Photovoltaic cleaning device operation path generation method based on unmanned aerial vehicle

Abstract

The invention discloses a photovoltaic cleaning device operation path generation method based on an unmanned aerial vehicle, and relates to the technical field of point cloud data processing. According to the method, an unmanned aerial vehicle carries LiDAR and RTK sensors to obtain point cloud and GPS data of a photovoltaic station, and the point cloud of a single photovoltaic panel is obtained through the steps of preprocessing, ground point removal through rasterization, photovoltaic panel point cloud extraction through an HSV color space, inclination angle calculation and the like; and fitting the lower edge of the point cloud of the single photovoltaic panel, obtaining a position coordinate in combination with the relative distance offset between the RTK of the cleaning device and the edge, converting the position coordinate into a GPS coordinate, and fitting a working path. The method reduces the manual collection cost, improves the cleaning efficiency and accuracy, enhances the operation safety, reduces the operation and maintenance cost, is suitable for different photovoltaic stations, and is high in flexibility.

Description

Technical Field

[0001] The present application belongs to the field of point cloud data processing technology, and in particular relates to a method for generating an operation path for a photovoltaic cleaning device based on an unmanned aerial vehicle. Background Art

[0002] With the booming development of the new energy industry, photovoltaic power generation, as a pillar of clean energy, is becoming increasingly important. However, photovoltaic sites typically cover vast areas, and photovoltaic panels are exposed to the elements for long periods of time, inevitably being attacked by pollutants such as dust, snow, and bird droppings. These pollutants can significantly reduce the panels' power generation efficiency, shorten the equipment's lifespan, and ultimately affect the economic benefits of the entire photovoltaic plant.

[0003] Traditional photovoltaic (PV) plant operation and maintenance methods rely primarily on manual inspections and cleaning, a method that is not only inefficient but also faces the dual challenges of operational safety risks and high O&M costs. With the rapid development of drone technology, its application in the intelligent O&M of PV plants is gradually demonstrating great potential. Drones can be equipped with high-precision LiDAR and RTK sensors to quickly and comprehensively collect point cloud and GPS data from PV plants, providing detailed coverage of PV panel layout and status.

[0004] While there have been several inventions related to drone route planning, point cloud data processing, and path planning, these technologies still have many deficiencies in practical application. For example, the random forest classifier may not achieve high-precision classification results when there is a lack of sufficient labeled data; the A* search algorithm performs poorly in dynamic environments and is subject to resource limitations of edge computing; and the octree, as a spatial data structure, may face high computational overhead when processing large-scale point cloud data, affecting the effectiveness of real-time path planning or applications in dynamic environments. Summary of the Invention

[0005] The purpose of this application is to provide a method for generating an operating path for a photovoltaic cleaning device based on a drone. By acquiring the original point cloud data and GPS data of the photovoltaic station and going through a series of data processing and analysis steps, the operating path of the photovoltaic cleaning device is finally fitted so that the photovoltaic cleaning device can complete the cleaning task efficiently and accurately.

[0006] To achieve the above objectives, the present invention provides a method for generating an operation path for a photovoltaic cleaning device based on a drone, comprising:

[0007] Obtain the original point cloud data and GPS data of the photovoltaic station, and preprocess the original point cloud data to obtain preprocessed point cloud data; process the preprocessed point cloud data based on the rasterization method, remove the ground point cloud data, and retain the non-ground point cloud data;

[0008] The non-ground point cloud data is processed based on the HSV color space to extract the first photovoltaic panel point cloud data; the photovoltaic panel inclination angle is obtained, and the first photovoltaic panel point cloud data is processed based on the photovoltaic panel slope to obtain the second photovoltaic panel point cloud data; the second photovoltaic panel point cloud data is processed row by row to obtain the photovoltaic panel point cloud data of each row, and the photovoltaic panel point cloud data of each row is processed to extract the point cloud data of a single photovoltaic panel;

[0009] The lower edge of the point cloud data of a single photovoltaic panel is fitted to obtain the lower edge point. The relative distance between the RTK installation position of the photovoltaic cleaning device and the edge position of the photovoltaic cleaning device is obtained. The lower edge point is offset by the relative distance to obtain the position coordinates of the photovoltaic cleaning device in the point cloud data of a single photovoltaic panel; the position coordinates are converted into GPS position coordinates, and the operation path of the photovoltaic cleaning device is obtained by fitting.

