Intelligent inspection method, device and equipment for photovoltaic power station

Abstract

The invention provides an intelligent inspection method, device and equipment for a photovoltaic power station, and the method comprises the steps: shooting a two-dimensional image of the photovoltaic power station, and constructing an earth rectangular coordinate system; determining a detection point position of the solar panel on the inspection path through the first unmanned aerial vehicle, and arranging an attitude correction object on the outer edge of the detection point position; and starting the second unmanned aerial vehicle to obtain an attitude correction image at the detection point position. Checking the attitude of the second unmanned aerial vehicle; if a preset condition is satisfied, controlling a second unmanned aerial vehicle to shoot the solar panel at the detection point position to obtain an original image of the solar panel, and constructing a fault identification network model of the solar panel to obtain a fault detection point; constructing an unmanned aerial vehicle rectangular coordinate system by taking the starting point of the second unmanned aerial vehicle as a coordinate origin; and the position information of the fault detection point is determined based on the inspection path in the earth rectangular coordinate system and the inspection path in the unmanned aerial vehicle rectangular coordinate system, so that the fault calculation complexity is reduced, and the calculation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic power generation technology, and in particular to an intelligent inspection method, device and equipment for photovoltaic power stations. Background Art

[0002] During the power generation process of photovoltaic power stations, since they are outdoors, they are inevitably damaged by strong winds or other things, causing the surface of the solar panels in the photovoltaic power station to be damaged, resulting in a decrease in power generation. Timely detection of problematic solar panels can improve the utilization rate of solar energy. At present, drones or other intelligent inspection methods are mostly used, but due to the large scale of photovoltaic power stations, drones convert between sensor coordinate systems, drone attitude coordinate systems, image coordinate systems and geodetic coordinate systems, and finally locate the position coordinates of the currently damaged solar panel.

[0003] However, the mutual conversion method of the above four coordinate systems is complicated and has large errors, resulting in poor accuracy in the location of the currently damaged solar panel. Summary of the invention

[0004] The present invention provides a photovoltaic power station intelligent inspection method, device and equipment, which are used to solve the defects of the prior art that the fault detection of solar panels of photovoltaic power stations is complicated and leads to poor accuracy.

[0005] In a first aspect, the present invention provides a photovoltaic power station intelligent inspection method, comprising:

[0006] Using satellite remote sensing to capture a two-dimensional image of the photovoltaic power station, and constructing a geodetic rectangular coordinate system based on the regional position of the photovoltaic power station in the two-dimensional image;

[0007] Performing an inspection task by the first drone to plan an inspection path, and determining the locations of inspection points of the solar panels on the inspection path;

[0008] Gridding the regional position of the photovoltaic power station and arranging posture correction objects at the outer edge of the detection point position;

[0009] Starting a second UAV to perform inspection along the inspection path, and controlling the second UAV to photograph the posture correction object at the detection point to obtain a posture correction image;

[0010] performing attitude verification on the second UAV according to the attitude correction image;

[0011] If the verification result meets the preset conditions, the second UAV is controlled to photograph the solar panel at the detection point to obtain an original image of the solar panel; if the verification result does not meet the preset conditions, the posture correction object is photographed again;

[0012] Building different types of image data sets based on the original images, and building a solar panel fault recognition network model, performing solar panel fault recognition at each detection point position, and obtaining a fault detection point;

[0013] Taking the starting point of the second UAV as the coordinate origin, constructing a UAV rectangular coordinate system;

[0014] Based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system, the position information of the fault detection point is determined.

[0015] According to a photovoltaic power station intelligent inspection method provided by the present invention, the posture correction object is a square panel structure, and a marking line with a width L is arranged in the square panel structure;

[0016] The square panel structure includes four vertices A, B, C, and D, and infrared distance sensors are respectively installed on the four vertices A, B, C, and D.

