Adaptive cleaning control method and device for photovoltaic module cleaning robot

Abstract

The application relates to a cleaning path technology field of a photovoltaic module cleaning robot, and discloses a self-adaptive cleaning control method and device for the photovoltaic module cleaning robot.The method comprises the following steps: acquiring first image information of a running area of the cleaning robot; identifying the category of foreign matters when the foreign matters exist in the first image information; acquiring second image information of a cleaned area of the cleaning robot; performing a cleanliness score on a specified area in the second image information; and determining a cleaning treatment strategy corresponding to the specified area according to the cleanliness score; wherein when the foreign matters are identified as hard obstacles, the cleaning robot is controlled to adjust a cleaning path so as to bypass the area corresponding to the foreign matters and keep a working state; and the area passed through by the cleaning robot in the bypassing process is marked as the cleaned area. It can be seen that the application can manage the obstacle avoidance process in the overall cleaning coverage range, avoid local missed cleaning caused by obstacle avoidance bypassing, and reduce unnecessary repeated cleaning of the cleaned area.

Description

Technical Field

[0001] This application relates to the field of cleaning path technology for photovoltaic module cleaning robots, and in particular to an adaptive cleaning control method and device for photovoltaic module cleaning robots. Background Technology

[0002] With the continuous expansion of photovoltaic power plant installed capacity, the use of cleaning robots to automatically clean the surface of photovoltaic modules has become a common operation and maintenance method. Existing photovoltaic module cleaning robots generally include a walking mechanism, a roller brush assembly, and a spray assembly, moving back and forth on the photovoltaic glass surface according to a preset path to remove dust, bird droppings, leaves, and other adhering substances. Some solutions also incorporate cameras or detection devices on the cleaning robot to identify foreign objects in front, avoid obstacles, or assess the level of dirt on the photovoltaic glass surface, thereby improving the automation level of the cleaning operation.

[0003] However, existing technologies still lack a good coordination between foreign object recognition, path adjustment, and cleaning effect judgment, making it difficult to simultaneously achieve obstacle avoidance, cleaning coverage, and cleaning effect. In particular, when there are different types of foreign objects on the surface of photovoltaic modules and the cleaning robot needs to clean continuously along a preset path, problems such as unreasonable path processing, missed or repeated cleaning of local areas, and difficulty in timely judgment of whether the requirements have been met after cleaning are likely to occur, thus affecting cleaning efficiency and actual use effect.

[0004] It should be noted that the above content is not necessarily prior art, nor is it intended to limit the scope of patent protection of this application. Summary of the Invention

[0005] This application provides an adaptive cleaning control method and device for a photovoltaic module cleaning robot to address the lack of good coordination between foreign object recognition, path adjustment, and cleaning effect judgment in existing photovoltaic module cleaning robots, making it difficult to simultaneously achieve obstacle avoidance, cleaning coverage, and cleaning effect. In particular, when there are different types of foreign objects on the surface of photovoltaic modules and the cleaning robot needs to continuously clean along a preset path, problems such as unreasonable path processing, missed or repeated cleaning of local areas, and difficulty in timely judgment of whether the requirements have been met after cleaning may occur.

[0006] The first aspect of this application discloses an adaptive cleaning control method for a photovoltaic module cleaning robot, the method comprising: Acquire first image information of the area where the cleaning robot is operating; Based on a pre-established foreign object recognition model, the first image information is identified to determine the category of the foreign object when it is present in the first image information. When a foreign object is identified as a hard obstacle, the cleaning robot is controlled to adjust its cleaning path based on a pre-established obstacle avoidance model in order to bypass the area corresponding to the foreign object. When a foreign object is identified as a cleanable contaminant, the cleaning robot is controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. Acquire second image information of the area already cleaned by the cleaning robot; Based on a pre-established cleanliness scoring model, a specified area in the second image information is analyzed to obtain the cleanliness score corresponding to the specified area. Based on a pre-established cleaning strategy model, the cleaning strategy for a specified area is determined according to the cleanliness score of the specified area. The process of controlling the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, based on a pre-established obstacle avoidance model, when the foreign object is identified as a hard obstacle, also includes: Control the cleaning components of the cleaning robot to keep them in working order; The areas traversed by the cleaning robot during its detour are marked as cleaned areas.

[0007] Optionally, after identifying the first image information based on a pre-established foreign object recognition model to identify the category of the foreign object when it is present in the first image information, the method further includes: Determine the height of the foreign object; Determine whether the height of the foreign object is greater than or equal to a preset height threshold; If the height of the foreign object is greater than or equal to the preset height threshold, the foreign object is identified as a hard obstacle, and the cleaning robot is controlled to adjust the cleaning path to bypass the area corresponding to the foreign object. If the height of the foreign object is less than the preset height threshold, the cleaning robot will be controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. Foreign objects with a height greater than or equal to a preset height threshold are all identified as hard obstacles. Even if the foreign object recognition model identifies the foreign object as a cleanable contaminant, it is still forcibly identified as a hard obstacle to prioritize the safety of the cleaning robot.

[0008] Optionally, after the process of marking the area traversed by the cleaning robot during its detour as a cleaned area, the method further includes: When the cleaning robot moves to the already cleaned area during subsequent cleaning, it obtains a cleanliness score for the cleaned area. Based on the cleanliness score within the specified range, determine whether to perform supplementary cleaning on the already cleaned areas.

[0009] Optionally, the process of analyzing a specified area in the second image information based on a pre-established cleanliness scoring model to obtain the cleanliness score corresponding to the specified area includes: The first sub-score is determined based on the reflectivity of the specified area; The second sub-score is determined based on the grayscale values ​​of a specified region; Input the image corresponding to the specified region into a pre-trained image classification model to obtain the third sub-score; The first, second, and third sub-scores are weighted and fused to obtain the cleanliness score.

[0010] Optionally, the process of determining the cleaning strategy for a specified area based on a pre-established cleaning strategy model and a cleanliness score includes: When the cleanliness score is less than the first threshold, the cleaning strategy is determined to be an alarm strategy, which is used to output a manual intervention cleaning prompt. When the cleanliness score is greater than or equal to the first threshold and less than the second threshold, the cleaning strategy is determined to be the re-sweeping strategy, which is used to perform re-sweeping on the specified area. When the cleanliness score is greater than or equal to the second threshold, the cleaning strategy is determined to be the normal pass strategy. The normal pass strategy is used to control the cleaning robot to continue cleaning along the cleaning path.

[0011] Optionally, the process of determining the cleaning strategy as a re-sweeping strategy includes: Perform a rescan on the designated area and re-obtain the cleanliness score for the designated area after the rescan; If the cleanliness score is still less than the second threshold after the rescan, and the current number of rescans is less than the preset maximum number of rescans, then the designated area will be rescanned again. If the current number of rescans reaches the preset limit and the cleanliness score is still less than the second threshold, the cleaning strategy for the specified area will be switched from the rescan strategy to the alarm strategy.