[0010] The above method according to the embodiment of the present application may also have the following additional technical features:

[0011] Furthermore, the drone is equipped with a LiDAR sensor and an RTK sensor, which can be used to obtain the original point cloud data and GPS data of the entire photovoltaic station via a fixed route.

[0012] Furthermore, the original point cloud data is transmitted wirelessly to the ground workstation for random downsampling and radius filtering denoising to obtain preprocessed point cloud data.

[0013] Furthermore, the preprocessed point cloud data is divided into grid units, and the preprocessed point cloud data is mapped to a two-dimensional grid. All points in each grid unit are processed, and the minimum Z value of the points in the grid unit is calculated as the ground height of the grid unit. By traversing each grid unit and recording the minimum Z value, a ground height model is obtained; according to the ground height model, the preprocessed point cloud data is divided into ground point point cloud data and non-ground point point cloud data, the ground point point cloud data is removed, and the non-ground point point cloud data is retained.

[0014] Furthermore, the first photovoltaic panel point cloud data is centralized to obtain centralized point cloud data, and the covariance matrix of the centralized point cloud data is calculated. The covariance matrix is eigenvalue decomposition is performed to obtain the first eigenvalue, the second eigenvalue and the third eigenvalue, as well as the first eigenvector, the second eigenvector and the third eigenvector corresponding to the first eigenvalue, the second eigenvalue and the third eigenvalue respectively; the eigenvector corresponding to the smallest eigenvalue is taken as the plane normal vector of the photovoltaic panel, and the plane equation of the photovoltaic panel is constructed, with the Z-axis direction as the reference direction, the unit vector of the Z-axis is obtained, the cosine value of the angle between the plane normal vector of the photovoltaic panel and the unit vector of the Z-axis is calculated, and the inverse cosine function is performed on the cosine value of the angle to obtain the angle between the plane normal vector of the photovoltaic panel and the unit vector of the Z-axis, which is the inclination angle of the photovoltaic panel. The first photovoltaic panel point cloud data is extracted according to the inclination angle of the photovoltaic panel to obtain the second photovoltaic panel point cloud data.

[0015] The method for generating an operation path for a photovoltaic cleaning device based on a drone provided in the embodiment of the present application has the following beneficial technical effects compared with the prior art:

[0016] The embodiment of the present application uses drones to quickly acquire point cloud data and GPS data of photovoltaic stations, reducing the time and cost of manual data collection; automatically processes and analyzes data, quickly generates operation paths, and improves the efficiency of cleaning operations.

[0017] The embodiment of the present application utilizes a rasterization method and HSV color space to process point cloud data, accurately extracts photovoltaic panel point cloud data, and reduces misjudgments and omissions; by calculating the inclination angle of the photovoltaic panel, the photovoltaic panel point cloud data is accurately processed, thereby improving the accuracy of the operation path.

[0018] By fitting the operation path in the embodiment of the present application, the photovoltaic cleaning device can avoid obstacles, reduce collision risks, and improve operation safety; the precise GPS position coordinates ensure the precise positioning of the cleaning device during the operation process, avoiding safety hazards caused by position deviation.

[0019] The automatic generation of operation paths in the embodiment of the present application reduces manual intervention and lowers labor costs; the efficient cleaning operation reduces dust and dirt accumulation on the photovoltaic panels, improves the power generation efficiency of the photovoltaic panels, and reduces operation and maintenance costs.

