video data.

It should be noted that, the unmanned aerial vehicle is used to carry the camera equipment or monitor equipment to obtain the video data of high-voltage power grid, ensure that the camera equipment possesses sufficient resolution to obtain high-quality video data.

The collected video data is transmitted to a computer or a server for subsequent preprocessing, and the integrity and the safety of the video data can be ensured through network transmission or storage equipment transmission.

The video data is cut and scaled according to the requirements to select the region of interest or adjust the size of the video, so that the method can help to concentrate on specific regions or reduce the data volume and improve the efficiency of subsequent processing.

The video data is de-noised and enhanced using image processing techniques to improve video quality and visualization.

S12, extracting an image sequence from the preprocessed video data, extracting an outline and matching the outline of the image sequence, and identifying a target component of the high-voltage power grid, wherein the target component at least comprises an electric wire, a pole tower and an insulator.

S13, extracting contour points from the matched contour lines, and detecting and dividing the electric wire dense areas in the image sequence.

The method for detecting and dividing the electric wire dense region in the image sequence comprises the following steps of:

s131, extracting contour points from the matched contour lines, and judging whether an electric wire dense area exists in the image sequence.

S132, acquiring chain code values of boundary points of the wire dense region in the image sequence, and carrying out boundary tracking on the wire dense region in a clockwise direction.

It should be noted that, performing chain code calculation on each contour, that is, mapping each boundary point in the contour into a corresponding chain code value, where the chain code is a coding mode for representing continuity of the contour, and a common chain code algorithm includes 8 connected chain codes and 4 connected chain codes; and extracting the chain code value of the wire dense region from the chain code calculation result, namely extracting the chain code value of the outline boundary point corresponding to the wire dense region.

S133, analyzing the chain code value to judge the chain code pair of the boundary point of the electric wire dense region, and judging whether the boundary point of the electric wire dense region is a characteristic point or not according to the chain code pair.

S134, detecting characteristic points and minimum circumscribed rectangles in the wire dense area, and detecting dense wires.

S135, performing bump detection on the dense electric wires to obtain the concave span and depth of the concave area, and if the preset condition is met, taking the concave area as the concave area of the dense electric wires and marking the direction of the concave area.

It should be noted that, the bump detection algorithm is used to detect the bump on the edge of the wire, where the bump is a local maximum value of the recess on the curve, and may represent the recess area of the wire; judging whether the conditions of the concave area are met or not according to the bump detection result and preset conditions; calculating a depression span according to the difference between the maximum and minimum abscissas of the depression area, and calculating a depression depth according to the difference between the maximum and minimum abscissas of the depression area; marking the concave areas meeting the preset conditions, and marking the directions of the concave areas.

S136, calculating the distance from the feature point in the concave area to the top points on two sides of the convex point, and taking the shortest distance as the distance from the current feature point to the top points of the convex point.

S137, comparing the characteristic points in the concave areas, taking the maximum characteristic point as a dividing point of the dense wire area, judging whether the concave areas meeting the preset conditions exist, if so, jumping to the step S136, otherwise, continuing to execute the step S138.

S138, connecting the division points with different directions, analyzing whether the divided wire dense areas meet the dense wire judging principle, and judging whether stranded wire characteristics exist in the dense wires.

The stranded wire feature refers to a wire harness formed by twisting two or more wires in a power line, and whether the stranded wire feature exists or not is judged by analyzing the shape, direction, bending degree and other features of the wire segment in the wire dense region. In addition, image processing technology such as Hough transformation, morphological operation and the like can be utilized to perform shape matching or feature extraction on the line segments, so that whether stranded wire features exist or not can be further judged.

It should be noted that, using hough transform and judging whether the twisted wire features exist in the dense wires includes the following steps:

Hough transform is applied to the boundary of the wire-dense region obtained by edge detection to detect the characteristics of a straight line or curve.

And extracting parameters of the straight line or curve, such as a starting point, an end point, an angle, a bending degree and the like, according to the Hough transformation result.

Judging whether the detected straight line or curve belongs to the dense electric wire or not according to preset judging conditions, wherein the judging conditions can be set according to actual requirements and experience.