[0017] According to a photovoltaic power station intelligent inspection method provided by the present invention, performing attitude verification on the second UAV according to the attitude correction image includes:

[0018] Using the infrared distance sensors on the four vertices A, B, C, and D, respectively, the distances between the second UAV and the second UAV are tested, and a first-level attitude check is performed;

[0019] If the first-level posture check meets the preset conditions, the posture correction image is cropped to obtain a standard image of the same size; if the first-level posture check does not meet the preset conditions, the posture correction object is re-photographed;

[0020] Calculating the probability density of the standard image using a Gaussian probability density function, and identifying the foreground area and the background area in the standard image according to the obtained probability density;

[0021] Using local adaptive binarization to calculate the threshold of the pixel points in the foreground area, and locating the identification line area on the square panel structure in the foreground area according to the threshold;

[0022] The width value L* of the marking line area is calculated using the Euclidean distance;

[0023] The difference between the calculated width L* and the laid width L is determined to achieve secondary posture verification.

[0024] According to a photovoltaic power station intelligent inspection method provided by the present invention, the infrared distance sensors on the four vertices A, B, C, and D are used to respectively test the distance between the second drone and the second drone, and perform a first-level attitude check, including:

[0025] The distance L between vertex A and the second drone is obtained by testing A0 , the distance L between vertex B and the second UAV B0 , the distance L between vertex C and the second UAV C0 , the distance L between vertex D and the second UAV D0 ;

[0026] When the L A0 , L B0 , L C0 , L D0 When the difference between any two of them is less than or equal to the preset threshold, the first-level posture verification meets the preset conditions.

[0027] According to a photovoltaic power station intelligent inspection method provided by the present invention, after obtaining the standard images of the same size, the method further includes:

[0028] Performing local weighted regression smoothing filtering on the standard image includes:

[0029] f(x)=Ave(y i |x i ∈N k (x i ));

[0030] Where f(x) is the average value of the y-axis coordinates of all pixels near pixel x, which is the smoothing value of the standard image data; y i is the y-axis coordinate of the pixel; x i is the x-axis coordinate of the pixel; N k (x i ) is a set of pixels;

[0031] Filter the image after local weighted regression smoothing filtering, including:

[0032]

[0033] Among them, f(x, y) is the standard image after filtering, w 1 is the image transfer weight, w 2 is the atmospheric light value weight, T(x) is the image transmission rate of the imaging device, and AOD is the atmospheric light value.

[0034] According to a photovoltaic power station intelligent inspection method provided by the present invention, the threshold of the pixel points in the foreground area is calculated by using local adaptive binarization, and the identification line area on the square panel structure in the foreground area is located according to the threshold, including:

[0035] Set the background threshold of the standard image f(x, y) to T;

[0036] According to the background threshold T, the standard image f(x, y) is binarized, including:

[0037]

[0038] Among them, F(x, y) is the standard image after thresholding;

[0039] If the grayscale value of the standard image f(x, y) is greater than the background threshold T, the identification line area of ​​the square panel structure is screened out.

[0040] According to a photovoltaic power station intelligent inspection method provided by the present invention, the Euclidean distance is used to calculate the width value L* of the identification line area, including:

[0041] Calculate the Euclidean distance L(P1, P2) between two pixels:

[0042]

[0043] Among them, x 1 is the number of rows of pixel P1, y 1 is the number of columns of pixel P1; x 2 is the number of rows of pixel P2, y 2 is the number of columns of pixel P2;

[0044] Calculate the width Lp of the identification line:

[0045] Lp=L(Pixel a ,Pixel (1) )×2;

[0046] Among them, Pixel (1) Pixel is the edge pixel closest to the pixel of the identification line. a is the pixel in the identification line area;

[0047] The calculated value L* of the marking line area width is calculated according to the width Lp of the marking line:

[0048] L*=Lp·A';

[0049] Wherein, A' is the width transformation parameter.