[0020] The embodiments of the present application are applicable to photovoltaic stations of different sizes and layouts, and have strong adaptability and versatility; by adjusting parameters and algorithms, it can adapt to photovoltaic panels of different tilt angles and layouts, thereby improving the flexibility of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flow chart of a method for generating an operation path for a photovoltaic cleaning device based on a drone according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some, rather than all, structures related to the present application are shown in the accompanying drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0023] As used herein, the terms "comprise," "comprising," and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0024] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0025] like Figure 1 As shown, the embodiment of the present application provides a method for generating an operation path of a photovoltaic cleaning device based on a drone, comprising the following steps:

[0026] Step 101: obtain the original point cloud data and GPS data of the photovoltaic station, and preprocess the original point cloud data to obtain preprocessed point cloud data; process the preprocessed point cloud data based on a rasterization method, remove the ground point cloud data, and retain the non-ground point cloud data.

[0027] The drone in the embodiment of the present application is equipped with a LiDAR sensor and an RTK sensor to obtain raw point cloud data and GPS data from the photovoltaic station. The LiDAR sensor is used to obtain three-dimensional point cloud data from the photovoltaic station, providing rich spatial information. The RTK sensor is used to obtain high-precision GPS data, providing accurate geographic location information for the point cloud data. LiDAR (Light Detection and Ranging) is a technology that determines target distance, speed, and other information by emitting a laser beam at the target object and measuring the time it takes for the reflected light to return to the receiver. The LiDAR system uses laser pulses instead of radio waves as a radiation source. The laser source emits pulses directed at the target of interest, such as terrain. After the pulse encounters the target, a portion of the laser energy is reflected back to the sensor near the light source. By measuring the round-trip travel time of the emitted laser pulse, the LiDAR system can determine the distance between the sensor and the target. RTK (Real-Time Kinematic) is a real-time dynamic differential positioning technology that uses carrier phase observations for real-time dynamic relative positioning. The RTK system consists of three parts: a base station receiver, a data link (usually using radio or network), and a rover receiver. The base station receiver, mounted at a reference point with known coordinates, continuously observes satellites and transmits this data via a data link to the rover receiver. The rover receiver receives both base station data and satellite signals, processing this data in real time to produce centimeter-level positioning results.

[0028] The drone flies along a fixed route, ensuring comprehensive and efficient coverage of the entire photovoltaic plant, thereby acquiring complete raw point cloud data and GPS data. The raw point cloud data is transmitted wirelessly to a ground station for subsequent processing. At the ground station, the raw point cloud data undergoes random downsampling and radius filtering denoising to reduce data redundancy and noise, resulting in preprocessed point cloud data. Random downsampling and radius filtering denoising can reduce data redundancy and noise, improving the accuracy and efficiency of subsequent processing. The raw point cloud data transmitted to the ground station, which reaches the hundreds of millions, requires downsampling to reduce computational burden. Random downsampling technology is used to compress and simplify the raw point cloud data, reducing the number of data points without losing key features. This step not only significantly reduces the data volume and improves the computational efficiency of subsequent processing, but also reduces data noise, ensuring the quality of the raw point cloud data.

[0029] Specifically, raw point cloud data often contains noise points, especially in complex environments, which can affect subsequent feature extraction and path planning. Local neighborhood analysis and radius filtering can effectively identify and remove these noise points. Radius filtering calculates the number of other points within each point's radius. Points with fewer than a certain threshold are filtered out.

[0030] Radius filtering is performed on the raw point cloud data. Radius filtering is a common denoising method used in point cloud processing, primarily for removing outliers. Unlike statistical filtering, radius filtering performs denoising based on the distance relationship between a point and its neighborhood. The core idea of this method is to check the number of neighboring points within a certain radius for each point in the point cloud. If the number of neighboring points is insufficient, the point is considered an outlier and removed.

[0031] Radius filtering first needs to define a radius range, that is, specify a radius distance threshold. The radius defines the neighborhood range of each point. For each point in the point cloud, the radius filter will find all its neighboring points within the specified radius. Specifically, for each point, calculate its Euclidean distance to all other points, and select neighboring points whose distance is less than the set radius. If the number of points in the neighborhood of a point is less than a preset threshold, the point is considered an outlier and is removed from the point cloud. This preset threshold is usually set according to the density of the point cloud. Usually, sparser areas will have fewer neighborhood points. For all points, check the number of neighboring points one by one. If the number of neighboring points of a point is less than the set threshold, the point is considered an outlier and is removed.

