Unmanned aerial vehicle control method and system integrating scanning and photovoltaic panel cleaning

Abstract

The invention provides an unmanned aerial vehicle control method and system integrating scanning and photovoltaic panel cleaning, and the method comprises the steps: scanning the surface of a photovoltaic panel, obtaining a visible light image, a near-infrared multispectral image and three-dimensional point cloud data, carrying out the fusion processing, and carrying out the calculation of the three-dimensional point cloud data based on the joint criterion of the near-infrared reflectivity difference and the three-dimensional space distribution. Distinguishing the types and pollution degrees of sheltering objects on the surfaces of the photovoltaic panels; according to the shielding object type and the pollution degree, a pollution thermodynamic diagram of the surface of the photovoltaic panel is generated, and a cleaning operation path of the unmanned aerial vehicle is dynamically planned based on the pollution thermodynamic diagram; a self-adaptive cleaning module of the unmanned aerial vehicle is controlled to execute cleaning operation according to the cleaning operation path; after cleaning operation is completed, the surface of the photovoltaic panel is scanned again, data obtained after cleaning is compared with data before cleaning, cleaning parameters are optimized through iterative learning, a detection-action-verification closed loop is formed, and full automation and intelligentization of the photovoltaic panel unmanned aerial vehicle in all links of detection, decision making, cleaning and verification are achieved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to a UAV control method and system that integrates scanning and photovoltaic panel cleaning. Background Technology

[0002] With the rapid development of the photovoltaic industry, the large-scale operation and maintenance of photovoltaic power plants faces enormous challenges. Pollution on the surface of photovoltaic panels (such as dust) and physical obstructions (such as bird droppings, fallen leaves, and shadows) can severely reduce power generation efficiency. Currently, using drones for photovoltaic panel inspection and cleaning has become an emerging technological trend.

[0003] However, existing drone operation and maintenance solutions suffer from significant technological gaps and limitations. Most solutions can only perform a single function: either using visible light cameras for initial inspections, with manual judgment and decision-making upon anomaly detection; or executing pre-defined flight and cleaning paths, lacking intelligent decision-making capabilities based on real-time perception. These solutions separate "detection" and "cleaning" into two independent stages, resulting in the following core flaws in the entire operation and maintenance process:

[0004] Inaccurate perception: It is difficult to effectively distinguish physical obstructions (such as shadows) from uniformly distributed dust using only visible light images, and it is also impossible to accurately determine the three-dimensional shape and type of obstructions.

[0005] Decision-making and execution are disconnected: the detection results cannot directly and automatically drive the cleaning execution unit. The drones cannot dynamically adjust the cleaning mode (such as pneumatic blowing or wet wiping) and operation path according to the type of obstruction and the degree of contamination, resulting in low cleaning efficiency and excessive energy consumption.

[0006] Lack of effectiveness verification: After the cleaning operation is completed, there is no immediate and automated effectiveness evaluation mechanism. Maintenance personnel cannot know whether the cleaning has met the standards, and the system cannot learn from this operation to optimize subsequent performance.

[0007] Therefore, how to provide a drone control method that integrates accurate detection, intelligent decision-making, adaptive execution, and effect verification to form an automated closed loop has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] This invention provides a drone control method and system that integrates scanning and photovoltaic panel cleaning, in order to solve the defects of existing photovoltaic panel drones in which the detection, decision-making, cleaning and verification stages are isolated from each other.

[0009] In a first aspect, the present invention provides a drone control method integrating scanning and photovoltaic panel cleaning, comprising:

[0010] The surface of the photovoltaic panel is scanned to obtain visible light images, near-infrared multispectral images, and three-dimensional point cloud data;

[0011] The visible light image, near-infrared multispectral image and three-dimensional point cloud data are fused and processed. Based on the joint criterion of near-infrared reflectivity difference and three-dimensional spatial distribution, the types of obstructions and the degree of pollution on the surface of the photovoltaic panel are distinguished.

[0012] Based on the type of obstruction and the degree of pollution, a pollution heat map of the photovoltaic panel surface is generated, and the cleaning operation path of the drone is dynamically planned based on the pollution heat map.