[0050] According to a photovoltaic power station intelligent inspection method provided by the present invention, the position information of the fault detection point is determined based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system, including:

[0051] Calculating the displacement loss between the inspection path in the rectangular coordinate system of the UAV and the planned inspection path in the rectangular coordinate system of the earth;

[0052]

[0053] Among them, G(x) represents the inspection path in the rectangular coordinate system of the UAV, and H(x) represents the planned inspection path in the rectangular coordinate system of the earth;

[0054] When the inspection path G(x) in the UAV rectangular coordinate system is shifted to coincide with the planned inspection path H(x) in the earth rectangular coordinate system, we have:

[0055]

[0056] get:

[0057] G(x)'=H(x)+n;

[0058] Among them, m1 and m2 are the left and right critical points of the inspection path respectively; Loss(H(x), G(x)) is the displacement loss; G(x)' is the new coordinate of the inspection path G(x) in the UAV rectangular coordinate system displaced to the earth rectangular coordinate system; n is the displacement vector.

[0059] In a second aspect, the present invention further provides an intelligent inspection device for a photovoltaic power station, comprising:

[0060] A construction module, used to take a two-dimensional image of a photovoltaic power station using satellite remote sensing, and to construct a geodetic rectangular coordinate system based on the regional position of the photovoltaic power station in the two-dimensional image;

[0061] An inspection module, configured to execute an inspection task by the first drone to plan an inspection path, and determine the locations of inspection points of the solar panel on the inspection path;

[0062] The correction module is used to perform grid processing on the regional position of the photovoltaic power station and arrange the posture correction object at the outer edge of the detection point position; start the second drone to inspect along the inspection path, and control the second drone to shoot the posture correction object at the detection point position to obtain a posture correction image; perform posture verification on the second drone according to the posture correction image; if the verification result meets the preset conditions, control the second drone to shoot the solar panel at the detection point position to obtain the original image of the solar panel; if the verification result does not meet the preset conditions, re-shoot the posture correction object;

[0063] A network module is used to construct different types of image data sets based on the original image, and to construct a solar panel fault recognition network model, to perform solar panel fault recognition at each detection point position, and to obtain a fault detection point;

[0064] The determination module is used to construct a drone rectangular coordinate system with the starting point of the second drone as the coordinate origin; based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system, determine the location information of the fault detection point.

[0065] In a third aspect, the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements any one of the above-mentioned intelligent inspection methods for photovoltaic power stations.

[0066] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described intelligent inspection methods for photovoltaic power stations.

[0067] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the intelligent inspection method for a photovoltaic power station as described above is implemented.

[0068] The present invention provides a photovoltaic power station intelligent inspection method, device and equipment. Since the inspection paths of the first drone and the second drone are exactly the same, the inspection paths are used as clues. In the later coordinate conversion, only the displacement vectors of the inspection paths in two different coordinate systems need to be calculated to realize the coordinate conversion from the drone rectangular coordinate system to the earth rectangular coordinate system, thereby determining the location information of the detection point with faults. The process of coordinate conversion is greatly reduced, thereby effectively reducing the calculation error, and the stability of solar panel shooting is guaranteed through attitude verification, thereby effectively improving the efficiency of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 is a flow chart of the intelligent inspection method for photovoltaic power stations provided in this embodiment;

[0071] Figure 2 is a schematic diagram of the geodetic rectangular coordinate system provided in this embodiment;

[0072] Figure 3 is a schematic diagram of the posture correction object provided in this embodiment;

[0073] Figure 4 This is a schematic diagram of the conversion between the rectangular coordinate system of the drone and the rectangular coordinate system of the earth provided in this embodiment;

[0074] Figure 5 is a schematic diagram of the structure of the intelligent inspection device for photovoltaic power stations provided in this embodiment;

[0075] Figure 6 It is a schematic diagram of the structure of the electronic device provided in this embodiment. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0077] Figure 1 It is a flow chart of the intelligent inspection method for photovoltaic power stations provided in this embodiment.

