An unmanned intelligent photovoltaic inspection method

Abstract

The present invention relates to the field of drone technology, and discloses an unmanned intelligent photovoltaic inspection method. The method comprises: controlling a long-endurance drone to inspect the photovoltaic panel area, obtaining multiple photovoltaic panel images taken by the long-endurance drone; performing clarity detection and fault detection on each photovoltaic panel image, screening target photovoltaic panel images that fail the clarity detection or have faults, and obtaining the target posture information carried by the target photovoltaic panel image, and determining the target photovoltaic panel code corresponding to the target posture information; using the shortest path planning algorithm, planning the flight path of a short-endurance drone according to the target photovoltaic panel code; and controlling the short-endurance drone to inspect according to the flight path. The present invention reduces the complexity and cost of cruising as a whole, and ensures the continuity and efficiency of inspections, through comprehensive inspections by long-endurance drones and key inspections by short-endurance drones.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to an unmanned intelligent photovoltaic inspection method. Background Art

[0002] Photovoltaic panel drone inspections are gradually replacing traditional manual inspections with their high efficiency, accuracy, safety and economy. With the continuous development and improvement of drone technology, as well as the continuous expansion of the scale of photovoltaic power stations and the increase in operation and maintenance needs, drone inspections will play an increasingly important role in the operation and maintenance of photovoltaic power stations.

[0003] In the existing technology, the inspection distance and flight time of drones are also limited due to their limited battery capacity and flight time. For areas that require extensive inspections, such as large photovoltaic power plants, drones may need to take off and land multiple times to complete the inspection task. This not only increases the complexity and cost of inspections, but may also affect the continuity and efficiency of inspections.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides an unmanned intelligent photovoltaic inspection method, which reduces the overall complexity and cost of patrolling through comprehensive inspections by long-endurance drones and key inspections by short-endurance drones, thereby ensuring the continuity and efficiency of inspections.

[0006] The embodiment of the present invention provides an unmanned intelligent photovoltaic inspection method, comprising:

[0007] Controlling a long-endurance drone to inspect the photovoltaic panel area, and obtaining a plurality of photovoltaic panel images taken by the long-endurance drone; each photovoltaic panel image carries the position information of the long-endurance drone when taking the image;

[0008] Performing clarity detection and fault detection on each photovoltaic panel image, screening target photovoltaic panel images that do not meet the clarity requirements or have faults, and obtaining target posture information carried by the target photovoltaic panel images, and determining at least one target photovoltaic panel code corresponding to the target posture information;

[0009] Planning a flight path of the short-endurance UAV according to the target photovoltaic panel code;

[0010] The short-endurance UAV is controlled to perform inspection according to the flight path; wherein the endurance of the short-endurance UAV is shorter than the endurance of the long-endurance UAV.

[0011] Optionally, after controlling the short-endurance UAV to perform inspection according to the flight path, the method further includes:

[0012] Acquire multiple target photovoltaic panel images taken by the short-endurance drone;

[0013] Perform fault detection on each target photovoltaic panel image, and if a fault is determined, notify maintenance personnel to repair the faulty target photovoltaic panel;

[0014] In response to the maintenance completion operation of the maintenance personnel, deleting the photovoltaic panel code whose maintenance has been completed from all target photovoltaic panel codes;

[0015] The flight path of the short-endurance UAV is replanned according to the remaining target photovoltaic panel codes, and the operation of controlling the short-endurance UAV to perform inspection according to the flight path is executed.

[0016] Optionally, after controlling the short-endurance UAV to perform inspection according to the flight path, the method further includes:

[0017] Acquire multiple target photovoltaic panel images taken by the short-endurance drone;

[0018] Perform fault detection on each target photovoltaic panel image, and if no fault exists, delete the photovoltaic panel code without fault from all target photovoltaic panel codes;

[0019] The flight path of the short-endurance UAV is replanned according to the remaining target photovoltaic panel codes, and the operation of controlling the short-endurance UAV to perform inspection according to the flight path is executed.