Further analyzing the straight line or curve of the dense wire to detect whether the stranded wire features exist, so that the stranded wire features of the straight line or curve, such as the intersection between the line segments, the relative position relationship and the like, are analyzed by using a specific algorithm or technology.

S14, acquiring depth information of each wire in the wire dense area by using a vision principle recovery algorithm, deducing the front-back connection relation of the wires, and describing the actual trend of the wires.

The depth information of each wire in the wire dense area is obtained by utilizing a vision principle recovery algorithm, the front-back connection relation of the wires is deduced, and the actual trend of the wires is described, wherein the method comprises the following steps of:

s141, dividing the contour line in the wire dense area into segments with equal lengths, and extracting the maximum tangent point of the current segment from each segment as a key point.

S142, establishing a corresponding relation between the current image and the key frame by utilizing homography transformation, and selecting a key point in the current image.

And S143, drawing the epipolar line corresponding to the selected key point in the adjacent image, and taking the key point with the minimum distance with the epipolar line as the corresponding point in the adjacent image.

S144, calculating the depth information of the corresponding point on the contour line by using a visual principle recovery algorithm, and using the average value of the horizontal and vertical interpolation results as the depth value of the pixels in the contour line.

It should be noted that, the depth information is calculated by using a visual principle recovery algorithm, and the visual principle recovery algorithm can be selected according to specific requirements and algorithms, for example, a calculation method based on stereoscopic vision or a calculation method based on optical flow, etc.

The pixel points in the contour line are subjected to horizontal and vertical interpolation according to the known depth information around the pixel points to obtain an approximate depth value, the linear interpolation or other interpolation methods can be used for calculation, and the average value of the horizontal and vertical interpolation results is calculated to obtain the depth value of the pixels in the contour line.

S145, deducing the front-back connection relation of the electric wires through adjacent depth values, and describing the actual trend of the electric wires according to the front-back connection relation.

S15, marking the distribution positions of target components in the high-voltage power grid by respectively combining the actual trend of the electric wires.

S2, setting an initial flight path by combining the distribution positions of the target components, and controlling the virtual unmanned aerial vehicle to simulate flight according to the initial flight path.

The method for controlling the virtual unmanned aerial vehicle to simulate flying according to the initial flying path comprises the following steps of:

s21, constructing a three-dimensional digital model based on the high-voltage power grid, and modeling each component of the high-voltage power grid in the three-dimensional digital model to obtain a three-dimensional component body.

It should be noted that, constructing a three-dimensional digital model based on a high-voltage power grid, and modeling each component of the high-voltage power grid in the three-dimensional digital model to obtain a three-dimensional component body, including the following steps:

and selecting a software tool suitable for three-dimensional modeling, and in the selected modeling software, carrying out three-dimensional modeling on each component of the high-voltage power grid according to the acquired data and related data.

According to the line topology structure of the high-voltage power grid, a three-dimensional topology structure of the whole power grid is constructed by using tools and functions in modeling software, and modeling can be performed according to the connection relation, the supporting structure and the like of the line.

According to actual conditions and requirements, texture mapping is carried out on the three-dimensional model so as to increase sense of realism, and mapping can be carried out by utilizing collected images or carrying out texture generation.

S22, setting an initial flight path based on the distribution position of the three-dimensional component body in the three-dimensional digital model, and carrying out simulated flight on the virtual unmanned aerial vehicle according to the initial flight path.

S23, adjusting control instructions of the virtual aircrafts in real time according to flight positions, flight speeds and flight postures of the virtual aircrafts in simulated flight.

S24, acquiring environment information, obstacle positions and height changes of the virtual unmanned aerial vehicle in simulated flight, and correcting an initial flight path by combining an environment sensing algorithm.

It should be noted that, obtaining the environmental information, the obstacle position and the height change of the virtual unmanned aerial vehicle in the simulated flight, and correcting the initial flight path by combining the environmental perception algorithm includes the following steps:

environmental information is acquired through sensors in the simulated environment, such as cameras, lidars and the like, and virtual environment is generated by using computer graphics technology, or real environmental information is acquired by using sensors in the real environment.