[0078] like Figure 1 As shown, the photovoltaic power station intelligent inspection method provided by the embodiment of the present invention mainly includes the following steps:

[0079] 101. Use satellite remote sensing to capture two-dimensional images of photovoltaic power stations, and construct a geodetic rectangular coordinate system based on the regional location of the photovoltaic power stations in the two-dimensional images.

[0080] In a specific implementation process, the constructed geodetic rectangular coordinate system can be constructed based on the assistance of Beidou navigation satellite, with the purpose of constructing a known coordinate system and including the entire area of ​​the photovoltaic power station in the geodetic rectangular coordinate system. Figure 2 As shown, the area where M is located is the entire area of ​​the photovoltaic power station.

[0081] 102. The first UAV is used to perform the inspection task to plan the inspection route and determine the inspection point position of the solar panel on the inspection route.

[0082] The inspection task can be controlled manually throughout the process to develop a reasonable inspection path, such as Figure 2 As shown, S is the inspection path and the detection point is Figure 2 1, 2, 3, 4, 5, 6, 7, 8, this embodiment does not limit the specific location and number of detection points, which can be determined according to the specific area conditions.

[0083] 103. The regional location of the photovoltaic power station is gridded, and posture correction objects are arranged at the outer edge of the detection point location.

[0084] like Figure 3 As shown, the posture correction object in this embodiment is a square panel structure, and a marking line with a width of L is arranged in the square panel structure; the square panel structure includes four vertices A, B, C, and D, and infrared distance sensors are respectively installed on the four vertices A, B, C, and D. The infrared distance sensor includes a transmitting part and a receiving part. During the specific test, the transmitting part transmits infrared rays, and the receiving part receives the emitted infrared rays to obtain the distance between the vertex and the drone. A marking line with a width of L is arranged in the square panel structure. When the marking line acts as a mark, it is convenient for image recognition. In addition, the most important thing is that it is used to verify the images taken by the drone to detect whether the drone posture and the shooting camera posture parameters are correct.

[0085] 104. Start the second UAV to inspect along the inspection path, and control the second UAV to photograph the posture correction object at the detection point to obtain a posture correction image.

[0086] 105. Based on the attitude correction image, perform attitude verification on the second UAV.

[0087] Use the infrared distance sensors on the four vertices A, B, C, and D to test the distance to the second drone and perform a first-level attitude check. Specifically, the distance L between vertex A and the second drone is obtained by testing. A0 , the distance L between vertex B and the second drone B0 , the distance L between vertex C and the second drone C0 , the distance L between vertex D and the second droneD0 ; When L A0 , L B0 , L C0 , L D0 When the difference between any two of them is less than or equal to the preset threshold, the first-level attitude check meets the preset conditions. The first-level check is a rough check, which is mainly used to adjust the attitude of the drone to adjust the drone to a horizontal level and directly above the target.

[0088] If the first-level posture verification meets the preset conditions, the posture correction image is cropped to obtain a standard image of the same size; if the first-level posture verification does not meet the preset conditions, the posture correction object is re-photographed.

[0089] Perform local weighted regression smoothing filtering on the standard image, including:

[0090] f(x)=Ave(y i |x i ∈N k (x i )) (1)

[0091] Where f(x) is the average value of the y-axis coordinates of all pixels near pixel x, which is the smoothing value of the standard image data; y i is the y-axis coordinate of the pixel; x i is the x-axis coordinate of the pixel; N k (x i ) is a set of pixel points.

[0092] Filter the image after local weighted regression smoothing filtering, including:

[0093]

[0094] Among them, f(x,y) is the standard image after filtering, w 1 is the image transfer weight, w 2 is the atmospheric light value weight, T(x) is the image transmission rate of the imaging device, and AOD is the atmospheric light value.

[0095] The probability density of the standard image is calculated using the Gaussian probability density function, and the foreground area and the background area in the standard image are identified based on the obtained probability density. Specifically, in image recognition, the Gaussian probability density function is used based on the double peaks and double valleys in the grayscale histogram, where one peak corresponds to the background area and the other peak corresponds to the foreground area. The foreground area here is the area where the posture correction object is located.

