A method for inspecting power grid faults

Abstract

The present invention provides a power grid fault inspection method, including: controlling an image acquisition device to inspect a target area according to a first preset route to obtain a first optical image during the inspection process; using a preset prior knowledge base to perform fault discrimination on an object to be detected in the first optical image to obtain a first fault identification result, wherein the preset prior knowledge base includes a Gaussian mixture model of each object to be detected; controlling an image acquisition device to inspect a target area according to a second preset route to obtain a second optical image during the inspection process; fusing the first optical image and the second optical image to obtain a fused image, using a preset prior knowledge base to perform fault discrimination on the object to be detected in the fused image to obtain a second fault identification result; obtaining a fault identification result based on the first fault identification result and the second fault identification result. Through the present invention, an prior knowledge base is established through a small number of samples, and an image to be detected with a fault can be accurately identified.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault detection, and in particular to a power grid fault inspection method. Background Art

[0002] Currently, a large number of power transmission networks are exposed to the natural environment and are extremely vulnerable to extreme weather conditions, such as typhoons and earthquakes. Post-disaster manual repair of power grid failures is constrained by extreme geographical environments and long-distance transmission networks, which requires huge manpower and time costs, and even the lives of repair personnel. Although deep learning object detection technology has made major breakthroughs in the computer field, the quality of deep learning models depends largely on the quantity and quality of training data. The data set generated in the process of post-disaster repair of power transmission lines is limited. Therefore, it is still impossible to accurately identify faults through deep learning. Summary of the invention

[0003] Therefore, the technical problem to be solved by the present invention is to overcome the defect that the prior art cannot accurately identify the fault, thereby providing a power grid fault inspection method.

[0004] A first aspect of the present invention provides a power grid fault inspection method, comprising: controlling an image acquisition device to inspect a target area along a first preset route, obtaining a first optical image during the inspection process, wherein the first optical image includes an object to be detected; using a preset priori knowledge base to perform fault discrimination on the object to be detected in the first optical image, and obtaining a first fault identification result, wherein the preset priori knowledge base includes a Gaussian mixture model of each object to be detected; controlling an image acquisition device to inspect a target area along a second preset route, obtaining a second optical image during the inspection process, wherein the first preset route and the second preset route are in opposite directions, and when the first optical image and the second optical image are obtained, the horizontal inclination angle of the image acquisition device is different; fusing the first optical image and the second optical image to obtain a fused image, using a preset priori knowledge base to perform fault discrimination on the object to be detected in the fused image, and obtaining a second fault identification result; obtaining a fault identification result based on the first fault identification result and the second fault identification result.

[0005] Optionally, in the power grid fault inspection method provided by the present invention, the first preset route is determined by the following steps: obtaining a height map of the target area, performing three-dimensional modeling of the target area based on the height map, and forming a virtual environment of the target area; determining the transmission network topology structure in the target area according to the virtual environment, and determining an initial route according to the transmission network topology structure; and optimizing the initial route using reinforcement learning to obtain the first preset route.

[0006] Optionally, in the power grid fault inspection method provided by the present invention, a preset prior knowledge base is used to perform fault identification on the object to be detected in the first optical image to obtain a first fault identification result, including: preprocessing the first optical image to obtain a denoised line feature space image; parsing the denoised line feature space image to obtain at least one candidate frame, which contains the object to be detected; using a preset prior knowledge base to perform fault identification on the object to be detected in the candidate frame to obtain a first fault identification result.

[0007] Optionally, in the power grid fault inspection method provided by the present invention, the first optical image is preprocessed to obtain a denoised line feature space image, including: mapping pixel points in the first optical image to the line feature space to obtain a connected edge contour of each object to be detected; based on the connected edge contour, a preset image filtering algorithm is used to remove the environmental background noise in the first optical image to obtain a denoised line feature space image.

[0008] Optionally, in the power grid fault inspection method provided by the present invention, the step of constructing a preset prior knowledge base includes: obtaining known images containing objects to be detected that are captured at multiple angles; preprocessing the known images to obtain denoised feature space images of the known images; modeling the denoised feature space images of the known images using Gaussian mixture models of plane views at different angles to obtain a first Gaussian mixture model for each angle; and performing secondary modeling on the Gaussian mixture models of plane views at each angle to obtain a second Gaussian mixture model for each object to be detected.