The obtained environmental information is detected and tracked by using an environmental perception algorithm, and the obstacle is identified and tracked by using technologies such as computer vision, deep learning and the like, and the information such as the position, the shape and the height of the obstacle is obtained.

The height change of the obstacle (the obstacle comprises a target component of a high-voltage power grid and other obstacles) is detected through an environment sensing algorithm, the height information of the obstacle is obtained through a sensor such as a laser radar, or the height estimation is carried out through an image processing technology.

And generating an initial flight path by using a path planning algorithm according to the starting point and the target point of the unmanned aerial vehicle.

Comparing and analyzing the acquired environmental information, the position and the height change of the obstacle and the like with the initial flight path, correcting the initial flight path by using an environment sensing algorithm, and performing path adjustment or avoidance action according to the position and the height change of the obstacle.

And applying the corrected flight path to the simulated flight of the virtual unmanned aerial vehicle, and visually displaying the virtual unmanned aerial vehicle in a simulated environment by utilizing a computer graphics technology.

S25, integrating the corrected flight path and the distribution positions of the three-dimensional components to ensure that the virtual unmanned aerial vehicle can cover all the three-dimensional components in simulated flight.

And S3, dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path, and collecting inspection video data in real time.

S4, extracting defect characteristics of the electric wire from the inspection video data by using a texture recognition technology, and performing defect classification and defect position positioning on the defect characteristics by using a deep learning model.

It should be noted that, extracting the defect feature of the wire from the inspection video data by using the texture recognition technology, and performing defect classification and defect position positioning on the defect feature by using the deep learning model includes the following steps:

preprocessing the collected video data, including video segmentation, frame extraction, etc., segmenting the video into individual frame images, and extracting key frames.

A texture feature extraction method such as a local binary pattern (Local Binary Patterns, LBP) may be used to extract a defect feature of the electric wire from the extracted frame image by using a texture recognition technique.

And marking the extracted defect characteristics, including defect types and defect positions, and dividing marked data into a training set, a verification set and a test set.

Deep learning models suitable for defect classification and position location tasks, such as convolutional neural networks (Convolutional Neural Network, CNN) and the like, are designed, and network structures combining image classification and target location are combined.

Training the deep learning model by using a training set, taking the extracted defect characteristics as input, classifying defects and predicting the position location, and performing model training by using a cross entropy loss function and an optimization algorithm.

And evaluating the trained model by using the verification set, and calculating indexes such as classification accuracy, position location accuracy and the like to evaluate the performance of the model.

And performing defect classification and position positioning on the defect characteristics in the test set by using the trained deep learning model, and identifying and positioning the defects of the electric wire according to the classification result and the positioning result output by the model.

S5, respectively dispatching unmanned aerial vehicle groups of corresponding types based on the defect characteristics of different types to treat the cable defects.

The method for processing the cable defects by dispatching the unmanned aerial vehicle groups of corresponding types based on the defect characteristics of different types comprises the following steps:

s51, obtaining defect characteristics of different types, wherein the defect characteristics at least comprise surface cracks, surface abrasion, floaters on the surface of the electric wire and wire strands.

S52, constructing a collaborative game model, taking the unmanned aerial vehicle as a game participant, and synchronously executing the flight direction and the flight step length of the self task decision of each game participant according to the step length constraint condition.

It should be noted that, the collaborative game model is a model in a game theory, and is used for describing a situation that a plurality of participants realize common interests through cooperation, in the collaborative game model, the participants reach consensus through negotiation, cooperation and coordination, and jointly formulate a strategy to realize an optimal result.

The method comprises the following steps of constructing a collaborative game model, taking an unmanned aerial vehicle as a game participant, and synchronously executing the flight direction and the flight step length of self task decision of each game participant according to step length constraint conditions:

s521, acquiring state variables and control variables of the unmanned aerial vehicle group at preset time, and constructing a cooperative game model of the unmanned aerial vehicle group.

S522, randomly generating initial positions of the unmanned aerial vehicle units, and setting initial step sizes of the unmanned aerial vehicle units by combining the sensing radius and the communication radius of the unmanned aerial vehicle units.