[0012] Optionally, the process of controlling the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, based on a pre-established obstacle avoidance model, when the foreign object is identified as a hard obstacle, includes: A virtual obstacle zone is generated by extending a preset safety buffer distance outward from the boundary of the area corresponding to the foreign object. Divide the path segment that intersects with the virtual obstacle area in the current bow-shaped cleaning path into the obstacle front segment and the obstacle rear segment; After controlling the cleaning robot to execute the first part of the obstacle, it travels along the boundary of the virtual obstacle area to the corresponding position of the second part of the obstacle and continues to execute the second part of the obstacle. After the current cleaning cycle's bow-shaped cleaning path is completed, a supplementary cleaning path is generated for the uncovered areas around the virtual obstacle zone, and supplementary cleaning is performed.

[0013] A second aspect of this application discloses an adaptive cleaning control device for a photovoltaic module cleaning robot, the device comprising: The image acquisition module is used to acquire first image information of the area where the cleaning robot is operating, and second image information of the area that the cleaning robot has cleaned. The foreign object recognition module is used to identify the first image information based on a pre-established foreign object recognition model, so as to identify the category of foreign object when foreign object is present in the first image information; The path control module is used to control the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object when the foreign object is identified as a hard obstacle, based on a pre-established obstacle avoidance model. It is also used to control the cleaning robot to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning when the foreign object is identified as a cleanable contaminant. The cleanliness assessment module is used to analyze a specified area in the second image information based on a pre-established cleanliness scoring model, and obtain the cleanliness score corresponding to the specified area. The strategy determination module is used to determine the cleaning strategy for a specified area based on a pre-established cleaning strategy model and the cleanliness score of the specified area.

[0014] A third aspect of this application discloses another adaptive cleaning control device for a photovoltaic module cleaning robot, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the adaptive cleaning control method for a photovoltaic module cleaning robot as described in any of the first aspects of the present invention.

[0015] A fourth aspect of this application discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute the adaptive cleaning control method for a photovoltaic module cleaning robot as described in any of the first aspects of this invention.

[0016] The embodiments of this application employing the above-described technical solution may have the following advantages: Implementing this invention allows for different approaches to handling foreign objects, such as detours or direct cleaning, avoiding the problems of unreasonable path processing and excessive path deflection caused by treating all detected foreign objects as detours in existing technologies. This establishes a clear correspondence between foreign object identification and path adjustment. Simultaneously, while the foreign object is identified as a hard obstacle and detours are performed, the cleaning robot's cleaning components remain operational, marking the areas covered during the detour as cleaned areas. This integrates the obstacle avoidance process into the overall cleaning coverage, preventing missed areas due to obstacle avoidance and reducing unnecessary re-cleaning of already cleaned areas. Furthermore, by acquiring second image information of the cleaned area and determining the corresponding cleaning processing strategy based on the cleanliness scoring model and cleaning strategy model, the cleaning effect judgment is based on the objective evaluation of the actual image data after cleaning. It can adaptively select the subsequent processing method according to the evaluation results, thereby forming a closed loop between foreign object recognition, path adjustment and cleaning effect judgment. This helps to balance obstacle avoidance, cleaning coverage and cleaning effect, and improve the overall cleaning efficiency and actual use effect. Attached Figure Description

[0017] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0018] Figure 1 A flowchart illustrating an adaptive cleaning control method for a photovoltaic module cleaning robot provided in an embodiment of this application; Figure 2 The differences provided for the embodiments of this application are Figure 1 A flowchart illustrating another adaptive cleaning control method for photovoltaic module cleaning robots; Figure 3 A schematic diagram of the preset cleaning path of a cleaning robot on a photovoltaic module, which applies the adaptive cleaning control method provided in the embodiments of this application. Figure 4 for Figure 3 A schematic diagram showing the tilt of the provided cleaning robot along the preset cleaning path of the photovoltaic module; Figure 5 To distinguish from Figure 3 Another preset cleaning path planar diagram; Figure 6 for Figure 5 A schematic diagram showing the tilt of the provided cleaning robot along the preset cleaning path of the photovoltaic module; Figure 7 for Figure 3A schematic diagram showing the inclined cleaning path of the cleaning robot as it navigates around the photovoltaic modules; Figure 8 This is a schematic diagram of the structure of an adaptive cleaning control device for a photovoltaic module cleaning robot provided in an embodiment of this application; Figure 9 The differences provided for the embodiments of this application are Figure 3 A schematic diagram of another adaptive cleaning control device for photovoltaic module cleaning robots.

[0019] Explanation of reference numerals in the attached figures: 301 - Image Acquisition Module; 302 - Foreign Object Recognition Module; 303 - Path Control Module; 304 - Cleanliness Assessment Module; 305 - Strategy Determination Module; 401 - Memory; 402 - Processor. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other. The application will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] In this application, the term "numerical interval" (i.e., numerical range) refers to a range of values. Unless otherwise specified, the distribution of selectable values ​​within this numerical interval is considered continuous, and includes the two endpoints (i.e., the minimum and maximum values) of the interval, as well as every value between these endpoints. Unless otherwise specified, when a numerical interval refers only to integers within that interval, it includes the two endpoints of the range and every integer between them, effectively listing every integer. When multiple numerical ranges are provided to describe features or characteristics, these ranges can be merged. In other words, unless otherwise specified, the numerical ranges disclosed in this application should be understood to include any and all subranges included therein. The "numerical value" in this numerical interval can be any quantitative value, such as a number, percentage, or proportion. The term "numerical interval" can broadly include percentage intervals, proportion intervals, ratio intervals, and other quantitative intervals.

[0023] Exemplary embodiments according to this application will now be described in more detail with reference to the accompanying drawings. It should be understood that these exemplary embodiments may be implemented in many different forms and should not be construed as being limited to the embodiments set forth herein. Example 1

[0024] Please see Figure 1 , Figure 1 This is a flowchart illustrating an adaptive cleaning control method for a photovoltaic module cleaning robot disclosed in an embodiment of the present invention. Figure 1 As shown, the adaptive cleaning control method may include the following steps:

[0025] 101. Obtain the first image information of the cleaning robot's operating area.

[0026] In this embodiment, the first image information can be acquired by an image acquisition device located at the front and / or top of the cleaning robot, used to capture scene images in front of the cleaning robot's running direction. The image acquisition device can continuously acquire images of the running area and transmit the acquired images to the control system in real time for analysis and processing. By acquiring image information of the area in front of the cleaning robot, a timely visual data foundation can be provided for subsequent foreign object detection and classification.

[0027] 102. Based on a pre-established foreign object recognition model, the first image information is identified to identify the category of the foreign object when it is present in the first image information.

[0028] In this embodiment, the foreign object recognition model can be a pre-trained visual recognition model, such as a recognition model built based on an object detection network, image segmentation network, or classification network. The foreign object recognition model is obtained by training on a large number of sample images of various foreign objects on the surface of the photovoltaic module. Specifically, the foreign object recognition model analyzes the first image information, and when a foreign object is detected in the image, it further identifies the category of the foreign object.