[0032] The pre-processed point cloud data is divided into grid cells and mapped to a two-dimensional grid. All points within each grid cell are processed, and the minimum Z value of the points within the grid cell is calculated as the ground height of the grid cell. By traversing each grid cell and recording the minimum Z value, a ground height model is obtained. According to the ground height model, the pre-processed point cloud data is divided into ground point point cloud data and non-ground point point cloud data. The ground point point cloud data is removed, and the non-ground point point cloud data is retained for subsequent extraction of photovoltaic panel point cloud data. The rasterization method can quickly process large amounts of point cloud data and improve data processing efficiency. By calculating the minimum Z value within the grid cell, the ground height model can be accurately constructed, thereby effectively separating ground points from non-ground points.

[0033] Specifically, a rasterization method is used to segment the pre-processed point cloud data into a regular grid. The lowest point in each grid is considered a ground point, while all other points are considered non-ground points. This method effectively removes ground information from massive point cloud data, retaining the point cloud data of photovoltaic panels and other superstructures, improving the accuracy and efficiency of subsequent analysis.

[0034] Select an appropriate cellsize. The cellsize depends on the density of the point cloud and is usually chosen experimentally. Smaller cellsizes capture more detail but are more computationally intensive. Conversely, larger cellsizes may miss some detail but are less computationally intensive. Map the preprocessed point cloud data onto a 2D grid using the selected cellsize. Typically, the X and Y coordinates of the point cloud are mapped to the rows and columns of the grid, while the Z coordinate corresponds to a value in the grid.

[0035] After rasterization, the next step is to construct a ground height model based on the point cloud data for each grid cell. All points within each grid cell are processed, and the minimum Z value of the points within that grid cell is calculated, which serves as the ground height for that grid cell. By traversing each grid cell and recording the minimum Z value, a rasterized ground height model is ultimately obtained. These ground height values will serve as a reference for ground removal in subsequent steps.

[0036] Through the ground height model, the pre-processed point cloud data can be divided into ground point point cloud data and non-ground point point cloud data. For each point, compare its Z value with the minimum Z value of the corresponding grid cell. If the Z value of the point is close to or less than the Z value of the ground model, it is considered to be ground point point cloud data; otherwise, it is considered to be non-ground point point cloud data. Set a reasonable height difference (threshold) to determine whether the point is point cloud data. Generally speaking, the Z value of point cloud data will be very close to the rasterized ground height value, while non-ground point point cloud data will deviate from this value. For each point, check whether its Z value is within the range of the minimum Z value of the grid cell plus a certain threshold. If so, it is considered to be ground point point cloud data. According to the above classification, the ground point point cloud data is removed from the pre-processed point cloud data. The non-ground point point cloud data will be retained for subsequent processing.

[0037] In practice, step 101 has been successfully applied to data acquisition and processing at multiple photovoltaic sites. Using drones equipped with LiDAR and RTK sensors, raw point cloud data and GPS data are acquired along fixed routes. After preprocessing and rasterization at a ground workstation, high-quality non-ground point cloud data is generated, providing strong support for subsequent photovoltaic panel point cloud data extraction and operation path generation.

[0038] Step 102: Process the non-ground point cloud data based on the HSV color space to extract the first photovoltaic panel point cloud data; obtain the photovoltaic panel inclination angle, process the first photovoltaic panel point cloud data based on the photovoltaic panel slope, and obtain the second photovoltaic panel point cloud data; perform row-by-row processing on the second photovoltaic panel point cloud data to obtain photovoltaic panel point cloud data for each row, process the photovoltaic panel point cloud data for each row, and extract the point cloud data of a single photovoltaic panel.

[0039] The HSV (Hue, Saturation, Value) color space represents points in the RGB color space within an inverted cone, which better aligns with human visual perception of color. Using the HSV color space, you can set a specific color threshold range to extract points in the non-ground point cloud data that correspond to the color of the photovoltaic panel, forming the first photovoltaic panel point cloud data.