S523, detecting all flight environment information of each unmanned aerial vehicle in a sensing range at the current position, and carrying out local information exchange with the adjacent unmanned aerial vehicle in a communication range.

S524, each unmanned aerial vehicle unit decides an optimal flight direction with the maximum profit function at the current moment according to the acquired local information.

The expression of the benefit function is:

in the method, in the process of the invention,respectively indicate that the predicted current moment i is at the selection theta _-i (k) The distance sum and the distance standard deviation between the next moment and all adjacent unmanned aerial vehicles;

n represents the number of adjacent unmanned aerial vehicles;

θ (k) and θ _-i (k) Respectively representing the flight direction and the optimal flight direction;

i represents the number of game participants;

j represents the j-th adjacent unmanned aerial vehicle;

θ represents a decision variable;

k represents the kth time;

u represents an effect function value;

f represents a benefit function value;

d represents the distance.

And S525, determining the maximum feasible step length at the current moment according to the determined optimal flight direction at the current moment, and taking the maximum feasible step length as the movement step length at the current moment.

The method comprises the following steps of determining the maximum feasible step length of the current moment according to the determined optimal flight direction of the current moment, and taking the maximum feasible step length as the movement step length of the current moment:

s5251, when the number of adjacent unmanned aerial vehicles is larger than a preset threshold, taking the maximum step length as the current movement step length.

And S5252, calculating the maximum feasible step length capable of keeping all the adjacent unmanned aerial vehicles according to the flight distance between the adjacent unmanned aerial vehicles and the optimal flight direction at the current moment when the number of the adjacent unmanned aerial vehicles is smaller than a preset threshold value, and taking the maximum feasible step length as the movement step length at the current moment.

S526, each unmanned aerial vehicle unit simultaneously executes the optimal flight direction and the motion step length which are independently determined, and moves to a new position, and finally returns to the step S522.

And S53, combining the cooperative game model with a task priority scheduling algorithm according to the importance and the emergency degree of the defect characteristics.

It should be noted that, according to the importance and the emergency degree of the defect feature, the combination of the collaborative game model and the task priority scheduling algorithm includes the following steps:

according to the importance and the emergency degree of the defect characteristics, corresponding priorities are assigned to each task, and the priorities of the tasks can be determined according to factors such as defect types, influence ranges, processing time limits and the like.

And constructing a collaborative game model according to the relationship between the participants and the tasks, wherein the collaborative game model comprises a strategy space and a utility function of the participants, and the utility function can consider factors such as the priority of the tasks, the resource investment of the participants, the collaborative relationship and the like.

And solving the optimal task strategy of each participant by using a solving method in the game theory, such as Nash equilibrium, optimal response and the like, wherein in the solving process, the priority of the task can be considered as a constraint condition, so that the task with high priority is ensured to be processed preferentially.

Tasks are scheduled according to an optimal task strategy and task priorities using task priority scheduling algorithms including shortest job priority (SJF), highest Priority (HPF), round robin scheduling, etc.

And executing tasks in sequence according to the priority according to the result of the task priority scheduling algorithm, and simultaneously, guiding the cooperative behavior among the participants by utilizing the cooperative game model to ensure the efficient execution of the tasks and optimize the cooperative effect.

And S54, solving the cooperative game model, determining an optimal task strategy of each game participant, and dispatching the unmanned aerial vehicle group of the corresponding type to the corresponding task for defect processing.

And S55, each unmanned aerial vehicle unit executes processing tasks of corresponding types of defect features according to the dispatching instructions, wherein the processing tasks at least comprise cleaning tasks by using the cleaning unmanned aerial vehicle unit, inspection tasks by using the inspection unmanned aerial vehicle unit, and cruising processing tasks by using the cruising unmanned aerial vehicle unit.

It should be noted that, the inspection unmanned aerial vehicle unit is utilized to carry out the inspection task, the inspection unmanned aerial vehicle unit is equipped with devices such as a high-definition camera and an infrared sensor, and can carry out all-round inspection and inspection on a high-voltage power grid, and according to task requirements, the inspection unmanned aerial vehicle unit can fly to a defect area, acquire images and data in real time, and monitor, detect and record defects.