[0029] In some implementations, the categories of foreign objects include at least hard obstacles and cleanable contaminants. Hard obstacles can refer to objects that cannot be directly removed by the cleaning components of the cleaning robot and may affect the safe passage of the cleaning robot, such as junction boxes, bracket connectors, jumper cables, bolt protrusions, etc. Cleanable contaminants can refer to contaminants that can be removed by the cleaning components, such as bird droppings, leaves, pollen clumps, sand piles, etc.

[0030] 103. When a foreign object is identified as a hard obstacle, the cleaning robot is controlled to adjust its cleaning path based on a pre-established obstacle avoidance model in order to bypass the area corresponding to the foreign object.

[0031] In this embodiment, when the foreign object identification result indicates that the object ahead is a hard obstacle, the control system calls the obstacle avoidance model to adjust the current cleaning path, allowing the cleaning robot to bypass the area corresponding to the hard obstacle. The obstacle avoidance model can generate a local detour path suitable for the current operating scenario based on factors such as the obstacle's position, outline range, the cleaning robot's body size, current motion state, and preset safety distance. By performing path adjustment after identifying hard obstacles, the risk of collision between the cleaning robot and hard structures can be reduced, avoiding damage to the cleaning robot body, cleaning components, and photovoltaic module surfaces, while also helping to maintain the continuity and safety of the cleaning operation.

[0032] 104. When a foreign object is identified as a cleanable contaminant, control the cleaning robot to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning.

[0033] In this embodiment, when a foreign object is identified as a cleanable contaminant, the control system does not trigger obstacle avoidance path adjustment. Instead, it controls the cleaning robot to continue running along the current cleaning path and performs cleaning on the area containing the contaminant using the cleaning components. Since cleanable contaminants can be removed by brushes, scraping, spraying, or other cleaning methods, they do not need to be treated as obstacles that must be avoided. By directly passing through and cleaning cleanable contaminants, unnecessary path detours can be reduced, cleaning efficiency can be improved, and overall cleaning coverage can be ensured.

[0034] 105. Obtain second image information of the area already cleaned by the cleaning robot.

[0035] In this embodiment, the second image information can be acquired by an image acquisition device facing the cleaned area, used to capture images of the photovoltaic module surface after the cleaning robot has passed. This second image information provides the image data basis for subsequent cleaning effect evaluation, enabling the judgment of cleaning quality to be based on objective image information.

[0036] 106. Based on the pre-established cleanliness scoring model, analyze the specified area in the second image information to obtain the cleanliness score corresponding to the specified area.

[0037] In this embodiment, the cleanliness scoring model is used to quantitatively analyze the second image information of a specified area and output a numerical cleanliness score. The cleanliness score can be represented as a score value within a certain numerical range, such as 0% to 100%, with a higher score indicating a better cleaning effect. By quantifying the cleaning effect into a comparable score value, a basis can be provided for the automatic decision-making of subsequent cleaning strategies.

[0038] 107. Based on the pre-established cleaning strategy model, determine the cleaning treatment strategy for the specified area according to the cleanliness score of the specified area.

[0039] In this embodiment, the cleaning strategy model automatically selects the corresponding cleaning strategy based on the cleanliness score, thereby achieving closed-loop control from the detection and classification of foreign objects ahead, path adjustment, to the evaluation of the cleaning effect of the cleaned area, and then to the determination of the subsequent processing strategy. The cleaning strategy may include continuing along the original cleaning path, performing re-sweeping on a designated area, and outputting manual intervention cleaning prompts, etc.

[0040] This shows that, Figure 1 In the illustrated embodiment, by linking the identification results of foreign objects in front of the operating area with the cleaning path adjustment, different types of foreign objects can be triggered to either be bypassed or directly cleaned. This avoids the problem of excessive path deflection caused by treating all detected foreign objects as bypass targets. Simultaneously, by scoring the cleanliness of the cleaned area and determining the cleaning strategy accordingly, automatic quantitative evaluation and adaptive decision-making of the cleaning effect can be achieved. This allows the cleaning robot to automatically take appropriate measures based on the actual cleaning effect, thereby helping to improve cleaning coverage, cleaning quality, and overall cleaning efficiency.

[0041] In an optional embodiment, before step 103, i.e. after identifying the foreign object and determining its category, the method may further include: determining the height of the foreign object; determining whether the height of the foreign object is greater than or equal to a preset height threshold; if the height of the foreign object is greater than or equal to the preset height threshold, then the foreign object is classified as a hard obstacle, and the step of controlling the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object is executed; if the height of the foreign object is less than the preset height threshold, then the cleaning robot is controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. Here, foreign objects with a height greater than or equal to the preset height threshold are all classified as hard obstacles. Even if the foreign object recognition model identifies the foreign object as a cleanable contaminant, it is still forcibly classified as a hard obstacle to prioritize the safety of the cleaning robot.

[0042] In this embodiment, the preset height threshold can be set based on the ground clearance of the cleaning robot's chassis, the installation height of the cleaning components, the structural parameters of the motion mechanism, or actual operating experience. For example, the preset height threshold can be set to a value between 3cm and 6cm. By introducing a unified height judgment mechanism, height judgment is performed on all identified foreign objects: for foreign objects with a height less than the preset height threshold, the cleaning robot can pass directly and perform cleaning, thereby reducing unnecessary path deflection; for foreign objects with a height greater than or equal to the preset height threshold, regardless of their identification category, they are determined to be hard obstacles and detours are performed. Since cleanable contaminants are usually low in height and generally do not exceed the preset height threshold; if a foreign object is identified as a cleanable contaminant but its height is greater than or equal to the preset height threshold, it is considered a misjudgment and is forcibly determined to be a hard obstacle and detours are performed, so as to prioritize the safety of the cleaning robot and avoid collisions or interference caused by the cleaning robot passing directly through it.

[0043] Therefore, this optional embodiment can reduce unnecessary path deflection by introducing a unified height judgment mechanism, using height as the core basis for judging hard obstacles, and only performing detour processing on foreign objects whose height reaches a preset threshold, and forcibly reclassifying foreign objects that are identified as cleanable contaminants but whose height exceeds the threshold as hard obstacles, thereby ensuring equipment safety.

[0044] In this optional embodiment, as an alternative implementation, the method may further include: controlling the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, and keeping the cleaning components of the cleaning robot in working order; and marking the area traversed by the cleaning robot during the detour as a cleaned area based on the detour trajectory and the effective coverage width of the cleaning components.

[0045] In this embodiment, the cleaning component may include at least one of a brush component and a spray component. During the detour, the cleaning component does not stop working due to obstacle avoidance actions; that is, the surface of the photovoltaic modules on the detour path segment can still be cleaned normally. Simultaneously, the control system records the detour trajectory and, combined with the effective coverage width of the cleaning component, determines the area actually cleaned during the detour and marks this area as a cleaned area. By marking this area, after the cleaning robot completes the detour and returns to the normal cleaning path, subsequent path planning can identify the cleaned area and adjust the subsequent processing method accordingly, avoiding unnecessary repeated cleaning of areas that are considered uncleaned.