[0009] Optionally, in the power grid fault inspection method provided by the present invention, the preset prior knowledge base also includes a rule model of the object to be detected, and the step of constructing the preset prior knowledge base also includes: obtaining a known image containing the object to be detected; preprocessing the known image to obtain a denoised feature space image of the known image; and summarizing the denoised feature space image of the known image to obtain a rule model of the object to be detected.

[0010] Optionally, in the power grid fault inspection method provided by the present invention, the denoised line feature space image is analyzed to obtain at least one candidate frame, including: determining an area in the denoised line feature space image where the pixel density is greater than a preset value as a candidate frame, and / or fitting each pixel point in the denoised line feature space image, and determining the circumscribed rectangle of the fitted curve as a candidate frame.

[0011] Optionally, the power grid fault inspection method provided by the present invention further includes determining the geographic coordinates of the image to be inspected in each candidate frame according to GPS information.

[0012] The second aspect of the present invention provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, thereby executing the power grid fault inspection method provided in the first aspect of the present invention.

[0013] A third aspect of the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the power grid fault inspection method provided by the first aspect of the present invention.

[0014] The technical solution of the present invention has the following advantages:

[0015] The power grid fault inspection method provided by the present invention adopts a preset priori knowledge base to perform fault discrimination on the objects to be detected in the first optical image and the second optical image. The preset priori knowledge base includes a Gaussian mixture model of each detection object. Since the dependence on samples is low when the Gaussian mixture model is established, after the priori knowledge base is established by a small amount of samples in the early stage, the image to be detected with faults can be accurately identified. In addition, in the method provided by the present invention, the target area is inspected according to the first preset route and the second preset route to obtain the first optical image and the second optical image respectively, the first optical image is used to obtain a first fault identification result, and the second optical image is used for secondary verification to obtain a second fault identification result, and finally the fault identification result is obtained. The directions of obtaining the first optical image and the second optical image are different, and the horizontal inclination angles are different. Therefore, the fault identification result obtained by combining the first optical image and the second optical image is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a flowchart of a specific example of a power grid fault inspection method according to an embodiment of the present invention;

[0018] Figure 2 is a principle block diagram of a specific example of a power grid fault inspection device in an embodiment of the present invention;

[0019] Figure 3 It is a principle block diagram of a specific example of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] In the description of the present invention, it should be noted that the terms “first” and “second” are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0022] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0023] The embodiment of the present invention provides a method for inspecting power grid faults. Figure 1 As shown, including:

[0024] Step S11: controlling the image acquisition device to inspect the target area along a first preset route, and acquiring a first optical image during the inspection process, wherein the first optical image contains the object to be detected.

[0025] In an optional embodiment, the target area includes a plurality of objects to be detected that need to be detected to see if a fault occurs, and the images to be detected include but are not limited to pole towers, transmission lines, and the like.

[0026] In an optional embodiment, the image acquisition device may be an aircraft with image acquisition function, such as a drone. Since the method provided in the embodiment of the present invention is used for the detection of power grid faults, the coverage range of the objects to be detected is wide and the target area is large, the drone can quickly acquire images of each object to be detected in the target area.

[0027] In an optional embodiment, the first optical image may be a video image, or may be a group of optical images taken continuously.

[0028] In an optional embodiment, the first preset route can be pre-set before the image acquisition device starts acquiring, or it can be generated by the image acquisition device during the acquisition process based on the distribution of each object to be detected in the target area, or it can be obtained by adjusting the pre-set path according to the distribution of each object to be detected during the image acquisition process.

[0029] In an optional embodiment, when the image acquisition device inspects the target area according to the first preset route, it maintains a certain distance and angle with the object to be inspected to acquire images.

[0030] Step S12: using a preset priori knowledge base to perform fault discrimination on the object to be detected in the first optical image to obtain a first fault identification result, wherein the preset priori knowledge base includes a Gaussian mixture model of each object to be detected.

[0031] The Gaussian mixture model is a generative model that assumes that the samples follow a Gaussian mixture distribution. When building a Gaussian model for a tower, it is assumed that there are k different types of towers in the sample, that is, the model target is k clusters. Its parameters are only k μ and σ. Here, only the Gaussian mixture model is used, and its idea is applied to make the assumption of a Gaussian mixture distribution on the data in the sample, thereby reducing the data dependence of modeling a small number of tower samples.

[0032] In an optional embodiment, if the first fault identification result determines that a fault exists in the object to be detected, the first fault identification result also includes the fault type and confidence level of the object to be detected.