It is to be noted that, utilize cleaning unmanned aerial vehicle to clean the task, cleaning unmanned aerial vehicle carries on cleaning device and sensor usually, can clean impurity such as floater, the filth that the electric wire surface exists, cleans unmanned aerial vehicle and can pinpoint and clean the defect area, and the clearance pollutant needs to guarantee the security in cleaning process, avoids causing the damage to the electric wire.

It should be noted that, the unmanned aerial vehicle is used to perform the cruising task, and the unmanned aerial vehicle generally has longer flight time and remote control capability, and is suitable for long-time and large-scale task execution, and the unmanned aerial vehicle can perform the cruising task according to the dispatching instruction.

S56, information sharing and resource coordination among the unmanned aerial vehicle units are achieved by utilizing a wireless communication technology, and the positions and defect processing progress of the unmanned aerial vehicle units are monitored in real time.

And S6, recording and reporting the processing result, and making an overhaul plan based on the health condition of the high-voltage power grid.

It should be noted that, recording and reporting the processing result, and making an overhaul plan based on the health condition of the high-voltage power grid includes the following steps:

and recording flight data and identification results of the inspection unmanned aerial vehicle, wherein the flight data and identification results comprise information such as flight paths, inspection time, inspection areas, identified defect characteristics and the like, and the records can comprise text descriptions, photos, videos and the like so as to facilitate subsequent analysis and reporting.

And evaluating the defect characteristics identified by the video inspection according to the recorded data and actual conditions, analyzing indexes such as types, quantity, severity and the like of the defect characteristics, and judging the health condition and maintenance requirement of the power grid.

And (3) according to the health condition evaluation and the maintenance requirements, a corresponding maintenance plan is formulated, and the maintenance plan comprises the information of the repair content, the time schedule, the resource requirements and the like of the defect characteristics.

Determining specific maintenance tasks and requirements according to the maintenance plan; according to the nature and the requirement of the maintenance task, proper maintenance personnel are allocated to ensure that the maintenance personnel have relevant skills and qualification; determining the specific working time and place of a maintenance task, and communicating and arranging with maintenance personnel; according to the requirement of maintenance tasks, necessary tools, equipment and protection equipment are provided, so that maintenance personnel can safely and efficiently carry out maintenance work.

As shown in fig. 2, according to another embodiment of the present invention, there is further provided an unmanned aerial vehicle high-voltage network video inspection recognition system based on deep learning, where the unmanned aerial vehicle high-voltage network video inspection recognition system includes a component position labeling module 1, a simulated flight module 2, an inspection video acquisition module 3, a defect recognition module 4, a defect processing module 5, and an inspection plan making module 6;

The component position labeling module 1 is used for acquiring video data of the high-voltage power grid in real time, extracting outline features from the video data and labeling the distribution positions of target components in the high-voltage power grid.

And the simulated flight module 2 is used for setting an initial flight path in combination with the distribution position of the target member and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path.

And the inspection video acquisition module 3 is used for dispatching the inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path and acquiring inspection video data in real time.

And the defect recognition module 4 is used for extracting defect characteristics of the electric wire from the inspection video data by utilizing a texture recognition technology, and carrying out defect classification and defect position positioning on the defect characteristics by utilizing a deep learning model.

And the defect processing module 5 is used for respectively dispatching unmanned aerial vehicle groups of corresponding types to process the cable defects based on the defect characteristics of different types.

And the overhaul plan making module 6 is used for recording and reporting the processing result and making an overhaul plan based on the health condition of the high-voltage power grid.