[0096] The threshold of the pixel points in the foreground area is calculated by local adaptive binarization, and the identification line area on the square panel structure in the foreground area is located according to the threshold. Specifically, the background threshold of the standard image f(x, y) is set to T;

[0097] According to the background threshold T, the standard image f(x,y) is binarized, including:

[0098]

[0099] Among them, F(x, y) is the standard image after thresholding.

[0100] If the gray value of the standard image f(x, y) is greater than the background threshold T, the identification line area of ​​the square panel structure is screened out. The background threshold T here can be set manually.

[0101] The width value L* of the marking line area is calculated using the Euclidean distance. Specifically, the Euclidean distance L(P1, P2) between two pixels is calculated:

[0102]

[0103] Among them, x 1 is the number of rows of pixel P1, y 1 is the number of columns of pixel P1; x 2 is the number of rows of pixel P2, y 2 is the number of columns of pixel P2;

[0104] Calculate the width Lp of the identification line:

[0105] Lp=L(Pixel a ,Pixel (1) )×2 (5)

[0106] Among them, Pixel (1) Pixel is the edge pixel closest to the pixel of the identification line. a is the pixel in the identification line area;

[0107] Calculate the calculated value L* of the identification line area width based on the identification line width Lp:

[0108] L*=Lp·A' (6)

[0109] Wherein, A' is the width transformation parameter.

[0110] Determine the difference between the calculated width L* of the mark and the laid width L to achieve the secondary posture verification. Specifically, when the difference between the calculated width L* of the mark line area and the laid width L is less than or equal to the threshold, it is determined that the verification result meets the preset condition.

[0111] 106. If the verification result meets the preset conditions, the second UAV is controlled to photograph the solar panel at the detection point to obtain the original image of the solar panel; if the verification result does not meet the preset conditions, the attitude correction object is photographed again.

[0112] Re-photographing the posture correction object means returning to step 104 and re-executing the process until the verification result meets the preset condition.

[0113] After obtaining the original image, the original image of the solar panel is preprocessed and manually classified and labeled to obtain different types of image data sets. Specifically, the preprocessing includes randomly cropping the original image and performing data enhancement. The pixel resolution of the randomly cropped image is 256×256. The enhanced original image data is normalized and the image pixel value range is scaled from 0-255 to 0-1.

[0114] Manual classification and labeling include three categories: crack defects, pit defects and missing defects of solar panels, and three data subsets are obtained based on manual classification and labeling: crack defect image data subset, pit defect image data subset and missing defect image data subset.

[0115] 107. Different types of image data sets are constructed based on the original images, and a fault recognition network model for solar panels is constructed. The solar panel fault is recognized at each detection point to obtain the fault detection point.

[0116] The fault identification network model is preferably a VGG model, which includes three linear layers. The first layer inputs a subset of image data, the output feature is set to 256, and a nonlinear activation function ReLU is added after the output; the second layer input feature is 256, the output feature is 128, a nonlinear activation function ReLU is added after the output, and a random inactivation layer Dropout is added after the second layer of nonlinear activation function; the third linear layer input feature is 128, and the output feature is N=3; finally, it is input to the activation function softmax to output the probabilities of three categories.

[0117] 108. Use the starting point of the second drone as the coordinate origin to construct the drone rectangular coordinate system.

[0118] like Figure 4 As shown, X'O'Y' is the rectangular coordinate system of the drone. The inspection path does not change in the coordinate system X'O'Y', but the coordinate expression of the corresponding points is transformed. Therefore, based on the unchanged shape of the inspection path, as long as the vector corresponding to the coordinate transformation can be found, the coordinate system X'O'Y' can be converted to the coordinate system XOY.

[0119] 109. Based on the inspection path in the earth's rectangular coordinate system and the inspection path in the drone's rectangular coordinate system, determine the location information of the fault detection point.