[0033] In an optional embodiment, when determining a fault, if it is detected that the object to be detected has a fault, the method provided by the embodiment of the present invention further includes determining the world coordinates of the object to be detected with the fault in combination with GPS information.

[0034] Step S13: Control the image acquisition device to inspect the target area along the second preset route to obtain a second optical image during the inspection process. The first preset route is in opposite directions to the second preset route. When acquiring the first optical image and the second optical image, the horizontal inclination angle of the image acquisition device is different.

[0035] In an optional embodiment, the first optical image is a plan view, and the second optical image is a 45° side view.

[0036] In an optional embodiment, the image acquisition device acquires the first optical image and the second optical image at the same frame rate.

[0037] In an optional embodiment, the first preset route and the second preset route have the same path but are in opposite directions.

[0038] In an optional embodiment, the second optical image may be a video image, or may be a group of optical images taken continuously.

[0039] In an optional embodiment, the image acquisition device activates a route memory function while acquiring the first optical image, and records the first preset route actually traversed during the acquisition of the first optical image. The image acquisition device acquires the second optical image while returning along the original route of the first preset route, thereby ensuring that the objects to be detected contained in the first optical image and the second optical image are the same, and that the positions of the same image to be detected in the first optical image and the second optical image are the same, so that the subsequent fused image obtained by fusing the first optical image and the second optical image is more efficient in fault identification and the identification result is more accurate.

[0040] Step S14: Fusing the first optical image and the second optical image to obtain a fused image, and using a preset priori knowledge base to perform fault identification on the object to be detected in the fused image to obtain a second fault identification result.

[0041] In an optional embodiment, after the first optical image and the second optical image are fused, a plan view and a 45° side view of the same image to be detected are obtained.

[0042] In an optional embodiment, the Gaussian mixture model used when executing step S12 is a plan view Gaussian mixture model, and the Gaussian mixture model used when executing step S14 is obtained by secondary modeling based on the plan view Gaussian mixture model.

[0043] In the embodiment of the present invention, the angles for acquiring the first optical image and the second optical image are different, so that the second fault identification results obtained according to the first optical image and the second optical image are integrated to achieve complementarity, thereby obtaining a more accurate fault identification result.

[0044] Step S15: Obtain a fault identification result according to the first fault identification result and the second fault identification result.

[0045] In an optional embodiment, the method for determining the fault result can be that when any one of the first fault identification result and the second fault identification result determines that the object to be detected has a fault, it is determined that the object to be detected has a fault; when both the first fault identification result and the second fault identification result determine that the object to be detected does not have a fault, it is determined that the object to be detected does not have a fault.

[0046] The power grid fault inspection method provided by the embodiment of the present invention adopts a preset priori knowledge base to perform fault discrimination on the objects to be detected in the first optical image and the second optical image. The preset priori knowledge base includes a Gaussian mixture model of each detection object. Since the dependence on samples is low when establishing the Gaussian mixture model, after the priori knowledge base is established by a small number of samples in the early stage, the image to be detected with faults can be accurately identified. In addition, in the method provided by the embodiment of the present invention, the target area is inspected according to the first preset route and the second preset route to obtain the first optical image and the second optical image, respectively, the first optical image is used to obtain a first fault identification result, and the second optical image is used for secondary verification to obtain a second fault identification result, and finally the fault identification result is obtained. The directions of obtaining the first optical image and the second optical image are different, and the horizontal inclination angles are different. Therefore, the fault identification result obtained by combining the first optical image and the second optical image is more accurate.

[0047] In an optional embodiment, the first preset route is determined by the following steps:

[0048] First, a height map of the target area is obtained, and three-dimensional modeling of the target area is performed based on the height map to form a virtual environment of the target area.

[0049] In an optional embodiment, a satellite image may be used to obtain a height map of the target area.

[0050] In an optional embodiment, the target area is three-dimensionally modeled through HeightMap, which is a grayscale map with grayscale values ​​of 0-255. The grayscale value of any image point reflects the height of the point in the original terrain.

[0051] Then, the transmission network topology in the target area is determined according to the virtual environment, and the initial route is determined according to the transmission network topology.