[0013] The adaptive cleaning module of the control drone performs cleaning operations according to the cleaning operation path;

[0014] After the cleaning operation is completed, the surface of the photovoltaic panel is scanned again to obtain the data after cleaning and compare it with the data before cleaning. The cleaning parameters are optimized through iterative learning to form a detection-action-verification closed loop.

[0015] According to the present invention, a drone control method integrating scanning and photovoltaic panel cleaning is provided, wherein the method for distinguishing the type of obstruction and the degree of contamination on the surface of the photovoltaic panel based on a joint criterion of near-infrared reflectivity difference and three-dimensional spatial distribution includes:

[0016] Spatiotemporal registration is performed between the two-dimensional image pixels acquired by the multispectral imaging module and the three-dimensional point cloud data acquired by the lidar module to ensure that each image pixel corresponds to three-dimensional spatial coordinates.

[0017] Calculate the average near-infrared reflectance of pixels within the registered region and compare it with the preset clean photovoltaic panel reference reflectance to generate a reflectance difference map;

[0018] Based on the registered 3D point cloud, the height variance map of the corresponding region relative to the photovoltaic panel reference plane is calculated, and the type of obstruction is identified by combining the reflectivity difference map.

[0019] According to the present invention, a drone control method integrating scanning and photovoltaic panel cleaning is provided, wherein the identification of obstruction type includes:

[0020] Areas where the reflectance difference exceeds a preset difference threshold and the absolute value of the height variance is less than a preset height tolerance are identified as dust pollution.

[0021] For areas where the reflectance difference value exceeds the preset difference threshold and the height variance is greater than the preset positive shape threshold, the area is determined to be a vegetation-type obstruction.

[0022] Areas with reflectivity differences exceeding the preset difference threshold and height variance less than the preset negative shape threshold are identified as building shadows.

[0023] According to the UAV control method integrating scanning and photovoltaic panel cleaning provided by the present invention, the preset difference threshold, preset height tolerance, positive shape threshold and negative shape threshold are periodically optimized based on a continuously updated sample database.

[0024] The sample database is constructed by adding successfully classified regional data from historical operations and their corresponding multispectral and point cloud raw data to it. The sample database is then used to retrain and optimize the preset difference threshold, preset height tolerance, positive shape threshold, and negative shape threshold, so that the differentiation method can adapt to the optical and physical characteristics of different regions, seasons, and photovoltaic panel models.

[0025] According to the present invention, a drone control method integrating scanning and photovoltaic panel cleaning is provided, wherein generating a contamination heat map of the photovoltaic panel surface based on the type and degree of obstruction includes:

[0026] Based on the identified types of obstructions and the degree of pollution, a quantified power generation efficiency impact factor is assigned to each identified area;

[0027] Based on the power generation efficiency impact factor and the corresponding regional area, calculate its weight in the overall pollution assessment;

[0028] Based on the weight and spatial location information of each region, a pollution heat map is generated, where different colors represent different levels of cleaning urgency, which is used for dynamic cleaning strategies and route planning.

[0029] According to the present invention, a drone control method integrating scanning and photovoltaic panel cleaning is provided, wherein the dynamic planning of the drone's cleaning operation path based on the pollution heat map includes:

[0030] The pollution heatmap is rasterized, and each raster is assigned a cost weight that is positively correlated with the urgency of cleaning.

[0031] Using the drone's starting position as the root node, a path search algorithm is used to iteratively expand the raster map to find the optimal path sequence that visits all rasters with cost weights higher than a preset threshold.

[0032] According to the present invention, a UAV control method integrating scanning and photovoltaic panel cleaning is provided, wherein the path search algorithm is a reinforcement learning algorithm incorporating energy consumption constraints, and the reward function of the reinforcement learning algorithm is configured as follows:

[0033] Positive rewards are given to high-cost cleaning grids, while negative rewards are given for changes in drone flight distance, turning angle, and cleaning mode switching. The policy network is updated in real time through online learning to achieve a dynamic balance between cleaning efficiency and energy consumption.