[0020] Optionally, controlling the short-endurance UAV to perform inspection according to the flight path includes:

[0021] Determining the number of times to photograph each target photovoltaic panel according to the clarity and fault type of the target photovoltaic panel image;

[0022] The short-endurance drone is controlled to photograph each target photovoltaic panel according to the number of times.

[0023] Optionally, a shortest path planning algorithm is used to plan a flight path of the short-endurance UAV according to the target photovoltaic panel code, including:

[0024] Determining the position information of the short-endurance UAV during shooting according to the target photovoltaic panel code;

[0025] The shortest path planning algorithm is used to determine the flight path of the short-endurance UAV according to the posture information and flight route of the short-endurance UAV during shooting.

[0026] Optionally, before controlling the long-endurance drone to inspect the photovoltaic panel area, it also includes:

[0027] The long-endurance UAV and the short-endurance UAV are calibrated in position and posture to determine the corresponding relationship between the position and posture information of the long-endurance UAV and the short-endurance UAV during shooting and the photovoltaic panel coding.

[0028] Optionally, performing posture calibration on the long-endurance UAV and the short-endurance UAV includes:

[0029] Controlling the calibration photovoltaic panel to flip according to a set angle; the calibration photovoltaic panel is any photovoltaic panel in the photovoltaic panel area;

[0030] Controlling the long-endurance drone and the short-endurance drone to photograph the calibration photovoltaic panel in multiple postures, and identifying the multiple calibration images taken;

[0031] When the target calibration image is a calibration photovoltaic panel flipped at a set angle, the calibration photovoltaic panel number is bound to the posture information when the target calibration image was taken.

[0032] Optionally, when the calibration image is a calibration photovoltaic panel flipped at a set angle, binding the calibration photovoltaic panel number with the pose information when the calibration photovoltaic panel is photographed includes:

[0033] When the area of ​​the calibration photovoltaic panel in the calibration image is smaller than a set value, the calibration photovoltaic panel number is bound to the posture information when the calibration photovoltaic panel is photographed; wherein the set value is determined according to the set angle.

[0034] Optionally, perform fault detection on each PV panel image, including:

[0035] Each photovoltaic panel image is input into the fault detection model to obtain the fault type;

[0036] The fault detection model comprises a feature extraction layer, a fully connected layer and a classification layer, and the fault detection model is trained based on photovoltaic panel images of known fault types.

[0037] The embodiments of the present invention have the following technical effects:

[0038] 1. Through comprehensive inspections by long-endurance drones and key inspections by short-endurance drones, the complexity and cost of patrols are reduced overall, ensuring the continuity and efficiency of inspections.

[0039] 2. Through intelligent clarity detection and fault detection, images with possible problems can be screened out from all photovoltaic panel images, improving detection efficiency.

[0040] 3. The present invention adopts the shortest path planning algorithm to plan the flight path of the short-endurance UAV according to the target photovoltaic panel code, and controls the short-endurance UAV to perform inspections according to the flight path, thereby minimizing the energy consumption of the short-endurance UAV. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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.

[0042] Figure 1 This is a flow chart of an unmanned intelligent photovoltaic inspection method provided by an embodiment of the present invention;

[0043] Figure 2 It is a top view of the existing photovoltaic panel area;

[0044] Figure 3 This is a comparison diagram of the areas of the unflipped photovoltaic panel and the flipped photovoltaic panel provided in this embodiment. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. 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 belong to the scope of protection of the present invention.

[0046] This embodiment provides an unmanned intelligent photovoltaic inspection method, which is applicable to the technical field of using drones to inspect photovoltaic panels. The method provided in this embodiment is executed by a host computer. Figure 1 , the method provided in this embodiment includes:

[0047] S110: Control a long-endurance drone to inspect a photovoltaic panel area, and obtain a plurality of photovoltaic panel images taken by the long-endurance drone.