[0046] Therefore, this optional implementation can continue the cleaning operation during obstacle avoidance and detour, and mark the areas already covered during the detour, which helps to reduce unnecessary repeated cleaning while ensuring cleaning coverage.

[0047] In this embodiment, as a further implementation, after marking the areas traversed by the cleaning robot during its detour as cleaned areas, the method may further include: When the cleaning robot moves to the already cleaned area during the subsequent cleaning process, it obtains the cleanliness score of the already cleaned area. Based on the cleanliness score within the specified range, determine whether to perform supplementary cleaning on the already cleaned areas.

[0048] In this embodiment, after the cleaning robot completes its detour and returns to the normal cleaning path, the subsequent arc-shaped path segment may spatially overlap with the detour trajectory. When the cleaning robot moves to the marked area covered by the detour trajectory, the second camera captures an image of that area, and the cleanliness scoring model calculates a cleanliness score. The control system determines whether supplementary cleaning of the area is needed based on the score range of the cleanliness score. Specifically, if the cleanliness score of the area has reached the standard level for normal cleaning, the cleaning robot can pass through directly without performing any cleaning action, avoiding invalid repeated cleaning of already compliant areas; if the cleanliness score of the area is in the re-cleaning range, the area is re-cleaned; if the cleanliness score of the area is in the alarm range, a manual intervention cleaning prompt is output. Through the above method, the subsequent path planning module can automatically identify the cleaned markings on the detour trajectory and determine the subsequent processing method based on the actual cleaning effect of the area. This avoids the detour path segment becoming a cleaning blind spot and also avoids unnecessary repeated cleaning of already compliant areas, which would aggravate the wear and tear on the photovoltaic glass surface.

[0049] Therefore, this further implementation method can combine the status marking of the detour trajectory with the cleanliness assessment and strategy decision in the subsequent path planning, so that the detour path segment is automatically identified in the subsequent cleaning process and treated differently according to the actual cleaning effect, which is beneficial to reduce invalid repeated cleaning while ensuring cleaning coverage.

[0050] In another optional embodiment, step 106, which involves analyzing a designated area in the second image information based on a pre-established cleanliness scoring model to obtain a cleanliness score, may specifically include: determining a first sub-score based on the surface reflection features of the designated area; determining a second sub-score based on the grayscale distribution features of the designated area; inputting the image corresponding to the designated area into a pre-trained image classification model to obtain a third sub-score; and performing weighted fusion of the first, second, and third sub-scores to obtain the cleanliness score.

[0051] In this embodiment, the first sub-score can be determined based on the surface reflectance characteristics of a specified area, and the mapping relationship between surface reflectance characteristics and surface cleanliness can be obtained through pre-calibration. The second sub-score can be determined based on the grayscale distribution characteristics of a specified area, and the relationship between grayscale distribution characteristics and surface cleanliness can be established through sample training or calibration. The third sub-score can be obtained by inputting an image of a specified area into a pre-trained image classification model, which can be obtained by training on photovoltaic module surface image samples of different cleanliness levels. These multiple evaluation dimensions assess cleanliness from different perspectives, and then a comprehensive score is obtained through weighted fusion. This helps to compensate for the inaccuracy of a single evaluation dimension under specific lighting conditions or specific types of contaminants, thereby improving the accuracy and reliability of the cleanliness score.

[0052] In some implementations, the cleanliness scoring model may also be used to comprehensively evaluate at least one of edge texture clarity, color deviation, proportion of local residual contaminants, and illumination normalization characteristics, but this embodiment of the invention does not limit this.

[0053] Therefore, this optional embodiment can quantify the cleaning effect of the cleaned area through multi-dimensional feature fusion, which helps to maintain a relatively stable cleanliness assessment effect under different lighting conditions and different types of pollution.

[0054] In another optional embodiment, step 107, which involves determining the cleaning strategy based on a pre-established cleaning strategy model and a cleanliness score, may specifically include: when the cleanliness score is less than a first threshold, determining the cleaning strategy as an alarm strategy, which is used to output a manual intervention cleaning prompt; when the cleanliness score is greater than or equal to the first threshold and less than a second threshold, determining the cleaning strategy as a re-sweeping strategy, which is used to perform re-sweeping on a specified area; and when the cleanliness score is greater than or equal to the second threshold, determining the cleaning strategy as a normal passage strategy, which is used to control the cleaning robot to continue cleaning along the cleaning path.

[0055] In this embodiment, the first threshold and the second threshold are preset parameters, and the first threshold is less than the second threshold. For example, the first threshold can be set to 30%, the second threshold can be set to 70%, and they can also be adjusted according to on-site operation and maintenance requirements. By dividing the cleanliness score into different intervals and corresponding to different cleaning treatment strategies, the control system can automatically select the appropriate subsequent processing method based on the quality of the cleaning effect: when the cleaning effect is seriously substandard, an alarm message is directly output to prompt manual intervention; when the cleaning effect is at an intermediate level, a re-sweeping is performed to try to further improve the cleanliness; when the cleaning effect meets the requirements, subsequent cleaning tasks continue to be executed. Through the above-mentioned grading strategy, it is helpful to achieve differentiated treatment for areas with different cleaning effects.

[0056] In this optional embodiment, as an optional implementation method, when the cleaning processing strategy is determined to be a re-sweeping strategy, the method may further include: performing a re-sweeping on the designated area, and re-acquiring the cleanliness score corresponding to the designated area after the re-sweeping; if the cleanliness score after the re-sweeping is still less than the second threshold, and the current number of re-sweeping times is less than the preset upper limit, then performing a re-sweeping on the designated area again; if the current number of re-sweeping times reaches the preset upper limit, and the cleanliness score is still less than the second threshold, then switching the cleaning processing strategy corresponding to the designated area from the re-sweeping strategy to the alarm strategy.

[0057] In this embodiment, the preset maximum number of cleaning cycles can be set according to the abrasion resistance of the photovoltaic module surface, cleaning efficiency requirements, and equipment operation strategies. For example, it can be set as an integer between 2 and 5 times. By setting a maximum number of re-sweeping cycles, the system can limit the number of repeated cleanings of the same area while attempting to improve the cleaning effect through re-sweeping, thus avoiding unnecessary wear on the photovoltaic module surface due to excessive mechanical friction. When the cleaning effect still does not meet the requirements after the maximum number of re-sweeping cycles has been reached, the system can automatically switch to an alarm strategy to prompt maintenance personnel to take further action.

[0058] Therefore, this optional implementation can introduce a maximum number of sweeps control mechanism when executing the resweeping strategy, which helps to achieve a balance between improving cleaning quality and controlling over-sweeping.

[0059] In another optional embodiment, the process of controlling the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object in step 103 may specifically include: expanding a preset safety buffer distance outward based on the boundary of the area corresponding to the foreign object to generate a virtual obstacle zone; dividing the path segment in the current bow-shaped cleaning path that intersects with the virtual obstacle zone into a front obstacle segment and a rear obstacle segment; controlling the cleaning robot to execute the front obstacle segment, then detouring along the boundary of the virtual obstacle zone to the corresponding position of the rear obstacle segment, and continuing to execute the rear obstacle segment; after the bow-shaped cleaning path of the current cleaning cycle is completed, generating a supplementary cleaning path for the uncovered area around the virtual obstacle zone and performing supplementary cleaning.