[0034] The UAV control method integrating scanning and photovoltaic panel cleaning provided by the present invention further includes:

[0035] The pollution heat map is divided into multiple clean sub-regions;

[0036] Based on the real-time location, remaining battery power, and cleaning capacity of each drone, a cleaning sub-area is allocated to each drone through a distributed auction algorithm;

[0037] Each drone uses the reinforcement learning algorithm to plan a local path within its assigned cleaning sub-area, and each drone shares its progress and remaining battery power through communication during the execution of the task, thereby triggering dynamic task reallocation.

[0038] According to the present invention, a drone control method integrating scanning and photovoltaic panel cleaning is provided, wherein the photovoltaic panel surface is scanned again, the data after cleaning is compared with the data before cleaning, and the cleaning parameters are optimized through iterative learning to form a detection-action-verification closed loop, including:

[0039] After cleaning, the surface of the photovoltaic panel is scanned again to obtain multispectral images and three-dimensional point cloud data after cleaning;

[0040] The data after cleaning is compared with the data before cleaning to quantify and calculate the power generation efficiency recovery rate and the degree of cleaning residue.

[0041] Based on the power generation efficiency recovery rate and cleaning residue, the cleaning parameters for the next working cycle are dynamically adjusted using the gradient descent method. The cleaning parameters include the working pressure of the cleaning head, the flow rate of the cleaning medium, and the moving speed of the robotic arm.

[0042] The optimized cleaning parameters are updated to the system's strategy database for subsequent cleaning operation decisions under the same or similar environmental conditions.

[0043] Secondly, the present invention provides a drone control method integrating scanning and photovoltaic panel cleaning, comprising:

[0044] The acquisition module is used to scan the surface of the photovoltaic panel to acquire visible light images, near-infrared multispectral images, and three-dimensional point cloud data;

[0045] The differentiation module is used to fuse the visible light image, near-infrared multispectral image and three-dimensional point cloud data, and to differentiate the type of obstruction and degree of pollution on the surface of the photovoltaic panel based on the joint criteria of near-infrared reflectance difference and three-dimensional spatial distribution.

[0046] The planning module is used to generate a pollution heat map of the photovoltaic panel surface based on the type and degree of the obstruction, and dynamically plan the cleaning operation path of the drone based on the pollution heat map.

[0047] A cleaning module is used to control the drone's adaptive cleaning module to perform cleaning operations according to the cleaning operation path;

[0048] The verification module is used to scan the surface of the photovoltaic panel again after the cleaning operation is completed, compare the data after cleaning with the data before cleaning, optimize the cleaning parameters through iterative learning, and form a detection-action-verification closed loop.

[0049] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the drone control method for integrated scanning and photovoltaic panel cleaning as described above.

[0050] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the unmanned aerial vehicle control method for integrated scanning and photovoltaic panel cleaning as described above.

[0051] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the unmanned aerial vehicle control method for integrated scanning and photovoltaic panel cleaning as described above.

[0052] Beneficial effects:

[0053] The solution provided by this invention achieves full automation and intelligence in the operation and maintenance process, from data collection to effect verification, without any manual intervention. Precise identification based on multi-sensor fusion and dynamic path planning based on pollution heat maps enable cleaning resources to be accurately deployed to the most needed areas, improving cleaning efficiency. Closed-loop verification of cleaning effects enables iterative learning and self-improvement capabilities, continuously enhancing overall operation and maintenance performance. It ensures that each cleaning task has a clear quality inspection process, fundamentally avoiding problems such as incomplete cleaning and missed areas, guaranteeing that photovoltaic panels can stably recover to a high power generation efficiency state after cleaning, and improving the controllability of operation and maintenance quality. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0055] Figure 1 This is a flowchart illustrating the drone control method for integrated scanning and photovoltaic panel cleaning provided in this embodiment;

[0056] Figure 2 This is a schematic diagram illustrating the principle of the drone control method integrating scanning and photovoltaic panel cleaning provided in this embodiment;

[0057] Figure 3 This is a schematic diagram of the drone hardware configuration provided in this embodiment;

[0058] Figure 4 This is a schematic diagram illustrating the principle of the data fusion and differentiation process provided in this embodiment;

[0059] Figure 5 This is a schematic diagram of the cleaning module provided in this embodiment;

[0060] Figure 6 This is a schematic diagram of the structure of the UAV control system integrating scanning and photovoltaic panel cleaning provided in this embodiment;

[0061] Figure 7 This is a schematic diagram of the structure of the electronic device provided in this embodiment. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0063] Figure 1 This is a flowchart illustrating the drone control method for integrated scanning and photovoltaic panel cleaning provided in this embodiment. Figure 2 This is a schematic diagram illustrating the principle of the drone control method for integrated scanning and photovoltaic panel cleaning provided in this embodiment.