[0120] Calculate the displacement loss between the inspection path in the UAV rectangular coordinate system and the planned inspection path in the earth rectangular coordinate system;

[0121]

[0122] Among them, G(x) represents the inspection path in the rectangular coordinate system of the UAV, and H(x) represents the planned inspection path in the rectangular coordinate system of the earth;

[0123] When the inspection path G(x) in the UAV rectangular coordinate system is shifted to coincide with the planned inspection path H(x) in the earth rectangular coordinate system, we have:

[0124]

[0125] get:

[0126] G(x)'= H(x)+ n (9)

[0127] Among them, m1 and m2 are the left and right critical points of the inspection path respectively; Loss(H(x), G(x)) is the displacement loss; G(x)' is the new coordinate of the inspection path G(x) in the UAV rectangular coordinate system displaced to the earth rectangular coordinate system; n is the displacement vector.

[0128] In this embodiment, the inspection paths of the first drone and the second drone are exactly the same. Based on this, the inspection paths are used as clues. In the later coordinate conversion, only the displacement vectors of the inspection paths in two different coordinate systems need to be calculated to realize the coordinate conversion from the drone rectangular coordinate system to the earth rectangular coordinate system, and then determine the location information of the detection point with faults. Compared with the traditional image coordinate system conversion to the sensor coordinate system, the sensor coordinate system conversion to the drone coordinate system, the drone coordinate system conversion to the earth coordinate system greatly reduces the amount of coordinate conversion process, and improves the conversion efficiency through clever design.

[0129] In addition, the use of the second-level attitude verification can make it possible to capture solar panel images more accurately and stably. First, a rough verification, namely the first-level attitude verification, is performed to determine whether the drone's attitude is stable and flush with the horizontal plane by detecting whether the distances between the four points on the ground and the drone are equal. After completing the first-level attitude verification, the captured image is used to compare the calculated value of the width of the identification line area in the image with the actual width to determine whether the clarity and stability of the current captured image meet the requirements. If not, the drone's attitude is further adjusted or the shooting camera parameters are adjusted until the second-level attitude verification is met at the same time, and then the formal solar panel image capture begins.

[0130] Based on the same general inventive concept, the present invention also protects an intelligent inspection device for a photovoltaic power station. The intelligent inspection device for a photovoltaic power station described below and the intelligent inspection method for a photovoltaic power station described above can refer to each other.

[0131] Figure 5 It is a schematic diagram of the structure of the intelligent inspection device for photovoltaic power stations provided in this embodiment.

[0132] like Figure 5 As shown, the photovoltaic power station intelligent inspection device provided in this embodiment includes:

[0133] A construction module 501 is used to take a two-dimensional image of a photovoltaic power station using satellite remote sensing, and construct a geodetic rectangular coordinate system based on the regional position of the photovoltaic power station in the two-dimensional image;

[0134] The inspection module 502 is used to execute the inspection task by the first drone to plan the inspection path, and determine the inspection point position of the solar panel on the inspection path;

[0135] The correction module 503 is used to perform grid processing on the regional position of the photovoltaic power station and arrange the attitude correction object on the outer edge of the detection point position; start the second drone to inspect along the inspection path, and control the second drone to shoot the attitude correction object at the detection point position to obtain an attitude correction image; perform attitude verification on the second drone according to the attitude correction image; if the verification result meets the preset conditions, control the second drone to shoot the solar panel at the detection point position to obtain the original image of the solar panel; if the verification result does not meet the preset conditions, re-shoot the attitude correction object;

[0136] The network module 504 is used to construct different types of image data sets based on the original image, and to construct a solar panel fault recognition network model, and to perform solar panel fault recognition at each detection point to obtain a fault detection point;

[0137] The determination module 505 is used to construct a UAV rectangular coordinate system with the starting point of the second UAV as the coordinate origin; and determine the location information of the fault detection point based on the inspection path in the earth rectangular coordinate system and the inspection path in the UAV rectangular coordinate system.