In summary, by means of the technical scheme, the position of the electric wire can be accurately identified by extracting the contour points and detecting and dividing the electric wire dense area, so that the accuracy and precision of electric wire detection are improved, the position and boundary of the electric wire can be positioned, more accurate electric power line information can be provided for unmanned aerial vehicle inspection, meanwhile, the area of the electric wire can be clearly displayed, further, the existence of stranded wire characteristics can be identified and judged, and the potential line fault risk is reduced; according to the invention, the depth information of each wire in the wire dense area can be deduced according to the parallax information in the image through the vision principle recovery algorithm, so that the three-dimensional position and distance information of the wires are obtained, the spatial distribution of the wires can be more comprehensively understood, the wires can be accurately positioned, meanwhile, the topological relation between the wires can be known through deducing the front-rear connection relation of the wires, the structure and the characteristics of the power lines can be more comprehensively analyzed, and more accurate power line information can be further provided for unmanned aerial vehicle inspection; according to the invention, by combining the distribution positions of the target components, the optimal initial flight path can be set and corrected, so that the navigation distance and time of the unmanned aerial vehicle are reduced to the greatest extent, the actual inspection efficiency and coverage rate are improved, the situation that the unmanned aerial vehicle repeatedly covers or does not fly in the actual inspection process is avoided, the risk that the unmanned aerial vehicle collides with other obstacles in the flight process is avoided, and the flight safety is improved; according to the invention, the deep learning model is utilized to classify the defect characteristics, and the cooperative game model is combined with the task priority scheduling algorithm, so that the unmanned aerial vehicle unit of the corresponding type can be dispatched to the corresponding defect processing task according to the defect characteristics, thereby improving the defect processing efficiency, shortening the fault repairing period, simultaneously helping to reasonably allocate the resources of the unmanned aerial vehicle unit, allocating the defect processing task with higher priority to the unmanned aerial vehicle unit with higher experience and expertise, further helping to ensure the safety of defect processing, and reducing the potential risk and possibility of accidents.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.

Claims (10)

1. The unmanned aerial vehicle high-voltage power network video inspection and identification method based on deep learning is characterized by comprising the following steps of:

s1, acquiring video data of a high-voltage power grid in real time, extracting outline features from the video data, and marking distribution positions of target components in the high-voltage power grid;

s2, setting an initial flight path by combining the distribution positions of the target components, and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path;

s3, dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path, and collecting inspection video data in real time;

s4, extracting defect characteristics of the electric wire from the inspection video data by using a texture recognition technology, and performing defect classification and defect position positioning on the defect characteristics by using a deep learning model;

s5, respectively dispatching unmanned aerial vehicle groups of corresponding types based on the defect characteristics of different types to treat the cable defects;

And S6, recording and reporting the processing result, and making an overhaul plan based on the health condition of the high-voltage power grid.

2. The unmanned aerial vehicle high-voltage power network video inspection recognition method based on deep learning according to claim 1, wherein the steps of acquiring video data of the high-voltage power network in real time, extracting outline features from the video data, and marking the distribution positions of target components in the high-voltage power network comprise the following steps:

s11, acquiring video data of a high-voltage power grid, and preprocessing the video data;

s12, extracting an image sequence from the preprocessed video data, extracting a contour line of the image sequence and matching the contour line, and identifying a target component of the high-voltage power grid, wherein the target component at least comprises an electric wire, a pole tower and an insulator;

s13, extracting contour points from the matched contour lines, and detecting and dividing an electric wire dense region in the image sequence;

s14, acquiring depth information of each wire in the wire dense area by utilizing a vision principle recovery algorithm, deducing a front-back connection relation of the wires, and describing the actual trend of the wires;

s15, marking the distribution positions of target components in the high-voltage power grid by respectively combining the actual trend of the electric wires.

3. The unmanned aerial vehicle high-voltage power network video inspection recognition method based on deep learning according to claim 2, wherein the steps of extracting contour points from the matched contour lines, and detecting and dividing the wire-dense areas in the image sequence comprise the following steps:

s131, extracting contour points from the matched contour lines, and judging whether an electric wire dense area exists in the image sequence;

s132, acquiring a chain code value of a boundary point of an electric wire dense region in an image sequence, and carrying out boundary tracking on the electric wire dense region in a clockwise direction;

s133, analyzing the chain code value to judge a chain code pair of the boundary point of the electric wire dense region, and judging whether the boundary point of the electric wire dense region is a characteristic point or not according to the chain code pair;

s134, detecting characteristic points and minimum circumscribed rectangles in the wire dense area, and detecting dense wires;

s135, performing bump detection on the dense electric wires to obtain the concave span and depth of the concave area, and if the preset condition is met, taking the concave area as the concave area of the dense electric wires and marking the direction of the concave area;

s136, calculating the distance from the feature point in the concave area to the top points on two sides of the convex point, and taking the shortest distance as the distance from the current feature point to the top points of the convex point;

S137, comparing the characteristic points in the concave areas, taking the maximum characteristic point as a dividing point of the dense wire area, judging whether the concave areas meeting the preset conditions exist, if so, jumping to the step S136, otherwise, continuing to execute the step S138;

s138, connecting the division points with different directions, analyzing whether the divided wire dense areas meet the dense wire judging principle, and judging whether stranded wire characteristics exist in the dense wires.