[0064] like Figure 1 and Figure 2 As shown in the figure, the drone control method integrating scanning and photovoltaic panel cleaning provided in this embodiment of the invention mainly includes the following steps:

[0065] 101. Scan the surface of the photovoltaic panel to obtain visible light images, near-infrared multispectral images, and three-dimensional point cloud data.

[0066] Specifically, such as Figure 3 As shown, a gimbal (including a horizontal laser and a multi-source data synchronous acquisition stabilizing gimbal) is installed under the fuselage of the drone, which integrates a visible light camera, a near-infrared multispectral imaging module, and a lidar. The gimbal can counteract the shaking of the drone during flight, ensuring that the sensor's attitude is stable during scanning, and that the acquired data will not become blurred or misaligned due to shaking.

[0067] The drone first flies within a preset initial range, during which three types of sensors work simultaneously. The visible light camera takes a two-dimensional image of the photovoltaic panel surface, which can show whether there is obvious dirt or obstruction on the surface; the near-infrared multispectral imaging module collects near-infrared images of different bands, and distinguishes different substances by spectral characteristics (for example, dust and bird droppings have different spectra); the lidar emits lasers to scan the photovoltaic panel, obtains three-dimensional point cloud data, and can determine the actual size, position, and protrusions and depressions of objects on the surface.

[0068] By simultaneously collecting three types of data—"visible appearance," "spectral properties of matter," and "spatial morphology"—the problem of missing information and inaccurate judgments caused by using only one sensor is avoided.

[0069] 102. The visible light image, near-infrared multispectral image and three-dimensional point cloud data are fused and processed. Based on the joint criterion of near-infrared reflectance difference and three-dimensional spatial distribution, the types of obstructions and the degree of pollution on the surface of photovoltaic panels are distinguished.

[0070] Specifically, such as Figure 4 As shown, the first step is spatiotemporal synchronization registration, aligning data from different sensors. Using the timestamps and location information recorded by the GPS / BeiDou positioning module and attitude sensor on the UAV, each pixel of the near-infrared multispectral image is mapped to the 3D point cloud coordinates of the LiDAR. Simply put, this ensures that a pixel knows both its position on the 2D image and its 3D coordinates in real space, avoiding matching errors caused by data asynchrony.

[0071] For each registered region, two core data points are calculated: Near-infrared reflectance difference: First, the average near-infrared reflectance of this region is calculated, then compared with the standard reflectance of a clean photovoltaic panel stored in the system to obtain the difference value, creating a reflectance difference map. Areas with low reflectance are likely to have shading or contamination. Height variance: Based on the 3D point cloud data from the lidar, the height change of this region relative to the flat photovoltaic panel reference surface is calculated, creating a height variance map. Areas with abnormal height may have protruding obstructions (such as leaves) or depressions (although rare, these can be ruled out).

[0072] Based on the two diagrams above, classify the types according to the rules. If the reflectance difference exceeds the preset difference threshold and the height variance is particularly small (close to flat), less than the height tolerance, it is judged as dust pollution, because dust is uniformly covered and will not have obvious height changes. If the reflectance difference exceeds the preset difference threshold and the height variance is large (with obvious protrusions), greater than the preset positive shape threshold, it is judged as vegetation-type obstructions (such as weeds, fallen leaves, bird droppings), which will protrude from the surface of the photovoltaic panel. If the reflectance difference exceeds the preset difference threshold and the height variance is small but negative, less than the preset negative shape threshold, it is judged as building shadow-type obstructions. Shadows have no solid form and only reduce reflectance without changing the height.