[0040] For each point (x i ,y i ,z i ) is centralized, that is, its mean is subtracted to obtain the centralized point cloud data. The coordinates after centering are:

[0041]

[0042] is the mean value of each point in the first photovoltaic panel point cloud data, that is:

[0043]

[0044] Calculate the covariance matrix of the centralized point cloud data to describe the correlation between data points. The formula is as follows:

[0045] The covariance matrix is a 3×3 symmetric matrix with the form:

[0046]

[0047] Here, cov(x,x) represents the covariance of the data in the x direction, cov(x,y) represents the covariance between the x and y directions, and so on.

[0048] Perform eigenvalue decomposition on the covariance matrix to obtain the first eigenvalue λ1, the second eigenvalue λ2 and the third eigenvalue λ3 corresponding to the first eigenvector v1, the second eigenvector v2 and the third eigenvector v3. These eigenvalues and eigenvectors represent the variance and main direction of the data in the three directions respectively. Take the minimum value of the first eigenvalue λ1, the second eigenvalue λ2 and the third eigenvalue λ3 as the minimum eigenvalue λ minThrough PCA analysis, the normal vector of the plane is the eigenvector corresponding to the minimum eigenvalue of the covariance matrix. PCA (Principal Component Analysis) is a commonly used statistical method and machine learning technique, mainly used for dimensionality reduction, data compression, and feature extraction. PCA uses an orthogonal transformation to convert a set of potentially correlated high-dimensional data into a set of linearly uncorrelated low-dimensional data. This set of low-dimensional data is called the principal component. The first principal component has the largest variance, and subsequent principal components have the next largest variance and are orthogonal to the previous principal component.

[0049] With the minimum eigenvalue λ min The corresponding eigenvector v min As the normal vector of the photovoltaic panel plane, let v min =(A,B,C), construct the plane equation of the photovoltaic panel, the formula is as follows:

[0050] Ax+By+Cz+D=0

[0051] Where D is the distance between the photovoltaic panel plane and the origin. The center of gravity of the point cloud is taken as a point (x0, y0, z0) on the photovoltaic panel plane, then D = -(Ax0+By0+Cz0)

[0052] Taking the Z-axis as the reference direction, calculate the cosine of the angle between the normal vector of the photovoltaic panel plane and the Z-axis unit vector z = (0, 0, 1). The formula is as follows:

[0053]

[0054] |z|=1 is the magnitude of the unit vector on the Z axis.

[0055] Therefore, the formula for the cosine value of the angle between the photovoltaic panel plane is as follows:

[0056]

[0057] Then, the included angle, i.e. the tilt angle of the photovoltaic panel, is obtained through the arc cosine function. The formula is as follows:

[0058]

[0059] Photovoltaic panel areas often span multiple panels, necessitating refinement of these areas. First, the panels are processed row by row, along the x and y axes. Each row may contain multiple adjacent panels. Then, point cloud data for each panel is extracted from each row. This row-by-row extraction method ensures accurate extraction of the geometric features of each panel, providing precise geometric information for path planning.

[0060] Based on the calculated panel tilt angle, the first panel point cloud data is further processed to extract second panel point cloud data that matches the panel tilt characteristics. The second panel point cloud data is then processed row by row, dividing the point cloud data into rows of panel point cloud data based on the panel arrangement. Each row of panel point cloud data is processed, and the point cloud data for a single panel is extracted by setting an appropriate threshold or algorithm.

[0061] Specifically, the photovoltaic panels are first divided into rows, and the photovoltaic panel row segmentation based on the Y coordinate value can realize the organization of point cloud data by row. The point cloud data of photovoltaic panels has a certain regularity in space. The rows of photovoltaic panels are applicable to the scenario that they are parallel, and the Y coordinate value range of each row is relatively stable, so the rows are divided according to the Y prior value; then the point cloud of each row of photovoltaic panels is Euclidean clustering. Create a KD tree. The KD tree has an efficient spatial index structure for quickly finding neighboring points in the point cloud. Set the clustering tolerance, that is, the maximum distance between two points (unit: meter). Here it is set to 0.02 meters, which means that points with a distance of less than 2 centimeters will be considered to belong to the same cluster.