[0060] In this embodiment, the preset safety buffer distance can be set according to the cleaning robot's dimensions, positioning error, turning radius, and operational safety requirements, for example, a value between 8cm and 25cm. The bow-shaped cleaning path can be understood as a reciprocating coverage cleaning path, which may also manifest as an S-shaped or Z-shaped path distribution in some scenarios. By generating a virtual obstacle zone based on the area corresponding to the foreign object, and performing local cutting and detour processing on the current path segment intersecting with the virtual obstacle zone, path adjustments can be made only in the local area where the obstacle is located while maintaining the overall cleaning path framework, thereby reducing the computational overhead of path replanning. After all path segments have been executed, supplementary cleaning is performed on the uncovered areas around the obstacle caused by detours, helping to ensure cleaning coverage.

[0061] In some implementations, if a hard obstacle at the same location is continuously detected in multiple cleaning cycles, the location and boundary information of the corresponding virtual obstacle area can be stored as permanent obstacle area information. In subsequent cleaning cycles, the path planning module directly retrieves the corresponding information from this permanent obstacle area information for path pre-planning, eliminating the need to rely on the real-time detection results of the image acquisition device to trigger obstacle avoidance decisions each time. By persistently storing fixed obstacle information, the computational overhead caused by repeated detection and re-judgment can be reduced, improving path planning efficiency.

[0062] Therefore, this optional embodiment can achieve relatively efficient path adjustment when encountering hard obstacles by delineating virtual obstacle zones and recombining local paths, while also taking into account the cleaning coverage effect of the area surrounding the obstacle.

[0063] For example, please refer to the following: Figure 3 and Figure 4 , Figure 3 This is a plan view of a cleaning robot using the adaptive cleaning control method provided in the embodiments of this application, showing a preset cleaning path on a photovoltaic module. Figure 4 This is a schematic diagram of the tilt state of the preset cleaning path of the cleaning robot on the photovoltaic module.

[0064] Those skilled in the art will understand that the irradiance is highest and the power generation of a photovoltaic module is greatest when sunlight shines perpendicularly onto the module's surface. Therefore, in practical applications, photovoltaic modules are typically placed at a certain tilt angle. Figure 4 As shown, the cleaning robot 100 moves along a preset cleaning path on the tilted photovoltaic module. The cleaning robot 100 starts from one end of a short side below the photovoltaic module, moves upward along the tilted surface to the opposite side of the other end of the short side, then shifts laterally by a cleaning width, and then moves downward along the tilted surface to the vicinity of the lower short side. This process is repeated, and the overall cleaning path is distributed in a bow shape, gradually covering the entire surface of the photovoltaic module.

[0065] Furthermore, considering the tilted arrangement of the photovoltaic modules, during the upward cleaning process along the tilted surface, dust, wastewater, and other debris swept up by the modules may slide down the tilted surface due to gravity, causing secondary pollution to the already cleaned area. This is especially true in the area near the lower short side traversed during the initial climbing phase of the cleaning robot 100, as this area is the lowest and more susceptible to the accumulation of debris sliding down from above. Therefore, in some implementations, methods such as... Figure 5 and Figure 6 The preset cleaning path shown is based on the bow-shaped cleaning path, with subsequent cleaning path segments partially overlapping and covering the path segments traversed by the cleaning robot 100 during its initial uphill phase. This path design allows for the subsequent cleaning process to re-clean the area below that has been secondary contaminated due to sliding down the slope, helping to ensure the overall cleanliness of the tilted photovoltaic module surface.

[0066] Please refer to the following: Figure 7 , Figure 7 This is a diagram illustrating the cleaning path adjustment of the cleaning robot 100 when it encounters a hard obstacle. For example... Figure 7As shown, area A is the region containing a hard obstacle on the surface of the photovoltaic module, and the height of this hard obstacle is greater than or equal to a preset height threshold. Based on the obstacle avoidance model described in the aforementioned embodiment, the control system controls the cleaning robot 100 to adjust its cleaning path to bypass area A. During the bypass process, the cleaning components remain operational, and the area covered by the bypass trajectory is marked as a cleaned area. Simultaneously, the control system records the position information of area A and adjusts the subsequent bow-shaped cleaning path accordingly. During subsequent cleaning, when the cleaning robot 100 moves along the bow-shaped cleaning path to a marked area near area A, the control system obtains the cleanliness score of the marked area and determines whether supplementary cleaning of that area is necessary based on the score range within which the cleanliness score falls. Specifically, if the cleanliness score of the area has reached the standard level for normal cleaning, the cleaning robot 100 will avoid the area when it approaches it along the bow-shaped cleaning path and will not perform cleaning actions to avoid invalid repeated cleaning of the already qualified area; if the cleanliness score of the area is in the re-cleaning range, the area will be re-cleaned; if the cleanliness score of the area is in the alarm range, a manual intervention cleaning prompt message will be output.

[0067] Therefore, by combining the above-mentioned path planning and obstacle handling methods with the actual working conditions of photovoltaic modules being placed at an angle, the overlapping coverage path design can reduce the problem of secondary pollution caused by sliding on the slope. When encountering hard obstacles, the subsequent cleaning behavior can be adaptively adjusted according to the actual cleaning effect of the area covered by the detour trajectory, avoiding ineffective repeated cleaning of areas that have already met the standards. This helps to improve the overall cleaning efficiency while ensuring cleaning coverage. Example 2

[0068] Please see Figure 2 , Figure 2 This is a flowchart illustrating another adaptive cleaning control method for a photovoltaic module cleaning robot disclosed in an embodiment of the present invention. Embodiment two, based on embodiment one, combines several optional technical features as part of the main process to form a more complete adaptive cleaning control scheme. For example... Figure 2 As shown, the method may include the following steps:

[0069] 201. Obtain the first image information of the cleaning robot's operating area.

[0070] In this embodiment, step 201 can be the same as step 101 in embodiment one, that is, by using an image acquisition device located at the front, top, or other suitable position for acquiring the scene in front of the cleaning robot, image information of the front of the cleaning robot in its running direction can be acquired. By continuously acquiring images of the scene in front, data support can be provided for foreign object detection, identification, and subsequent path adjustment.

[0071] 202. Based on a pre-established foreign object recognition model, the first image information is identified to identify the category of the foreign object when it is present in the first image information.

[0072] In this embodiment, step 202 utilizes a foreign object recognition model to detect and classify foreign objects in the first image information, distinguishing between hard obstacles and cleanable contaminants. The foreign object recognition model is a pre-trained visual recognition model, which can be constructed based on a target detection network architecture and trained by annotating a large number of image samples of various foreign objects on the surface of photovoltaic modules. Specifically, the foreign object recognition model performs real-time analysis of the first image information. When a foreign object is detected in the image, it further classifies and identifies the foreign object to determine whether it is a hard obstacle or a cleanable contaminant. By identifying the category of the foreign object, a basis for judgment can be provided for subsequent detour or direct cleaning.