[0073] The preset difference threshold, height tolerance, preset positive shape threshold, and preset negative shape threshold mentioned above are not fixed values. The system builds a sample database, storing successfully classified region data (such as spectral and point cloud data of areas identified as dust) from each operation. This database is used periodically to retrain and adjust the thresholds, allowing the classification method to adapt to different situations, such as accurately distinguishing between snow accumulation in northern winters, moss in rainy southern regions, and the surface characteristics of different brands of photovoltaic panels.

[0074] By integrating multiple data sources and dynamically optimizing criteria, the accuracy of distinguishing between obstructions and contamination is significantly improved, providing a reliable basis for subsequent cleaning decisions.

[0075] 103. Generate a pollution heat map of the photovoltaic panel surface based on the type of obstruction and the degree of pollution, and dynamically plan the cleaning operation path of the drone based on the pollution heat map.

[0076] Specifically, first, assign a cleaning urgency level to each designated area. Then, based on the type of obstruction and the degree of contamination, assign a power generation efficiency impact factor to each area. For example, shadows and thick dust have a greater impact on power generation, resulting in a higher factor; thin dust has a smaller impact, resulting in a lower factor. Next, combine this with the area's size to calculate its weight in the overall photovoltaic panel cleaning task. Areas with larger areas and higher impact factors have the highest weight and the highest cleaning priority.

[0077] By combining the weight and location of each area to create a heat map, different colors are used to represent the urgency of cleaning (e.g., red means "clean immediately," and yellow means "can wait"). This way, both the system and staff can immediately see which area needs cleaning most, making it intuitive and clear.

[0078] The heatmap is transformed into a computable map to find the optimal path. Step 1: Divide the entire photovoltaic panel area into equally sized grids, with each grid's cost weight linked to its urgency (urgent grids have higher cost weights). Step 2: Use a reinforcement learning algorithm with energy consumption constraints to find the path. Starting from the drone's takeoff point (or current location), the algorithm tries different routes, considering two points: Positive reward: Flying to a grid with a high cost weight (i.e., an urgent cleaning area) incentivizes the algorithm to prioritize these areas. Negative reward: A penalty is imposed for excessively long flight distances, frequent turns, or frequent switching of cleaning modes (e.g., from aerodynamic to wet cleaning) to prevent wasted electricity. The algorithm updates its strategy in real time, ultimately finding the optimal path that covers all grids requiring cleaning while minimizing energy consumption.

[0079] By transforming abstract pollution situations into concrete cleaning routes, the problems of missed or excessive cleaning when following fixed routes are solved, allowing cleaning resources to be precisely deployed to where they are most needed, thus improving efficiency and saving energy.

[0080] 104. The adaptive cleaning module of the control drone performs cleaning operations according to the cleaning operation path.

[0081] Specifically, the adaptive cleaning module on the drone, such as Figure 5 As shown, the core component is a retractable robotic arm driven by shape memory alloy, equipped with three cleaning heads (pneumatic, wet, and dry) and a pressure feedback system, as well as a compensation nozzle on the side. The retractable robotic arm can flexibly adjust its length, ensuring a suitable cleaning distance regardless of whether the photovoltaic panel is laid flat or at a slight angle. The three cleaning heads—pneumatic (for blowing away dust), dry (for wiping light stains), and wet (with cleaning media for wiping stubborn stains such as bird droppings)—can be switched as needed. The pressure feedback system measures the contact pressure between the cleaning head and the photovoltaic panel surface in real time, preventing excessive pressure from scratching the panel or insufficient pressure from leaving it unclean. The side nozzle can be used to clean areas missed when detouring around obstacles.

[0082] After the drone flies to the designated cleaning area, the cleaning process is as follows: Select the cleaning head; for example, use the pneumatic head to blow away dust, and switch to the wet head for stubborn bird droppings. Adjust parameters: the pressure feedback system transmits pressure data to the controller. If the pressure is insufficient, slightly increase the cleaning head pressure; if the pressure is about to exceed the limit, decrease it, and simultaneously adjust the robotic arm's movement speed (slower for dirty areas, faster for clean areas). Obstacle avoidance compensation: if the drone encounters a small obstacle during flight, such as a drifting branch, requiring a detour, the side compensation nozzles will immediately activate to catch up on any small areas missed during the detour, ensuring no areas are missed.