[0052] In an optional embodiment, when determining the topology of the power transmission network, the points in the topology of the power transmission network are first determined. Since the height of the main grid tower base is about 80-120 meters, there is a significant difference in the grayscale map with the surrounding landforms. The specific manifestation may be an abnormal point in a continuous value image on the HeightMap. From this conversion, it can be known that there is a height difference between the point and the surrounding area, and it is inferred that the point is the tower base, that is, the point in the graph structure. After determining the point in the graph structure, the edge of the graph structure is determined according to expert knowledge, thereby forming the topology of the power transmission network.

[0053] In an optional embodiment, the initial route determined according to the power transmission network topology structure can pass through every point in the power transmission network topology structure.

[0054] Finally, the initial route is optimized by using reinforcement learning to obtain a first preset route. When the image acquisition device acquires the first optical image according to the first preset route, it moves at a certain distance and angle from the route to be detected in the target area.

[0055] In an optional embodiment, in the above step S12, a preset prior knowledge base is used to perform fault identification on the object to be detected in the first optical image to obtain a first fault identification result, which specifically includes:

[0056] Firstly, the first optical image is preprocessed to obtain a denoised line feature space image.

[0057] In an optional embodiment, the first optical image is preprocessed, including extracting line features of the first optical image to form a connected contour, and performing background filtering on the first optical image to distinguish between foreground and background. Exemplarily, the foreground and background of the image can be distinguished by a machine learning algorithm, and the background can be set to empty.

[0058] Then, the denoised line feature space image is parsed to obtain at least one candidate frame, which contains the object to be detected.

[0059] In an optional embodiment, different methods are used to calculate candidate boxes of different types of objects to be detected.

[0060] In an optional embodiment, the method for obtaining the candidate frame is: determining the area in the denoised line feature space image whose pixel density is greater than a preset value as the candidate frame. For example, if the object to be detected is a pole tower, this method can be used to determine the candidate frame.

[0061] In an optional embodiment, the method for obtaining the candidate frame is: fitting each pixel point in the denoised line feature space image, and determining the circumscribed rectangle of the fitted curve as the candidate frame. For example, if the object to be detected is a transmission line, this method can be used to determine the candidate frame.

[0062] Finally, a preset priori knowledge base is used to perform fault identification on the object to be detected in the candidate frame to obtain a first fault identification result.

[0063] In an optional embodiment, in addition to the Gaussian mixture model, the prior knowledge base also includes a variety of other different fault discrimination models. For different objects to be detected, different fault discrimination models are used to perform fault identification. When fault identification is performed on the object to be detected, the object to be detected is first classified, and the model of the object to be detected in the prior knowledge base is judged. If the model corresponding to the object to be detected is a Gaussian mixture model, the fault is judged according to the probability of the Gaussian mixture model to obtain a first fault identification result. By way of example, for poles and towers, a mixed Gaussian model is used for fault identification; if the model corresponding to the object to be detected is a rule model, the classifier is trained according to the rule to obtain a first fault identification result. By way of example, for transmission lines, a rule model is used for fault identification.

[0064] In an optional embodiment, the above step S14 uses a preset prior knowledge base to perform fault identification on the object to be detected in the fused image, specifically including:

[0065] First, the second optical image is preprocessed to obtain a denoised line feature space image. For details, refer to the description of the preprocessing of the first optical image in the above embodiment, which will not be repeated here.

[0066] Then, the denoised line feature space image is parsed to obtain at least one candidate frame, which contains the object to be detected. For details, see the description of determining the candidate frame in the first optical image in the above embodiment, which will not be repeated here.

[0067] Finally, the preset prior knowledge base is used to perform fault identification on the object to be detected in the candidate frame to obtain a second fault identification result. For details, please refer to the description of identifying the fault of the object to be detected in the first optical image in the above embodiment, which will not be repeated here.

[0068] In an optional embodiment, the step of preprocessing the first optical image to obtain a denoised line feature space image specifically includes:

[0069] Firstly, the pixel points in the first optical image are mapped to the line feature space to obtain the connected edge contours of each object to be detected.

[0070] In an optional embodiment, after mapping the pixel points in the first optical image to the line feature space, an edge contour is obtained. To prevent the line feature pixel points from being broken or discontinuous, corresponding line segments are connected according to preset rules to form a connected edge contour.

[0071] Then, based on the connected edge contour, a preset image filtering algorithm is used to remove the environmental background noise in the optical image to obtain a denoised line feature space image.