[0073] 203. After identifying the foreign object and determining its category, determine the height of the foreign object and judge whether the height of the foreign object is greater than or equal to the preset height threshold.

[0074] In this embodiment, the height of the foreign object can be estimated using monocular depth estimation, binocular stereo vision, structured light ranging, laser ranging, or other ranging methods. By comparing the height of the foreign object with a preset height threshold, it can be further determined whether to perform detour processing: for foreign objects with a height greater than or equal to the preset height threshold, they are identified as hard obstacles and detour processing is performed; for foreign objects with a height less than the preset height threshold, they are passed directly and cleaning is performed. This step helps avoid unnecessary detours for foreign objects that are low in height and do not affect the normal passage of the robot, thereby reducing path deflection.

[0075] 204. If the height of the foreign object is greater than or equal to the preset height threshold, the foreign object is identified as a hard obstacle. Based on the pre-established obstacle avoidance model, the cleaning robot is controlled to adjust the cleaning path to bypass the area corresponding to the foreign object. During the detour, the cleaning components of the cleaning robot are controlled to remain in working state. Based on the detour trajectory and the effective coverage width of the cleaning components, the area passed by the cleaning robot during the detour is marked as the cleaned area.

[0076] In this embodiment, when the height of a foreign object reaches or exceeds a preset height threshold, the object is identified as a hard obstacle. The control system then performs a detour, causing the cleaning robot to bypass the area corresponding to the foreign object. Simultaneously, the cleaning components continue operating during the detour to ensure that the surface of the photovoltaic modules along the detour path is still cleaned. The control system can also record the detour trajectory and, combined with the effective coverage width of the cleaning components, determine the actual covered area and mark it as the cleaned area. This process ensures robot safety and reduces missed areas caused by detours.

[0077] Furthermore, the process of controlling the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object may specifically include: expanding a preset safety buffer distance outward based on the boundary of the area corresponding to the foreign object to generate a virtual obstacle zone; dividing the path segment in the current bow-shaped cleaning path that intersects with the virtual obstacle zone into a front obstacle segment and a rear obstacle segment; controlling the cleaning robot to execute the front obstacle segment, then detouring along the boundary of the virtual obstacle zone to the corresponding position of the rear obstacle segment, and continuing to execute the rear obstacle segment; after the bow-shaped cleaning path of the current cleaning cycle is completed, generating a supplementary cleaning path for the uncovered area around the virtual obstacle zone and performing supplementary cleaning.

[0078] In some implementations, if an obstacle at the same location is continuously detected in multiple consecutive cleaning cycles, the location and boundary information of the corresponding virtual obstacle area can be stored as permanent obstacle area information. In subsequent cleaning cycles, the permanent obstacle area information is retrieved for path planning to reduce the computational overhead caused by repeated detection and judgment. For pre-stored permanent obstacle areas, the system can still update and correct their positions and boundaries based on real-time detection results; this embodiment of the invention does not limit this.

[0079] 205. If the height of the foreign object is less than the preset height threshold, the cleaning robot will be controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning.

[0080] In this embodiment, for foreign objects with a height below a preset threshold, the cleaning robot can directly pass along the original path and perform cleaning without adjusting the cleaning path, thereby reducing unnecessary path deflection. In some implementations, this direct passage is contingent upon the foreign object not interfering with the robot's chassis, walking mechanism, or cleaning components. This step improves path continuity and operational efficiency while ensuring safety.

[0081] 206. When a foreign object is identified as a cleanable contaminant and its height is less than a preset height threshold, the cleaning robot is controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. If the foreign object is identified as a cleanable contaminant but its height is greater than or equal to the preset height threshold, the foreign object is forcibly identified as a hard obstacle, and the cleaning robot is controlled to adjust its cleaning path to bypass the area corresponding to the foreign object.

[0082] In this embodiment, the cleanable contaminants are typically of low height, generally not exceeding a preset height threshold. When a foreign object is identified as a cleanable contaminant and its height is less than the preset height threshold, the control system controls the cleaning robot to pass along the original cleaning path and perform cleaning. If a foreign object is identified as a cleanable contaminant but its height is greater than or equal to the preset height threshold, it is considered a misidentification. The control system forcibly classifies the foreign object as a hard obstacle and controls the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, prioritizing the safety of the cleaning robot. In this way, it is possible to avoid misidentifying large but cleanable contaminants as objects to bypass, and to provide a safety fallback for foreign objects with abnormal heights.

[0083] 207. Obtain second image information of the area cleaned by the cleaning robot.

[0084] In this embodiment, step 207 is used to acquire image information of the cleaned area after the cleaning robot completes cleaning the corresponding area, as the basis data for subsequent cleanliness scoring and cleaning strategy determination. The second image information can be acquired by an image acquisition device located at the bottom, rear, or other suitable positions for capturing images of the cleaned area of ​​the cleaning robot, or by other detection devices installed in conjunction with the cleaning robot. By acquiring images of the cleaned area after cleaning is completed, subsequent cleaning effect judgments can be based on the actual cleaning results, thereby improving the objectivity and accuracy of cleaning quality assessment.

[0085] 208. Based on the pre-established cleanliness scoring model, analyze the specified area in the second image information to obtain the cleanliness score corresponding to the specified area.

[0086] In this embodiment, the designated area can be a local evaluation unit within the cleaned area. The local evaluation unit can be divided according to a preset area, preset length, areas corresponding to multiple adjacent battery cells, sub-areas of multiple adjacent components, or other methods; this embodiment does not limit this. In some implementations, the designated area can be divided according to areas corresponding to N adjacent battery cells, where N can be an integer between 8 and 12, such as 10. In other implementations, the designated area can also be divided according to the maximum lateral cleaning distance achievable by the cleaning robot's cleaning components, i.e., using the effective cleaning width of the cleaning components as the lateral division basis for the local evaluation unit. Since the lateral coverage of a single cleaning operation by the cleaning robot is the effective cleaning width of the cleaning components, dividing the evaluation unit based on this width ensures consistency between the evaluation unit and the actual cleaning coverage, facilitating a direct correlation between the cleanliness score and the actual cleaning effect. Furthermore, it eliminates the need for manually set parameters such as preset area and preset length, making the division method more aligned with the cleaning robot's operational characteristics. By dividing the cleaned area into appropriately sized evaluation units, a balance can be achieved between evaluation accuracy and computational overhead.

[0087] Furthermore, the process of obtaining a cleanliness score based on a pre-established cleanliness scoring model may specifically include: determining a first sub-score based on the surface reflection characteristics of a specified area; determining a second sub-score based on the grayscale distribution characteristics of the specified area; inputting the image corresponding to the specified area into a pre-trained image classification model to obtain a third sub-score; and performing weighted fusion of the first, second, and third sub-scores to obtain the cleanliness score. For the specific implementation methods of each sub-score and the weighted fusion, please refer to the relevant description in Example 1.