[0048] The photovoltaic panel area includes a plurality of photovoltaic panels arranged in sequence. The size of each photovoltaic panel can be the same or different. Figure 2It is a top view of the existing photovoltaic panel area. A host computer is set on the ground close to the photovoltaic panel area. The host computer is connected to the long-endurance drone for communication. The host computer sends a mode 1 inspection control instruction to the long-endurance drone, that is, to inspect all photovoltaic panels in the photovoltaic panel area. When flying over each photovoltaic panel, it hovers and stabilizes in a certain posture and photographs the photovoltaic panel to obtain an image of each photovoltaic panel. Among them, each photovoltaic panel image carries the posture information of the long-endurance drone when shooting, including the three-dimensional position and posture in space, which can be determined by the positioning system and sensors such as gyroscopes in the long-endurance drone. For details, please refer to the prior art, which will not be repeated here.

[0049] It should be noted that before controlling the long-endurance drone to inspect the photovoltaic panel area, the host computer needs to write the posture information required for shooting each photovoltaic panel into the long-endurance drone so that the long-endurance drone can hover and shoot according to the written posture information. Then, the long-endurance drone is calibrated to determine the correspondence between the posture information of the long-endurance drone and the photovoltaic panel code when shooting. The photovoltaic panel code is used to uniquely identify a photovoltaic panel. Specifically, any photovoltaic panel in the photovoltaic panel area is taken as the calibration photovoltaic panel A, and the calibration photovoltaic panel A is flipped according to the set angle by the host computer control; the set angle is the angle that makes the top view of the photovoltaic panel have a significant shape change, for example, 60 degrees. The remaining photovoltaic panels are not flipped, so that the calibration photovoltaic panel A is obviously different from the other photovoltaic panels in the top view. The long-endurance drone is controlled by the host computer to shoot the calibration photovoltaic panel A in multiple postures, and a calibration image is taken in each posture. Since it is currently unknown what posture can be used to photograph the calibration photovoltaic panel A, the long-endurance drone is controlled to continuously fly and hover over the calibration photovoltaic panel A to capture images in different postures. The multiple calibration images captured are identified, that is, whether the calibration image contains the calibration photovoltaic panel A flipped at a set angle. When the target calibration image is the calibration photovoltaic panel A flipped at a set angle, the calibration photovoltaic panel number (i.e. A) is bound to the posture information when the target calibration image was captured.

[0050] Preferably, considering that the shape of the photovoltaic panel is relatively regular and thin, the area of ​​the calibration photovoltaic panel in the top view after flipping the set angle is greatly reduced compared to when it is not flipped. By using this feature, whether the calibration image is the calibration photovoltaic panel flipped according to the set angle can be determined in the following simpler and more accurate method: Figure 3This is a comparison chart of the areas of the unflipped photovoltaic panel and the flipped photovoltaic panel provided in this embodiment. The area of ​​the calibration photovoltaic panel is determined by identifying the number of black pixels in the calibration image. When the area of ​​the calibration photovoltaic panel is less than the set value (determined according to the set angle of 60 degrees, that is, the image area of ​​the calibration photovoltaic panel under the top-down perspective when flipped 60 degrees), the calibration photovoltaic panel number is bound to the posture information when shooting the calibration photovoltaic panel. According to the above calibration method, the corresponding relationship between each photovoltaic panel code and the posture information when shooting by the long-endurance drone can be obtained.

[0051] This embodiment provides a simple and easy calibration method by utilizing the characteristics of the photovoltaic panel, which is flippable, regular in shape and very thin.

[0052] S120, performing clarity detection and fault detection on each photovoltaic panel image, screening target photovoltaic panel images that fail the clarity detection or have faults, and obtaining target posture information carried by the target photovoltaic panel images, and determining at least one target photovoltaic panel code corresponding to the target posture information.

[0053] Since the long-endurance drone is hovering in the air to shoot, it is affected by shaking or the wind, insects, birds, floating dust, etc. in nature, which will cause the image to be blurred and unable to illuminate whether the photovoltaic panel is operating normally. Therefore, the edge detection method or the statistical characteristic method is used to detect the clarity of each photovoltaic panel image. If the clarity test passes, further fault detection is performed; if the clarity test fails, the photovoltaic panel image is used as the target photovoltaic panel image.