4. The unmanned aerial vehicle high-voltage network video inspection recognition method based on deep learning according to claim 3, wherein the method for obtaining depth information of each wire in the wire-intensive area by using a vision principle recovery algorithm and deducing a front-back connection relationship of the wires, and describing the actual trend of the wires comprises the following steps:

s141, dividing the contour lines in the wire dense area into segments with equal lengths, and extracting the maximum tangent point of the current segment from each segment as a key point;

s142, establishing a corresponding relation between the current image and the key frame by utilizing homography transformation, and selecting a key point in the current image;

s143, drawing a polar line corresponding to the selected key point in the adjacent image, and taking the key point with the minimum distance with the polar line as a corresponding point in the adjacent image;

S144, calculating depth information of a corresponding point on the contour line by using a visual principle recovery algorithm, and using an average value of horizontal and vertical interpolation results as a depth value of pixels in the contour line;

s145, deducing the front-back connection relation of the electric wires through adjacent depth values, and describing the actual trend of the electric wires according to the front-back connection relation.

5. The deep learning-based unmanned aerial vehicle high-voltage power network video inspection recognition method according to claim 4, wherein setting an initial flight path in combination with the distribution position of the target member, and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path comprises the following steps:

s21, constructing a three-dimensional digital model based on a high-voltage power grid, and modeling each component of the high-voltage power grid in the three-dimensional digital model to obtain a three-dimensional component body;

s22, setting an initial flight path based on the distribution position of the three-dimensional component body in the three-dimensional digital model, and carrying out simulated flight on the virtual unmanned aerial vehicle according to the initial flight path;

s23, adjusting control instructions of the virtual aircrafts in real time according to flight positions, flight speeds and flight postures of the virtual aircrafts in simulated flight;

S24, acquiring environment information, obstacle positions and height changes of the virtual unmanned aerial vehicle in simulated flight, and correcting an initial flight path by combining an environment sensing algorithm;

s25, integrating the corrected flight path and the distribution positions of the three-dimensional components to ensure that the virtual unmanned aerial vehicle can cover all the three-dimensional components in simulated flight.

6. The deep learning-based unmanned aerial vehicle high-voltage power network video inspection recognition method according to claim 5, wherein the step of respectively dispatching unmanned aerial vehicle groups of corresponding types to treat cable defects based on different types of defect features comprises the following steps:

s51, obtaining defect characteristics of different types, wherein the defect characteristics at least comprise surface cracks, surface abrasion, floaters on the surface of the electric wire and wire strands;

s52, constructing a collaborative game model, taking an unmanned aerial vehicle as a game participant, and synchronously executing the flight direction and the flight step length of the task decision of each game participant according to the step length constraint condition;

s53, combining the cooperative game model with a task priority scheduling algorithm according to the importance and the emergency degree of the defect characteristics;

s54, solving the cooperative game model, determining an optimal task strategy of each game participant, and dispatching an unmanned aerial vehicle group of a corresponding type to a corresponding task for defect processing;

S55, each unmanned aerial vehicle unit executes processing tasks of corresponding types of defect features according to the dispatching instructions, wherein the processing tasks at least comprise cleaning tasks by using the cleaning unmanned aerial vehicle unit, inspection tasks by using the inspection unmanned aerial vehicle unit, and cruising processing tasks by using the cruising unmanned aerial vehicle unit;

s56, information sharing and resource coordination among the unmanned aerial vehicle units are achieved by utilizing a wireless communication technology, and the positions and defect processing progress of the unmanned aerial vehicle units are monitored in real time.