[0088] 209. Based on the pre-established cleaning strategy model, determine the cleaning treatment strategy for the specified area according to the cleanliness score of the specified area.

[0089] In this embodiment, the process of determining the cleaning strategy based on the cleanliness score can specifically include: when the cleanliness score is less than a first threshold, determining the cleaning strategy as an alarm strategy; when the cleanliness score is greater than or equal to the first threshold and less than a second threshold, determining the cleaning strategy as a re-sweeping strategy; and when the cleanliness score is greater than or equal to the second threshold, determining the cleaning strategy as a normal pass strategy. The first threshold is less than the second threshold; for example, the first threshold can be set to 30%, and the second threshold can be set to 70%, or it can be adjusted according to actual operation and maintenance standards. By comparing the cleanliness score with the preset thresholds, the corresponding subsequent processing method can be automatically selected for different cleaning effects.

[0090] Furthermore, when the cleaning strategy is determined to be a re-sweeping strategy, a re-sweeping can be performed on a designated area, and the cleanliness score corresponding to the designated area can be obtained again after the re-sweeping. If the cleanliness score after the re-sweeping is still less than the second threshold, and the current number of re-sweeping is less than the preset maximum number of times, then the designated area is re-sweeped again. If the current number of re-sweeping reaches the preset maximum number of times, and the cleanliness score is still less than the second threshold, then the cleaning strategy corresponding to the designated area is switched from the re-sweeping strategy to the alarm strategy.

[0091] In some implementations, for local areas marked as cleaned areas during the detour, when the system reaches the vicinity of the area again during subsequent cleaning processes, the area can also be included in the above-mentioned second image acquisition, cleanliness scoring and strategy determination process to determine whether supplementary cleaning, re-sweeping or alarm processing is required.

[0092] This shows that, Figure 2 In the embodiment shown, the detection and classification of foreign objects in front, height judgment, obstacle avoidance path adjustment, cleaning status management during detour, cleaning effect evaluation of the cleaned area, and determination of subsequent cleaning strategies can be integrated into one. This allows the cleaning robot to adaptively adjust its cleaning behavior based on the situation of foreign objects in front and the cleaning effect behind when cleaning continuously along the preset path, thereby helping to balance cleaning efficiency while ensuring cleaning coverage and cleaning effect. Example 3

[0093] Please see Figure 8 , Figure 8 This is a schematic diagram of an adaptive cleaning control device for a photovoltaic module cleaning robot disclosed in an embodiment of the present invention. The adaptive cleaning control device can be deployed on the controller of the cleaning robot body, or it can be deployed on a remote server and interact with the cleaning robot body via a communication module. This embodiment of the present invention does not limit this. The adaptive cleaning control device is a specific implementation of the control system involved in the aforementioned Embodiments 1 and 2. The control system, through the cooperation of the image acquisition module 301, foreign object recognition module 302, path control module 303, cleanliness assessment module 304, and strategy determination module 305 in the adaptive cleaning control device, realizes functions such as foreign object recognition, path control, cleanliness assessment, and strategy determination. The adaptive cleaning control device includes an image acquisition module 301, a foreign object recognition module 302, a path control module 303, a cleanliness assessment module 304, and a strategy determination module 305, wherein:

[0094] The image acquisition module 301 is used to acquire first image information of the area where the cleaning robot is operating, and second image information of the area that the cleaning robot has cleaned.

[0095] In some implementations, the image acquisition module 301 can be communicatively connected to one or more image acquisition units to receive images of the operating area and the cleaned area, respectively, providing data input for subsequent foreign object identification and cleanliness assessment.

[0096] The foreign object recognition module 302 is used to recognize the first image information based on a pre-established foreign object recognition model, so as to identify the category of foreign object when foreign object is present in the first image information.

[0097] In some implementations, the foreign object identification module 302 may output foreign object category information, location information, contour information and / or confidence information, wherein the foreign object category includes at least hard obstacles and cleanable contaminants.

[0098] The path control module 303 is used to control the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object when the foreign object is identified as a hard obstacle, based on a pre-established obstacle avoidance model, and to control the cleaning robot to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning when the foreign object is identified as a cleanable contaminant.

[0099] In some implementations, the path control module 303 can also control the cleaning components to remain in working condition during the detour, and mark the areas traversed by the cleaning robot during the detour as cleaned areas based on the detour trajectory and the effective coverage width of the cleaning components.

[0100] Furthermore, the path control module 303 can also generate a virtual obstacle zone based on the boundary of the area corresponding to the foreign object, and perform local cutting, detour and supplementary cleaning processing on the bow-shaped cleaning path that intersects with the virtual obstacle zone.

[0101] In other embodiments, the path control module 303 can also establish permanent obstacle zone information for obstacles continuously detected in multiple consecutive cleaning cycles, for use in subsequent path pre-planning.

[0102] The cleanliness assessment module 304 is used to analyze a specified area in the second image information based on a pre-established cleanliness scoring model to obtain the cleanliness score corresponding to the specified area.

[0103] In some implementations, the cleanliness assessment module 304 can quantitatively assess the cleaning effect based on at least one of surface reflection characteristics, grayscale distribution characteristics, image classification results, and other auxiliary characteristics.

[0104] The strategy determination module 305 is used to determine the cleaning treatment strategy for a specified area based on a pre-established cleaning strategy model and the cleanliness score of the specified area.

[0105] In some implementations, the strategy determination module 305 can select between an alarm strategy, a re-sweeping strategy, and a normal passage strategy based on the comparison result between the cleanliness score and a preset threshold, and output the corresponding control command to the actuator of the cleaning robot or the upper control system.

[0106] This shows that, Figure 3 In the embodiment shown, the cooperation between the image acquisition module 301, the foreign object recognition module 302, the path control module 303, the cleanliness assessment module 304, and the strategy determination module 305 enables closed-loop processing from foreign object detection and classification, path control to cleaning effect assessment and subsequent strategy decision-making.

[0107] In an optional embodiment, the path control module 303 is further configured to, when a foreign object is identified as a hard obstacle, first determine whether the height of the hard obstacle is greater than or equal to a preset height threshold, and only control the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object when the height is greater than or equal to the preset height threshold.

[0108] In another optional embodiment, the process by which the cleanliness assessment module 304 obtains a cleanliness score specifically includes: determining a first sub-score based on the surface reflection features of a specified area; determining a second sub-score based on the grayscale distribution features of a specified area; inputting the image corresponding to the specified area into a pre-trained image classification model to obtain a third sub-score; and performing weighted fusion of the first sub-score, the second sub-score, and the third sub-score to obtain a cleanliness score.