[0054] The fault detection method is as follows: each photovoltaic panel image is input into the fault detection model to obtain the fault type; the fault types include no fault, stain, water stain and light spot. Among them, the fault detection model includes a feature extraction layer, a fully connected layer and a classification layer. The fault detection model is trained based on photovoltaic panel images with known fault types. For example, 1,000 photovoltaic panel images are collected in advance, and the photovoltaic panel surfaces on 800 images have faults such as hot spots, stains, and water stains, which are marked as stain, water stain and light spot labels; the photovoltaic panel surfaces on 200 images have no faults, which are marked as no fault labels. These 1,000 photovoltaic panel images are input into the feature extraction layer, and the features of the photovoltaic panel images are extracted through operations such as convolution, activation function and pooling. In this process, each vector represents some information in the image, such as edges, textures, etc. In the fully connected layer, all feature vectors are flattened into a vector to represent the features of the entire photovoltaic panel image. This feature vector can be regarded as splitting the photovoltaic panel image from left to right and from top to bottom into several blocks and arranging them in order. In this way, complex image information can be converted into a vector representation. The aforementioned vector representation is mapped into two probability values ​​through the classification layer (i.e., softmax layer): the probability value of no fault and the probability value of fault. The classification result corresponding to the probability value greater than the threshold (e.g., 90%) is taken as the fault type of the photovoltaic panel image. During training, the parameters in the fault detection model are iterated by minimizing the distance between the fault type output by the model and the label, thereby obtaining a trained fault detection model.

[0055] This embodiment proposes a deep learning method to detect whether a photovoltaic panel has a fault and the type of fault, thereby improving detection efficiency and intelligence.

[0056] For the convenience of description and interval, the photovoltaic panel images that fail the clarity test or have faults are called target photovoltaic panel images. There are multiple target photovoltaic panel images, and the target pose information carried by each target photovoltaic panel image is obtained, that is, the pose information of each target photovoltaic panel image taken by the long-endurance drone. Since the pose information is pre-calibrated and the correspondence between the pose information and the photovoltaic panel code is determined, the target photovoltaic panel code corresponding to the target pose information can be obtained. When there are multiple target pose information, multiple target photovoltaic panel codes can be determined, and these target photovoltaic panels need to be re-photographed.

[0057] S130: Using the shortest path planning algorithm, plan the flight path of the short-endurance UAV according to the target photovoltaic panel code.

[0058] Specifically, the position information of the short-endurance UAV during shooting is determined according to the target photovoltaic panel code; and the flight path of the short-endurance UAV is determined according to the position information and flight route of the short-endurance UAV during shooting by using the shortest path planning algorithm.

[0059] The host computer is connected to the short-endurance drone, and the flight time of the short-endurance drone is shorter than that of the long-endurance drone. Since the target photovoltaic panel is only a part of all photovoltaic panels, the short-endurance drone can be enabled for inspection to save energy and cost.

[0060] The host computer sends a mode 2 inspection control command to the short-endurance drone, that is, to inspect the target photovoltaic panels identified by the target photovoltaic panel codes in the photovoltaic panel area. Prior to this, the host computer needs to write the posture information required for shooting each target photovoltaic panel into the short-endurance drone, so that the short-endurance drone can hover and shoot according to the written posture information. Then, the short-endurance drone is calibrated to determine the corresponding relationship between the posture information of the short-endurance drone when shooting and the photovoltaic panel code. For details, please refer to the calibration method of the long-endurance UAV, which controls the calibration photovoltaic panel to flip according to the set angle; the calibration photovoltaic panel is any photovoltaic panel in the photovoltaic panel area; the short-endurance UAV is controlled to shoot the calibration photovoltaic panel in multiple postures, and recognize the multiple calibration images taken; when the target calibration image is a calibration photovoltaic panel flipped at a set angle, the calibration photovoltaic panel number is bound to the posture information when the target calibration image is taken; further, when the area of ​​the calibration photovoltaic panel in the calibration image is less than the set value, the calibration photovoltaic panel number is bound to the posture information when the calibration photovoltaic panel is taken.