7. The deep learning-based unmanned aerial vehicle high-voltage network video tour inspection identification method according to claim 6, wherein the steps of constructing a collaborative game model, taking an unmanned aerial vehicle group as a game participant, and synchronously executing the flight direction and the flight step length of self task decision of each game participant according to step length constraint conditions comprise the following steps:

s521, acquiring a state variable and a control variable of the unmanned aerial vehicle unit at a preset moment, and constructing a cooperative game model of the unmanned aerial vehicle unit;

s522, randomly generating initial positions of the unmanned aerial vehicle units, and setting initial step sizes of the unmanned aerial vehicle units by combining the perceived radius and the communication radius of the unmanned aerial vehicle units;

s523, detecting all flight environment information of each unmanned aerial vehicle in a sensing range at the current position, and carrying out local information exchange with the adjacent unmanned aerial vehicle in a communication range;

S524, each unmanned aerial vehicle unit decides an optimal flight direction with the maximum profit function at the current moment according to the acquired local information;

s525, determining the maximum feasible step length of the current moment according to the determined optimal flight direction of the current moment, and taking the maximum feasible step length as the movement step length of the current moment;

s526, each unmanned aerial vehicle unit simultaneously executes the optimal flight direction and the motion step length which are independently determined, and moves to a new position, and finally returns to the step S522.

8. The deep learning-based unmanned aerial vehicle high-voltage power network video inspection recognition method according to claim 7, wherein the determining the maximum feasible step length of the current moment according to the determined optimal flight direction of the current moment and taking the maximum feasible step length as the motion step length of the current moment comprises the following steps:

s5251, when the number of adjacent unmanned aerial vehicles is greater than a preset threshold, taking the maximum step length as the current movement step length;

and S5252, calculating the maximum feasible step length capable of keeping all the adjacent unmanned aerial vehicles according to the flight distance between the adjacent unmanned aerial vehicles and the optimal flight direction at the current moment when the number of the adjacent unmanned aerial vehicles is smaller than a preset threshold value, and taking the maximum feasible step length as the movement step length at the current moment.

9. The deep learning-based unmanned aerial vehicle high-voltage power network video inspection recognition method according to claim 8, wherein the expression of the profit function is:

in the method, in the process of the invention,respectively indicate that the predicted current moment i is at the selection theta _-i (k) The distance sum and the distance standard deviation between the next moment and all adjacent unmanned aerial vehicles;

n represents the number of adjacent unmanned aerial vehicles;

θ (k) and θ _-i (k) Respectively representing the flight direction and the optimal flight direction;

i represents the number of game participants;

j represents the j-th adjacent unmanned aerial vehicle;

θ represents a decision variable;

k represents the kth time;

u represents an effect function value;

f represents a benefit function value;

d represents the distance.

10. The unmanned aerial vehicle high-voltage network video inspection recognition system based on deep learning is used for realizing the unmanned aerial vehicle high-voltage network video inspection recognition method based on deep learning as claimed in any one of claims 1-9, and is characterized in that the unmanned aerial vehicle high-voltage network video inspection recognition system comprises a component position marking module, a simulated flight module, an inspection video acquisition module, a defect recognition module, a defect processing module and an inspection plan making module;

the component position labeling module is used for acquiring video data of the high-voltage power grid in real time, extracting outline features from the video data and labeling the distribution positions of target components in the high-voltage power grid;

The simulated flight module is used for setting an initial flight path in combination with the distribution position of the target component and controlling the virtual unmanned aerial vehicle to perform simulated flight according to the initial flight path;

the inspection video acquisition module is used for dispatching an inspection unmanned aerial vehicle to automatically inspect the high-voltage power grid according to the corrected flight path and acquiring inspection video data in real time;

the defect recognition module is used for extracting defect characteristics of the electric wire from the inspection video data by utilizing a texture recognition technology, and carrying out defect classification and defect position positioning on the defect characteristics by utilizing a deep learning model;

the defect processing module is used for respectively dispatching unmanned aerial vehicle groups of corresponding types to process the cable defects based on the defect characteristics of different types;

and the overhaul plan making module is used for recording and reporting the processing result and making an overhaul plan based on the health condition of the high-voltage power grid.