[0062] A KD tree (K-Dimensional Tree) is a data structure used to organize points in k-dimensional space. It is particularly suitable for fast nearest neighbor searches and range searches in multidimensional space. A KD tree is a binary tree used to partition k-dimensional space. Each node represents a k-dimensional point, and the data space is divided into two parts along a single dimension. At each level of the tree, a dimension is selected such that the point corresponding to the median of that dimension serves as the split point. The left subtree contains points less than the split point, and the right subtree contains points greater than the split point.

[0063] By accurately calculating the HSV color space and the tilt angle of the photovoltaic panels, we can accurately extract the point cloud data of the photovoltaic panels, providing a reliable foundation for subsequent operation path generation. The entire processing process is highly automated, reducing manual intervention and improving processing efficiency. This method is applicable to photovoltaic panels with different tilt angles and arrangements, and has strong flexibility and adaptability.

[0064] In summary, the implementation of step 102 is crucial for generating the operating path for photovoltaic cleaning equipment. By accurately extracting the point cloud data of photovoltaic panels, accurate data support is provided for subsequent bottom edge fitting, position coordinate conversion, and operating path fitting, thereby improving the efficiency and accuracy of photovoltaic cleaning operations. This method also provides a new technical means and solution for the operation and maintenance management of photovoltaic stations.

[0065] Step 103: Perform lower edge fitting on the point cloud data of a single photovoltaic panel to obtain the lower edge point, obtain the relative distance between the RTK installation position of the photovoltaic cleaning device and the edge position of the photovoltaic cleaning device, offset the lower edge point by the relative distance, and obtain the position coordinates of the photovoltaic cleaning device in the point cloud data of the single photovoltaic panel; convert the position coordinates into GPS position coordinates, and fit the operation path of the photovoltaic cleaning device.

[0066] This embodiment first extracts the bottom edge points of a photovoltaic panel from the point cloud data of a single photovoltaic panel. A point cloud processing algorithm (such as least squares or RANSAC) is used to fit the point cloud data of the single photovoltaic panel to obtain the bottom edge points of the photovoltaic panel. These points represent the lowest point of the photovoltaic panel in the vertical direction and are the key locations that require attention in the photovoltaic cleaning device.

[0067] Next, determine the relative distance between the RTK installation location of the photovoltaic cleaning device and the edge of the cleaning device. This distance is obtained through actual measurement or design parameters. This distance is the key parameter for offsetting the lower edge point to obtain the cleaning device's location coordinates.

[0068] The bottom edge point is offset by a relative distance to determine the location coordinates of the PV cleaning device within the point cloud data of a single PV panel. Based on the obtained relative distance, the coordinates of the bottom edge point in the point cloud data are offset. The offset direction is typically aligned with the tilt of the PV panel to ensure that the cleaning device accurately reaches the bottom edge of the panel.

[0069] Convert the offset position coordinates to GPS position coordinates. Utilizing the conversion relationship between point cloud data and GPS data (e.g., using the coordinate conversion parameters of an RTK positioning system), convert the offset position coordinates from the point cloud coordinate system to the GPS coordinate system. This allows the cleaning device to navigate and locate based on the GPS position coordinates.

[0070] The PV cleaning equipment's operating path is fitted based on the converted GPS coordinates. Using path planning algorithms (such as B-spline and Bezier curves), the converted GPS coordinates are fitted to create a smooth, continuous operating path. This path guides the cleaning equipment to perform efficient cleaning operations within the PV plant.

[0071] By using bottom edge fitting and relative distance offset, the cleaning device's position on the photovoltaic panel can be accurately determined, improving cleaning accuracy. The entire process is automated by an algorithm, reducing manual intervention and improving efficiency. The system can adapt to photovoltaic panels of varying tilt angles and shapes, ensuring wide applicability.