[0061] In this step, the position information required to photograph each target photovoltaic panel constitutes each point in space. The starting point of the short-endurance drone is fixed, which is a fixed position on the ground. After hovering and photographing at each point from the starting point, it can return directly to the starting point from the last point. The shortest path planning algorithm can be used to plan the shortest path between each point in space. The shortest path planning algorithm can be a greedy algorithm for solving the single-source shortest path: Dijkstra algorithm.

[0062] The host computer needs to write the posture information of the short-endurance drone when shooting the target photovoltaic panel, as well as the flight route between each point into the short-endurance drone, so that the short-endurance drone can fly, hover and shoot according to the posture and flight route of each point.

[0063] Since the target photovoltaic panel is likely to have a fault, more images need to be taken to increase fault tolerance. Optionally, the number of shots of each target photovoltaic panel is determined according to the clarity and fault type of the target photovoltaic panel image; the short-endurance drone is controlled to shoot each target photovoltaic panel according to the number of shots. When calculating the clarity, taking the grayscale variance method as an example, when fully focused, the image is the clearest and the high-frequency components in the image are the most. Therefore, the grayscale change can be used as the basis for focusing evaluation, and the clarity can be evaluated by calculating the image grayscale variance. The fault types of the target photovoltaic panel are quantified as hot spots 0.2, stains 0.5, and water stains 0.3. The numerical value of clarity (such as image grayscale variance) and the quantized value of the fault type are summed and multiplied by a fixed coefficient of 100 to obtain the number of shots of each target photovoltaic panel. The method of quantizing the value and the method of calculating the number of shots can be determined according to business needs.

[0064] S140, controlling the short-endurance UAV to perform inspection according to the flight path.

[0065] When the short-endurance drone flies back to the starting point, the host computer reads the captured image and the posture information carried by the short-endurance drone to determine which photovoltaic panel the image is. Fault detection is performed on each target photovoltaic panel image, and a specific fault detection model is used, which will not be repeated here. If it is determined that there is a fault, the maintenance personnel are notified to repair the faulty target photovoltaic panel. After the maintenance personnel completes the maintenance, they input the code of the target photovoltaic panel to the host computer to perform the maintenance completion operation. In response to the maintenance personnel's maintenance completion operation, the host computer deletes the maintenance completed photovoltaic panel code in the target photovoltaic panel code; re-plans the flight path of the short-endurance drone according to the remaining target photovoltaic panel code, and executes the operation of controlling the short-endurance drone to perform inspections according to the flight path. That is, re-plan the next flight path to inspect again.

[0066] Optionally, if there is no fault, delete the photovoltaic panel codes that do not have a fault from all target photovoltaic panel codes. For example, an unclear image taken by a long-endurance drone may be fault-free. Replan the flight path of the short-endurance drone based on the remaining target photovoltaic panel codes, and perform the operation of controlling the short-endurance drone to perform inspections based on the flight path. That is, replan the next flight path for another inspection.

[0067] The embodiments of the present invention have the following technical effects:

[0068] 1. Through comprehensive inspections by long-endurance drones and key inspections by short-endurance drones, the complexity and cost of patrols are reduced overall, ensuring the continuity and efficiency of inspections.

[0069] 2. Through intelligent clarity detection and fault detection, images with possible problems can be screened out from all photovoltaic panel images, improving detection efficiency.

[0070] 3. The present invention adopts the shortest path planning algorithm to plan the flight path of the short-endurance UAV according to the target photovoltaic panel code, and controls the short-endurance UAV to perform inspections according to the flight path, thereby minimizing the energy consumption of the short-endurance UAV.

[0071] It should be noted that the terms used in the present invention are only for describing specific embodiments, rather than limiting the scope of the present application. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular, but may also include the plural. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the presence of other identical elements in the process, method or device including the elements.