[0138] Figure 6 It is a schematic diagram of the structure of the electronic device provided in this embodiment.

[0139] like Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the intelligent inspection method of the photovoltaic power station.

[0140] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0141] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent inspection method of the photovoltaic power station provided by the above methods.

[0142] On the other hand, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the intelligent inspection method for photovoltaic power stations provided by the above methods.

[0143] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0144] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A photovoltaic power station intelligent inspection method, characterized in that: include: Using satellite remote sensing to capture a two-dimensional image of the photovoltaic power station, and constructing a geodetic rectangular coordinate system based on the regional position of the photovoltaic power station in the two-dimensional image; Performing an inspection task by the first drone to plan an inspection path, and determining the locations of inspection points of the solar panels on the inspection path; Gridding the regional position of the photovoltaic power station and arranging posture correction objects at the outer edge of the detection point position; Starting a second UAV to perform inspection along the inspection path, and controlling the second UAV to photograph the posture correction object at the detection point to obtain a posture correction image; performing attitude verification on the second UAV according to the attitude correction image; If the verification result meets the preset conditions, the second UAV is controlled to photograph the solar panel at the detection point to obtain an original image of the solar panel; if the verification result does not meet the preset conditions, the posture correction object is photographed again; Building different types of image data sets based on the original images, and building a solar panel fault recognition network model, performing solar panel fault recognition at each detection point position, and obtaining a fault detection point; Taking the starting point of the second UAV as the coordinate origin, constructing a UAV rectangular coordinate system; Based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system, the position information of the fault detection point is determined.

2. The intelligent inspection method for photovoltaic power stations according to claim 1, characterized in that: The posture correction object is a square panel structure, and a marking line with a width of L is arranged in the square panel structure; The square panel structure includes four vertices A, B, C, and D, and infrared distance sensors are respectively installed on the four vertices A, B, C, and D.

3. The intelligent inspection method for photovoltaic power stations according to claim 2, characterized in that: The step of performing attitude verification on the second UAV according to the attitude correction image includes: Using the infrared distance sensors on the four vertices A, B, C, and D, respectively, the distances between the second UAV and the second UAV are tested, and a first-level attitude check is performed; If the first-level posture check meets the preset conditions, the posture correction image is cropped to obtain a standard image of the same size; if the first-level posture check does not meet the preset conditions, the posture correction object is re-photographed; Calculating the probability density of the standard image using a Gaussian probability density function, and identifying the foreground area and the background area in the standard image according to the obtained probability density; Using local adaptive binarization to calculate the threshold of the pixel points in the foreground area, and locating the identification line area on the square panel structure in the foreground area according to the threshold; The width value L* of the marking line area is calculated using the Euclidean distance; The difference between the calculated width L* and the laid width L is determined to achieve secondary posture verification.

4. The intelligent inspection method for photovoltaic power stations according to claim 3, characterized in that: The infrared distance sensors on the four vertices A, B, C, and D are used to respectively test the distance between the second UAV and the second UAV, and perform a first-level attitude check, including: The distance L between vertex A and the second drone is obtained by testing A0 , the distance L between vertex B and the second UAV B0 , the distance L between vertex C and the second UAV C0 , the distance L between vertex D and the second UAV D0 ; When the L A0 , L B0 , L C0 , L D0 When the difference between any two of them is less than or equal to the preset threshold, the first-level posture verification meets the preset conditions.

5. The intelligent inspection method for photovoltaic power station according to claim 3, characterized in that: After obtaining the standard images of the same size, the method further includes: Performing local weighted regression smoothing filtering on the standard image includes: f(x)=Ave(y i |x i ∈N k (x i )); Where f(x) is the average value of the y-axis coordinates of all pixels near pixel x, which is the smoothing value of the standard image data; y i is the y-axis coordinate of the pixel; x i is the x-axis coordinate of the pixel; N k (x i ) is a set of pixel points; The image after local weighted regression smoothing filtering is filtered, including: Among them, f(x,y) is the standard image after filtering, w1 is the image transmission weight, w2 is the atmospheric light value weight, T(x) is the image transmission rate of the imaging device, and AOD is the atmospheric light value.