[0083] When dealing with large-scale photovoltaic power plants (such as those covering hundreds of acres), multiple drones are needed to work together to improve overall efficiency and avoid problems such as a single drone not being able to complete the task or running out of power. The specific collaboration method is as follows:

[0084] First, the pollution heat map of the entire power plant is divided into multiple sub-regions based on location and pollution density, such as the number of rows of photovoltaic panel arrays. The cleaning workload for each sub-region is roughly the same, preventing any single drone from being overloaded. A distributed auction algorithm is used to assign tasks to each drone, with each drone transmitting its real-time status (location, remaining battery power, and cleaning head availability). Each sub-region is treated as an "auction item," and each drone "bids" based on its status (e.g., drones closer to the site or with sufficient battery power have a higher bid). The system allocates sub-regions according to the bids, ensuring that each drone's tasks match its capabilities. During operation, each drone also shares real-time information on "how much work has been done" and "how much battery power remains." For example, if a drone's battery is low and it hasn't finished its work in a sub-region, the system will immediately assign it to a nearby drone with sufficient battery power. Within its own sub-region, each drone uses reinforcement learning algorithms to plan short paths, ensuring the efficiency of each individual drone.

[0085] By selecting the appropriate cleaning method based on the type of pollution, parameters can be adjusted in real time to protect photovoltaic panels, and emergency avoidance can be handled to ensure cleaning coverage and effectiveness. Multi-machine collaboration for cleaning large-scale power plants can significantly shorten operation and maintenance time and handle single-machine failures, ensuring uninterrupted operation.

[0086] 105. After the cleaning operation is completed, the surface of the photovoltaic panel is scanned again to obtain the data after cleaning and compare it with the data before cleaning. The cleaning parameters are optimized through iterative learning to form a closed loop of detection-action-verification.

[0087] Specifically, whether a single robot cleans a section or multiple robots clean the entire power station, it will fly over again to collect all the data after cleaning. The new data after cleaning is compared area by area with the original data before cleaning to calculate two key indicators: Power generation efficiency recovery rate: This measures how much the near-infrared reflectance of the photovoltaic panels has recovered compared to before cleaning, and the corresponding recovery in power generation efficiency. Cleaning residue: This checks for any remaining uncleaned areas and the impact of residual pollution or obstructions on power generation. These two indicators clearly indicate whether the cleaning has met standards. If a certain type of area is found to be uncleaned (e.g., dust not completely removed by the pneumatic head), the gradient descent method is used to adjust the cleaning parameters for the next cleaning cycle. For example, the pressure of the pneumatic head might be increased, or the robotic arm's movement speed might be slowed down. The optimized parameters are stored in the system's strategy database. The next time a similar situation occurs (e.g., the same dust thickness, the same type of photovoltaic panels), these parameters can be directly adjusted without retrying, resulting in increasingly stable cleaning results.

[0088] Through verification, the entire solution forms a closed loop of detection-cleaning-verification-optimization, ensuring that the photovoltaic panels can stably restore high power generation efficiency after each cleaning.

[0089] The method described in this embodiment achieves full automation and intelligence in the operation and maintenance process, from data acquisition to effect verification, without any manual intervention. Precise identification based on multi-sensor fusion and dynamic path planning based on pollution heat maps enable cleaning resources to be accurately deployed to the most needed areas, improving cleaning efficiency. Closed-loop verification of cleaning effects allows for iterative learning and self-improvement, continuously enhancing overall operation and maintenance performance. It ensures a clear quality inspection process for each cleaning task, fundamentally avoiding incomplete cleaning and missed areas, guaranteeing that photovoltaic panels can stably recover to high power generation efficiency after cleaning, and improving the controllability of operation and maintenance quality.

[0090] Figure 6 This is a schematic diagram of the structure of the drone control system that integrates scanning and photovoltaic panel cleaning provided in this embodiment.