[0072] In practice, step 103 has been successfully applied to cleaning operations at multiple photovoltaic plants. Through precise bottom-edge fitting and position coordinate conversion, the cleaning device can accurately locate the bottom edge of the photovoltaic panel and perform efficient cleaning operations according to the fitted operation path. This not only improves cleaning efficiency and quality, but also reduces labor costs and operational risks.

[0073] It should be noted that, in the present application, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0074] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. A method for generating an operation path for a photovoltaic cleaning device based on an unmanned aerial vehicle, characterized in that: The method comprises: Acquire original point cloud data and GPS data of the photovoltaic station, and pre-process the original point cloud data to obtain pre-processed point cloud data; process the pre-processed point cloud data based on a rasterization method, remove ground point cloud data, and retain non-ground point cloud data; The non-ground point cloud data is processed based on the HSV color space to extract the first photovoltaic panel point cloud data; the inclination angle of the photovoltaic panel is obtained, and the first photovoltaic panel point cloud data is processed based on the photovoltaic panel slope to obtain the second photovoltaic panel point cloud data; the second photovoltaic panel point cloud data is processed row by row to obtain photovoltaic panel point cloud data for each row, and the photovoltaic panel point cloud data for each row is processed to extract the point cloud data of a single photovoltaic panel; Perform lower edge fitting on the point cloud data of the single photovoltaic panel to obtain the lower edge point, obtain the relative distance between the RTK installation position of the photovoltaic cleaning device and the edge position of the photovoltaic cleaning device, offset the lower edge point by the relative distance, and obtain the position coordinates of the photovoltaic cleaning device in the point cloud data of the single photovoltaic panel; convert the position coordinates into GPS position coordinates, and obtain the operation path of the photovoltaic cleaning device by fitting.

2. The method for generating an operation path of a photovoltaic cleaning device based on a drone according to claim 1, wherein: The drone is equipped with a LiDAR sensor and an RTK sensor, and the original point cloud data and GPS data of the entire photovoltaic station are obtained by the drone along a fixed route.

3. The method for generating an operation path of a photovoltaic cleaning device based on a drone according to claim 1, wherein: The method comprises: The original point cloud data is transmitted to a ground workstation by wireless transmission for random downsampling and radius filtering denoising processing to obtain preprocessed point cloud data.

4. The method for generating an operation path of a photovoltaic cleaning device based on a drone according to claim 1, wherein: The method comprises: The pre-processed point cloud data is divided into grid units, and the pre-processed point cloud data is mapped to a two-dimensional grid, all points in each grid unit are processed, and the minimum Z value of the points in the grid unit is calculated as the ground height of the grid unit. A ground height model is obtained by traversing each grid unit and recording the minimum Z value; according to the ground height model, the pre-processed point cloud data is divided into ground point point cloud data and non-ground point point cloud data, the ground point point cloud data is removed, and the non-ground point point cloud data is retained.

5. The method for generating an operation path of a photovoltaic cleaning device based on a drone according to claim 1, wherein: The method comprises: The first photovoltaic panel point cloud data is centralized to obtain centralized point cloud data, and the covariance matrix of the centralized point cloud data is calculated. The covariance matrix is eigenvalue decomposition is performed to obtain the first eigenvalue, the second eigenvalue and the third eigenvalue, and the first eigenvector, the second eigenvector and the third eigenvector corresponding to the first eigenvalue, the second eigenvalue and the third eigenvalue respectively; the eigenvector corresponding to the smallest eigenvalue is taken as the plane normal vector of the photovoltaic panel, and the plane equation of the photovoltaic panel is constructed, with the Z-axis direction as the reference direction, the unit vector of the Z-axis is obtained, the cosine value of the angle between the plane normal vector of the photovoltaic panel and the unit vector of the Z-axis is calculated, and the inverse cosine function is performed on the cosine value of the angle to obtain the angle between the plane normal vector of the photovoltaic panel and the unit vector of the Z-axis, the angle is the inclination angle of the photovoltaic panel, and the first photovoltaic panel point cloud data is extracted according to the inclination angle of the photovoltaic panel to obtain the second photovoltaic panel point cloud data.

Images