[0109] In another optional embodiment, the process of the strategy determination module 305 determining the cleaning process strategy specifically includes: when the cleanliness score is less than a first threshold, determining the cleaning process strategy as an alarm strategy; when the cleanliness score is greater than or equal to the first threshold and less than a second threshold, determining the cleaning process strategy as a re-sweeping strategy; and when the cleanliness score is greater than or equal to the second threshold, determining the cleaning process strategy as a normal pass strategy. Example 4

[0110] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of another adaptive cleaning control device for a photovoltaic module cleaning robot disclosed in an embodiment of the present invention. Figure 9 As shown, the adaptive cleaning control device for a photovoltaic module cleaning robot may include: Memory 401 containing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the adaptive cleaning control method for photovoltaic module cleaning robots described in Embodiment 1 or Embodiment 2 of the present invention. Example 5

[0111] This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the adaptive cleaning control methods for photovoltaic module cleaning robots disclosed in Embodiment 1 of this invention. Example 6

[0112] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the adaptive cleaning control method for a photovoltaic module cleaning robot described in Embodiment 1.

[0113] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. 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.

[0114] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method 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, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0115] Finally, it should be noted that the adaptive cleaning control method and apparatus for photovoltaic module cleaning robots disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. 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. Such 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. An adaptive cleaning control method for a photovoltaic module cleaning robot, characterized in that, The method includes: Acquire first image information of the area where the cleaning robot is operating; Based on a pre-established foreign object recognition model, the first image information is identified to determine the category of the foreign object when it is present in the first image information. When a foreign object is identified as a hard obstacle, the cleaning robot is controlled to adjust its cleaning path based on a pre-established obstacle avoidance model in order to bypass the area corresponding to the foreign object. When a foreign object is identified as a cleanable contaminant, the cleaning robot is controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. Acquire second image information of the area already cleaned by the cleaning robot; Based on a pre-established cleanliness scoring model, a specified area in the second image information is analyzed to obtain the cleanliness score corresponding to the specified area. Based on a pre-established cleaning strategy model, the cleaning strategy for a specified area is determined according to the cleanliness score of the specified area. The process of controlling the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, based on a pre-established obstacle avoidance model, when the foreign object is identified as a hard obstacle, also includes: Control the cleaning components of the cleaning robot to keep them in working order; The areas traversed by the cleaning robot during its detour are marked as cleaned areas.

2. The adaptive cleaning control method according to claim 1, characterized in that, After identifying the first image information based on the pre-established foreign object recognition model, and identifying the category of the foreign object when it is present in the first image information, the method further includes: Determine the height of the foreign object; Determine whether the height of the foreign object is greater than or equal to a preset height threshold; If the height of the foreign object is greater than or equal to the preset height threshold, the foreign object is identified as a hard obstacle, and the cleaning robot is controlled to adjust the cleaning path to bypass the area corresponding to the foreign object. If the height of the foreign object is less than the preset height threshold, the cleaning robot will be controlled to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning. Foreign objects with a height greater than or equal to a preset height threshold are all identified as hard obstacles. Even if the foreign object recognition model identifies the foreign object as a cleanable contaminant, it is still forcibly identified as a hard obstacle to prioritize the safety of the cleaning robot.

3. The adaptive cleaning control method according to claim 1, characterized in that, After the process of marking the areas traversed by the cleaning robot during its detour as cleaned areas, the method further includes: When the cleaning robot moves to the already cleaned area during subsequent cleaning, it obtains a cleanliness score for the cleaned area. Based on the cleanliness score within the specified range, determine whether to perform supplementary cleaning on the already cleaned areas.

4. The adaptive cleaning control method according to claim 1, characterized in that, The process of analyzing a designated area in the second image information based on a pre-established cleanliness scoring model to obtain the cleanliness score corresponding to the designated area includes: The first sub-score is determined based on the reflectivity of the specified area; The second sub-score is determined based on the grayscale values ​​of a specified region; Input the image corresponding to the specified region into a pre-trained image classification model to obtain the third sub-score; The first, second, and third sub-scores are weighted and fused to obtain the cleanliness score.

5. The adaptive cleaning control method according to claim 1, characterized in that, The process of determining the cleaning strategy for a specified area based on a pre-established cleaning strategy model and a cleanliness score includes: When the cleanliness score is less than the first threshold, the cleaning strategy is determined to be an alarm strategy, which is used to output a manual intervention cleaning prompt. When the cleanliness score is greater than or equal to the first threshold and less than the second threshold, the cleaning strategy is determined to be the re-sweeping strategy, which is used to perform re-sweeping on the specified area. When the cleanliness score is greater than or equal to the second threshold, the cleaning strategy is determined to be the normal pass strategy. The normal pass strategy is used to control the cleaning robot to continue cleaning along the cleaning path.

6. The adaptive cleaning control method according to claim 5, characterized in that, The process of determining the cleaning strategy as a re-sweeping strategy includes: Perform a rescan on the designated area and re-obtain the cleanliness score for the designated area after the rescan; If the cleanliness score is still less than the second threshold after the rescan, and the current number of rescans is less than the preset maximum number of rescans, then the designated area will be rescanned again. If the current number of rescans reaches the preset limit and the cleanliness score is still less than the second threshold, the cleaning strategy for the specified area will be switched from the rescan strategy to the alarm strategy.

7. The adaptive cleaning control method according to claim 1, characterized in that, When a foreign object is identified as a hard obstacle, the process of controlling the cleaning robot to adjust its cleaning path to bypass the area corresponding to the foreign object, based on a pre-established obstacle avoidance model, includes: A virtual obstacle zone is generated by extending a preset safety buffer distance outward from the boundary of the area corresponding to the foreign object. Divide the path segment that intersects with the virtual obstacle area in the current bow-shaped cleaning path into the obstacle front segment and the obstacle rear segment; After controlling the cleaning robot to execute the first part of the obstacle, it travels along the boundary of the virtual obstacle area to the corresponding position of the second part of the obstacle and continues to execute the second part of the obstacle. After the current cleaning cycle's bow-shaped cleaning path is completed, a supplementary cleaning path is generated for the uncovered areas around the virtual obstacle zone, and supplementary cleaning is performed.

8. An adaptive cleaning control device for a photovoltaic module cleaning robot, characterized in that, The device includes: The image acquisition module is used to acquire first image information of the area where the cleaning robot is operating, and second image information of the area that the cleaning robot has cleaned. The foreign object recognition module is used to identify the first image information based on a pre-established foreign object recognition model, so as to identify the category of foreign object when foreign object is present in the first image information; The path control module is used to control the cleaning robot to adjust the cleaning path to bypass the area corresponding to the foreign object when the foreign object is identified as a hard obstacle, based on a pre-established obstacle avoidance model. It is also used to control the cleaning robot to pass through the area corresponding to the foreign object along the cleaning path and perform cleaning when the foreign object is identified as a cleanable contaminant. The cleanliness assessment module is used to analyze a specified area in the second image information based on a pre-established cleanliness scoring model, and obtain the cleanliness score corresponding to the specified area. The strategy determination module is used to determine the cleaning strategy for a specified area based on a pre-established cleaning strategy model and the cleanliness score of the specified area.

9. An adaptive cleaning control device for a photovoltaic module cleaning robot, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the adaptive cleaning control method for a photovoltaic module cleaning robot as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the adaptive cleaning control method for a photovoltaic module cleaning robot as described in any one of claims 1-7.

Images