[0072] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0073] 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 replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. An unmanned intelligent photovoltaic inspection method, characterized in that: include: The photovoltaic panel used for control calibration is flipped according to the set angle; The calibration photovoltaic panel is any photovoltaic panel in the photovoltaic panel area; the long-endurance drone and the short-endurance drone are controlled to shoot the calibration photovoltaic panel in multiple postures, and multiple calibration images taken are identified; when the target calibration image is a calibration photovoltaic panel flipped at a set angle, the calibration photovoltaic panel number is bound to the posture information when the target calibration image is shot; to determine the corresponding relationship between the posture information of the long-endurance drone and the short-endurance drone when shooting and the photovoltaic panel code; Controlling the long-endurance drone to inspect the photovoltaic panel area, and obtaining a plurality of photovoltaic panel images taken by the long-endurance drone; each photovoltaic panel image carries the position information of the long-endurance drone when taking the image; Perform clarity detection and fault detection on each photovoltaic panel image, screen target photovoltaic panel images that fail the clarity detection or have faults, obtain target posture information carried by the target photovoltaic panel images, and determine the target photovoltaic panel code corresponding to the target posture information; Using the shortest path planning algorithm, the flight path of the short-endurance UAV is planned according to the target photovoltaic panel code; The short-endurance UAV is controlled to perform inspection according to the flight path; wherein the endurance of the short-endurance UAV is shorter than the endurance of the long-endurance UAV.

2. The method according to claim 1, characterized in that: After controlling the short-endurance UAV to perform inspection according to the flight path, the method further includes: Acquire multiple target photovoltaic panel images taken by the short-endurance drone; Perform fault detection on each target photovoltaic panel image, and if a fault is determined, notify maintenance personnel to repair the faulty target photovoltaic panel; In response to the maintenance completion operation of the maintenance personnel, deleting the photovoltaic panel code whose maintenance has been completed from all target photovoltaic panel codes; The flight path of the short-endurance UAV is replanned according to the remaining target photovoltaic panel codes, and the operation of controlling the short-endurance UAV to perform inspection according to the flight path is executed.

3. The method according to claim 2, characterized in that After controlling the short-endurance UAV to perform inspection according to the flight path, the method further includes: Acquire multiple target photovoltaic panel images taken by the short-endurance drone; Perform fault detection on each target photovoltaic panel image, and if no fault exists, delete the photovoltaic panel code without fault from all target photovoltaic panel codes; The flight path of the short-endurance UAV is replanned according to the remaining target photovoltaic panel codes, and the operation of controlling the short-endurance UAV to perform inspection according to the flight path is executed.

4. The method according to claim 3, characterized in that: Controlling the short-endurance UAV to perform inspection according to the flight path includes: Determining the number of times to photograph each target photovoltaic panel according to the clarity and fault type of the target photovoltaic panel image; The short-endurance drone is controlled to photograph each target photovoltaic panel according to the number of times.

5. The method according to claim 4, characterized in that The shortest path planning algorithm is used to plan the flight path of the short-endurance UAV according to the target photovoltaic panel code, including: Determining the position information of the short-endurance UAV during shooting according to the target photovoltaic panel code; The shortest path planning algorithm is used to determine the flight path of the short-endurance UAV according to the posture information and flight route of the short-endurance UAV during shooting.

6. The method according to claim 1, characterized in that When the calibration image is a calibration photovoltaic panel flipped at a set angle, the calibration photovoltaic panel number is bound to the posture information when the calibration photovoltaic panel is photographed, including: When the area of ​​the calibration photovoltaic panel in the calibration image is smaller than a set value, the calibration photovoltaic panel number is bound to the posture information when the calibration photovoltaic panel is photographed; wherein the set value is determined according to the set angle.

7. The method according to claim 6, characterized in that Perform fault detection on each PV panel image, including: Each photovoltaic panel image is input into the fault detection model to obtain the fault type; The fault detection model comprises a feature extraction layer, a fully connected layer and a classification layer, and the fault detection model is trained based on photovoltaic panel images of known fault types.

Images