[0072] In an optional embodiment, the step of preprocessing the second optical image to obtain a denoised line feature space image specifically includes:

[0073] First, the pixel points in the second optical image are mapped to the line feature space to obtain the connected edge contours of each object to be detected. For details, please refer to the description of obtaining the connected edge contour according to the first optical image in the above embodiment, which will not be repeated here.

[0074] Then, based on the connected edge contour, a preset image filtering algorithm is used to remove the environmental background noise in the optical image to obtain a denoised line feature space image.

[0075] In an optional embodiment, the step of constructing a preset priori knowledge base includes:

[0076] First, a known image containing the object to be detected collected at multiple angles is obtained;

[0077] Secondly, the known image is preprocessed to obtain a denoised feature space image of the known image. For details, please refer to the step of obtaining the denoised feature space image of the first optical image in the above embodiment, which will not be repeated here.

[0078] Then, the denoised feature space image of the known image is modeled with a planar Gaussian mixture model at different angles to obtain a first Gaussian mixture model at each angle. In an optional embodiment, when fault identification is performed on the object to be detected in the first optical image and the second optical image, fault identification is performed using the first Gaussian mixture model.

[0079] Finally, the Gaussian mixture models of the plane images at each angle are subjected to secondary modeling fusion to obtain a second Gaussian mixture model of each object to be detected. In an optional embodiment, when fault identification is performed based on the first optical image or the second optical image, the first Gaussian mixture model is used for fault identification, and when fault identification is performed based on a fused image after the first optical image and the second optical image are fused, the second Gaussian mixture model is used for fault identification.

[0080] In an optional embodiment, during secondary modeling, for the same object to be detected, pictures from different angles are mapped into clusters in two models, and the two clusters are regarded as the types of the object to be detected, completing the correspondence between different model clusters and clusters, and realizing fusion.

[0081] In an optional embodiment, the preset prior knowledge base also includes a rule model of the object to be detected, and the step of constructing the preset prior knowledge base also includes:

[0082] First, obtain a known image containing the object to be detected;

[0083] Then, the known image is preprocessed to obtain a denoised feature space image of the known image. For details, please refer to the step of obtaining the denoised feature space image of the first optical image in the above embodiment, which will not be repeated here.

[0084] Finally, the denoised feature space image of the known image is summarized to obtain the rule model of the object to be detected.

[0085] In an optional embodiment, a regular model can be established for the transmission line. For different transmission lines, several different natural splines are used to simulate the curvature of the transmission line, and their parameters are recorded and regarded as rules, that is, it is considered that the type of transmission line should obey the style of the spline, thereby obtaining a regular model of the transmission line.

[0086] The embodiment of the present invention provides a power grid fault inspection device, such as Figure 2 As shown, including:

[0087] The first inspection module 21 is used to control the image acquisition device to inspect the target area according to the first preset route, and obtain the first optical image during the inspection process. The first optical image contains the object to be detected. For details, please refer to the description of step S11 in the above embodiment, which will not be repeated here.

[0088] The first fault identification module 22 is used to use a preset priori knowledge base to perform fault identification on the object to be detected in the first optical image to obtain a first fault identification result. The preset priori knowledge base includes a Gaussian mixture model of each object to be detected. For details, please refer to the description of step S12 in the above embodiment, which will not be repeated here.

[0089] The second inspection module 23 is used to control the image acquisition device to inspect the target area along the second preset route, and obtain a second optical image during the inspection process. The first preset route is in opposite directions to the second preset route. When acquiring the first optical image and the second optical image, the horizontal inclination angle of the image acquisition device is different. For details, please refer to the description of step S13 in the above embodiment, which will not be repeated here.

[0090] The second fault identification module 24 is used to fuse the first optical image and the second optical image to obtain a fused image, and use a preset priori knowledge base to perform fault identification on the object to be detected in the fused image to obtain a second fault identification result. For details, please refer to the description of step S14 in the above embodiment, which will not be repeated here.

[0091] The fault determination module 25 is used to obtain a fault identification result according to the first fault identification result and the second fault identification result. For details, please refer to the description of step S15 in the above embodiment, which will not be repeated here.

[0092] An embodiment of the present invention provides a computer device, such as Figure 3 As shown, the computer device mainly includes one or more processors 31 and a memory 32. Figure 3 A processor 31 is taken as an example.

[0093] The computer device may further include: an input device 33 and an output device 34 .