[0091] like Figure 6 As shown in the figure, this embodiment provides a drone control method integrating scanning and photovoltaic panel cleaning, including:

[0092] The acquisition module 601 is used to scan the surface of the photovoltaic panel to acquire visible light images, near-infrared multispectral images and three-dimensional point cloud data;

[0093] The differentiation module 602 is used to fuse visible light images, near-infrared multispectral images and three-dimensional point cloud data, and to differentiate the type of obstruction and degree of pollution on the surface of photovoltaic panels based on the joint criteria of near-infrared reflectance difference and three-dimensional spatial distribution.

[0094] The planning module 603 is used to generate a pollution heat map of the photovoltaic panel surface based on the type of obstruction and the degree of pollution, and to dynamically plan the cleaning operation path of the drone based on the pollution heat map.

[0095] The cleaning module 604 is used to control the drone's adaptive cleaning module to perform cleaning operations according to the cleaning operation path;

[0096] The verification module 605 is used to scan the surface of the photovoltaic panel again after the cleaning operation is completed, compare the data after cleaning with the data before cleaning, and optimize the cleaning parameters through iterative learning to form a detection-action-verification closed loop.

[0097] Figure 7 This is a schematic diagram of the structure of the electronic device provided in this embodiment.

[0098] like Figure 7As shown, the electronic device may include a processor 701, a communications interface 702, a memory 703, and a communication bus 704. The processor 701, communications interface 702, and memory 703 communicate with each other via the communication bus 704. The processor 701 can call logic instructions from the memory 703 to execute a drone control method integrating scanning and photovoltaic panel cleaning.

[0099] Furthermore, the logical instructions in the aforementioned memory 703 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0100] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the UAV control method for integrated scanning and photovoltaic panel cleaning provided by the above methods.

[0101] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the UAV control method for integrated scanning and photovoltaic panel cleaning provided by the methods described above.

[0102] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A drone control method integrating scanning and photovoltaic panel cleaning, characterized in that, include: The surface of the photovoltaic panel is scanned to obtain visible light images, near-infrared multispectral images, and three-dimensional point cloud data; The visible light image, near-infrared multispectral image and three-dimensional point cloud data are fused and processed. Based on the joint criterion of near-infrared reflectivity difference and three-dimensional spatial distribution, the types of obstructions and the degree of pollution on the surface of the photovoltaic panel are distinguished. Based on the type of obstruction and the degree of pollution, a pollution heat map of the photovoltaic panel surface is generated, and the cleaning operation path of the drone is dynamically planned based on the pollution heat map. The adaptive cleaning module of the control drone performs cleaning operations according to the cleaning operation path; After the cleaning operation is completed, the surface of the photovoltaic panel is scanned again to obtain the data after cleaning and compare it with the data before cleaning. The cleaning parameters are optimized through iterative learning to form a detection-action-verification closed loop.

2. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 1, characterized in that, The criteria for distinguishing the type of obstruction and degree of contamination on the surface of photovoltaic panels based on the joint criteria of near-infrared reflectivity difference and three-dimensional spatial distribution include: Spatiotemporal registration is performed between the two-dimensional image pixels acquired by the multispectral imaging module and the three-dimensional point cloud data acquired by the lidar module to ensure that each image pixel corresponds to three-dimensional spatial coordinates. Calculate the average near-infrared reflectance of pixels within the registered region and compare it with the preset clean photovoltaic panel reference reflectance to generate a reflectance difference map; Based on the registered 3D point cloud, the height variance map of the corresponding region relative to the photovoltaic panel reference plane is calculated, and the type of obstruction is identified by combining the reflectivity difference map.

3. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 2, characterized in that, The types of obstructions identified include: Areas where the reflectance difference exceeds a preset difference threshold and the absolute value of the height variance is less than a preset height tolerance are identified as dust pollution. For areas where the reflectance difference value exceeds the preset difference threshold and the height variance is greater than the preset positive shape threshold, the area is determined to be a vegetation-type obstruction. Areas with reflectivity differences exceeding the preset difference threshold and height variance less than the preset negative shape threshold are identified as building shadows.

4. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 3, characterized in that, The preset difference threshold, preset height tolerance, positive shape threshold, and negative shape threshold are periodically optimized based on a continuously updated sample database. The sample database is constructed by adding successfully classified regional data from historical operations and their corresponding multispectral and point cloud raw data to it. The sample database is then used to retrain and optimize the preset difference threshold, preset height tolerance, positive shape threshold, and negative shape threshold, so that the differentiation method can adapt to the optical and physical characteristics of different regions, seasons, and photovoltaic panel models.

5. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 1, characterized in that, The step of generating a pollution heat map of the photovoltaic panel surface based on the type and degree of the obstruction includes: Based on the identified types of obstructions and the degree of pollution, a quantified power generation efficiency impact factor is assigned to each identified area; Based on the power generation efficiency impact factor and the corresponding regional area, calculate its weight in the overall pollution assessment; Based on the weight and spatial location information of each region, a pollution heat map is generated, where different colors represent different levels of cleaning urgency, which is used for dynamic cleaning strategies and route planning.

6. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 5, characterized in that, The method of dynamically planning the cleaning operation path of the UAV based on the pollution heat map includes: The pollution heatmap is rasterized, and each raster is assigned a cost weight that is positively correlated with the urgency of cleaning. Using the drone's starting position as the root node, a path search algorithm is used to iteratively expand the raster map to find the optimal path sequence that visits all rasters with cost weights higher than a preset threshold.

7. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 6, characterized in that, The path search algorithm is a reinforcement learning algorithm that incorporates energy consumption constraints, and the reward function of the reinforcement learning algorithm is configured as follows: Positive rewards are given to high-cost cleaning grids, while negative rewards are given for changes in drone flight distance, turning angle, and cleaning mode switching. The policy network is updated in real time through online learning to achieve a dynamic balance between cleaning efficiency and energy consumption.

8. The UAV control method for integrated scanning and photovoltaic panel cleaning according to claim 7, characterized in that, Also includes: The pollution heat map is divided into multiple clean sub-regions; Based on the real-time location, remaining battery power, and cleaning capacity of each drone, a cleaning sub-area is allocated to each drone through a distributed auction algorithm; Each drone uses the reinforcement learning algorithm to plan a local path within its assigned cleaning sub-area, and each drone shares its progress and remaining battery power through communication during the execution of the task, thereby triggering dynamic task reallocation.

9. The UAV control method for integrated scanning and photovoltaic panel cleaning according to any one of claims 1-8, characterized in that, The process of re-scanning the photovoltaic panel surface, comparing the post-cleaning data with the pre-cleaning data, and optimizing cleaning parameters through iterative learning to form a detection-action-verification closed loop includes: After cleaning, the surface of the photovoltaic panel is scanned again to obtain multispectral images and three-dimensional point cloud data after cleaning; The data after cleaning is compared with the data before cleaning to quantify and calculate the power generation efficiency recovery rate and the degree of cleaning residue. Based on the power generation efficiency recovery rate and cleaning residue, the cleaning parameters for the next working cycle are dynamically adjusted using the gradient descent method. The cleaning parameters include the working pressure of the cleaning head, the flow rate of the cleaning medium, and the moving speed of the robotic arm. The optimized cleaning parameters are updated to the system's strategy database for subsequent cleaning operation decisions under the same or similar environmental conditions.

10. A method for controlling a drone that integrates scanning and photovoltaic panel cleaning, characterized in that, include: The acquisition module is used to scan the surface of the photovoltaic panel to acquire visible light images, near-infrared multispectral images, and three-dimensional point cloud data; The differentiation module is used to fuse the visible light image, near-infrared multispectral image and three-dimensional point cloud data, and to differentiate the type of obstruction and degree of pollution on the surface of the photovoltaic panel based on the joint criteria of near-infrared reflectance difference and three-dimensional spatial distribution. The planning module is used to generate a pollution heat map of the photovoltaic panel surface based on the type and degree of the obstruction, and dynamically plan the cleaning operation path of the drone based on the pollution heat map. A cleaning module is used to control the drone's adaptive cleaning module to perform cleaning operations according to the cleaning operation path; The verification module is used to scan the surface of the photovoltaic panel again after the cleaning operation is completed, compare the data after cleaning with the data before cleaning, optimize the cleaning parameters through iterative learning, and form a detection-action-verification closed loop.

Images