6. The intelligent inspection method for photovoltaic power station according to claim 3, characterized in that: The method of using local adaptive binarization to calculate the threshold of the pixel points in the foreground area, and locating the identification line area on the square panel structure in the foreground area according to the threshold, includes: Set the background threshold of the standard image f(x,y) to T; According to the background threshold T, the standard image f(x, y) is binarized, including: Among them, F(x,y) is the standard image after thresholding; If the grayscale value of the standard image f(x, y) is greater than the background threshold T, the identification line area of ​​the square panel structure is screened out.

7. The intelligent inspection method for photovoltaic power station according to claim 3, characterized in that: The method of using the Euclidean distance to calculate the width value L* of the marking line area includes: Calculate the Euclidean distance L(P1, P2) between two pixels: Wherein, x1 is the row number of pixel P1, y1 is the column number of pixel P1; x2 is the row number of pixel P2, y2 is the column number of pixel P2; Calculate the width Lp of the identification line: Lp=L(Pixel a ,Pixel (1) )×2; Among them, Pixel (1) Pixel is the edge pixel closest to the pixel of the identification line. a is the pixel in the identification line area; The calculated value L* of the marking line area width is calculated according to the width Lp of the marking line: L*=Lp·A'; Where A' is the width transformation parameter.

8. The intelligent inspection method for photovoltaic power stations according to any one of claims 1 to 7, characterized in that: The determining the location information of the fault detection point based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system includes: Calculating the displacement loss between the inspection path in the rectangular coordinate system of the UAV and the planned inspection path in the rectangular coordinate system of the earth; Among them, G(x) represents the inspection path in the rectangular coordinate system of the UAV, and H(x) represents the planned inspection path in the rectangular coordinate system of the earth; When the inspection path G(x) in the UAV rectangular coordinate system is shifted to coincide with the planned inspection path H(x) in the earth rectangular coordinate system, we have: get: G(x)'=H(x)+n; Among them, m1 and m2 are the left and right critical points of the inspection path respectively; Loss(H(x), G(x)) is the displacement loss; G(x)' is the new coordinate of the inspection path G(x) in the UAV rectangular coordinate system displaced to the earth rectangular coordinate system; n is the displacement vector.

9. An intelligent inspection device for a photovoltaic power station, characterized in that: include: A construction module, used to take a two-dimensional image of a photovoltaic power station using satellite remote sensing, and to construct a geodetic rectangular coordinate system based on the regional position of the photovoltaic power station in the two-dimensional image; An inspection module, configured to execute an inspection task by using a first drone to plan an inspection path, and determine the locations of inspection points of the solar panel on the inspection path; A correction module, used for performing grid processing on the regional position of the photovoltaic power station and arranging posture correction objects at the outer edge of the detection point position; Starting a second drone to inspect along the inspection path, and controlling the second drone to photograph the posture correction object at the inspection point to obtain a posture correction image; performing posture verification on the second drone based on the posture correction image; if the verification result meets the preset conditions, controlling the second drone to photograph the solar panel at the inspection point to obtain an original image of the solar panel; if the verification result does not meet the preset conditions, re-photographing the posture correction object; A network module is used to construct different types of image data sets based on the original image, and to construct a solar panel fault recognition network model, to perform solar panel fault recognition at each detection point position, and to obtain a fault detection point; The determination module is used to construct a drone rectangular coordinate system with the starting point of the second drone as the coordinate origin; based on the inspection path in the earth rectangular coordinate system and the inspection path in the drone rectangular coordinate system, determine the location information of the fault detection point.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the intelligent inspection method for a photovoltaic power station as claimed in any one of claims 1 to 8 is implemented.

Images