[0094] The processor 31, the memory 32, the input device 33 and the output device 34 may be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0095] The processor 31 may be a central processing unit (CPU). The processor 31 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, or a combination of the above chips. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The memory 32 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the power grid fault inspection device, etc. In addition, the memory 32 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 32 may optionally include a memory remotely arranged relative to the processor 31, and these remote memories may be connected to the power grid fault inspection device via a network. The input device 33 can receive a calculation request (or other digital or character information) input by a user, and generate a key signal input related to the power grid fault inspection device. The output device 34 can include a display device such as a display screen to output the calculation result.

[0096] The embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions, and the computer storage medium stores computer executable instructions, which can execute the power grid fault inspection method in any of the above method embodiments. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0097] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A power grid fault inspection method, characterized in that: include: Controlling the image acquisition device to inspect the target area according to a first preset route, and acquiring a first optical image during the inspection process, wherein the first optical image contains the object to be detected; Using a first Gaussian mixture model in a preset priori knowledge base to perform fault discrimination on the object to be detected in the first optical image to obtain a first fault identification result, wherein the preset priori knowledge base includes a first Gaussian mixture model and a second Gaussian mixture model of each object to be detected; Controlling the image acquisition device to inspect the target area along a second preset route to acquire a second optical image during the inspection process, wherein the first preset route is in opposite directions to the second preset route, and when acquiring the first optical image and the second optical image, the horizontal inclination angle of the image acquisition device is different; The first optical image and the second optical image are fused to obtain a fused image, and a second Gaussian mixture model in a preset priori knowledge base is used to perform fault identification on the object to be detected in the fused image to obtain a second fault identification result; A fault identification result is obtained according to the first fault identification result and the second fault identification result.

2. The power grid fault inspection method according to claim 1, characterized in that: The first preset route is determined by the following steps: Acquire a height map of the target area, and perform three-dimensional modeling of the target area based on the height map to form a virtual environment of the target area; Determine a power transmission network topology structure within the target area according to the virtual environment, and determine an initial route according to the power transmission network topology structure; The initial route is optimized using reinforcement learning to obtain the first preset route.

3. The power grid fault inspection method according to claim 1, characterized in that: Using a preset prior knowledge base to perform fault identification on the object to be detected in the first optical image to obtain a first fault identification result includes: Preprocessing the first optical image to obtain a denoised line feature space image; Parsing the denoised line feature space image to obtain at least one candidate frame, wherein the candidate frame contains an object to be detected; The preset prior knowledge base is used to perform fault identification on the object to be detected in the candidate frame to obtain the first fault identification result.

4. The power grid fault inspection method according to claim 3, characterized in that: Preprocessing the first optical image to obtain a denoised line feature space image includes: Mapping the pixel points in the first optical image to a line feature space to obtain a connected edge contour of each object to be detected; Based on the connected edge contour, a preset image filtering algorithm is used to remove environmental background noise in the first optical image to obtain the denoised line feature space image.

5. The power grid fault inspection method according to claim 1, characterized in that: The steps of constructing the preset prior knowledge base include: Acquire known images containing the object to be detected collected at multiple angles; Preprocessing the known image to obtain a denoised feature space image of the known image; Modeling the denoised feature space image of the known image using a planar Gaussian mixture model at different angles to obtain a first Gaussian mixture model at each angle; The Gaussian mixture models of the plane images at each angle are subjected to secondary modeling fusion to obtain a second Gaussian mixture model of each object to be detected.

6. The power grid fault inspection method according to claim 5, characterized in that: The preset prior knowledge base also includes a rule model of the object to be detected, and the step of constructing the preset prior knowledge base also includes: Acquire a known image containing an object to be detected; Preprocessing the known image to obtain a denoised feature space image of the known image; The denoised feature space image of the known image is summarized to obtain a rule model of the object to be detected.

7. The power grid fault inspection method according to claim 3, characterized in that: The denoised line feature space image is parsed to obtain at least one candidate frame, including: Determine the region in the denoised line feature space image where the pixel density is greater than a preset value as the candidate frame, and / or, Each pixel point in the denoised line feature space image is fitted, and a circumscribed rectangle of the curve obtained by fitting is determined as the candidate frame.

8. The power grid fault inspection method according to claim 3, characterized in that: Also includes: The geographic coordinates of the image to be detected in each candidate frame are determined according to the GPS information.

9. A computer device, characterized in that: include: at least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to perform the power grid fault inspection method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the power grid fault inspection method according to any one of claims 1 to 8.

Images