Path determination method, device and equipment for polycrystalline silicon photovoltaic cleaning robot

Abstract

The invention discloses a path determination method, device and equipment for a polycrystalline silicon photovoltaic cleaning robot, and relates to the field of data processing. The method comprises the following steps: performing multidirectional scanning on a photovoltaic park to obtain a plurality of photovoltaic park pictures at different visual angles; constructing a three-dimensional point cloud model corresponding to the photovoltaic park; according to the three-dimensional point cloud model, obtaining position information and inclination angle information of the plurality of polycrystalline silicon photovoltaic modules; taking the current position as a starting point, performing layered random sampling on the park ground in the three-dimensional point cloud model, and generating a plurality of alternative navigation points; based on the plurality of alternative navigation points, screening to obtain a plurality of target navigation points, and based on the plurality of target navigation points, generating an advancing route; the cleaning height of the cleaning robot is obtained, and a cleaning path is generated by integrating the position information, the inclination angle information and the cleaning height; and cleaning the polycrystalline silicon photovoltaic module according to the advancing route and the cleaning path. By implementing the technical scheme provided by the invention, the cleaning efficiency of the photovoltaic module is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and particularly relates to a method, device, and equipment for determining the path of a polysilicon photovoltaic cleaning robot. Background Art

[0002] With the transformation of the global energy structure and the improvement of environmental protection awareness, the photovoltaic industry has developed rapidly. Large-scale photovoltaic parks have become common renewable energy production bases, which usually cover a vast geographical area and deploy a large number of polysilicon photovoltaic modules. These modules need to be cleaned regularly to maintain high energy conversion efficiency. Therefore, efficient photovoltaic module cleaning technology has become particularly important.

[0003] Currently, in the related art, the cleaning of photovoltaic modules adopts the method of manual cleaning. Manual cleaning requires a large amount of labor, especially in vast photovoltaic parks, and the moving speed and cleaning speed of manual labor are relatively slow, further prolonging the overall cleaning time. Therefore, the related art has the problem of low cleaning efficiency.

[0004] Therefore, there is an urgent need for a method, device, and equipment for determining the path of a polysilicon photovoltaic cleaning robot. Summary of the Invention

[0005] The present application provides a method, device, and equipment for determining the path of a polysilicon photovoltaic cleaning robot, which improves the cleaning efficiency of photovoltaic modules.

[0006] In the first aspect of the present application, a method for determining the path of a polysilicon photovoltaic cleaning robot is provided. The method includes: scanning the photovoltaic park in multiple directions to obtain multiple photovoltaic park pictures from different perspectives; constructing a three-dimensional point cloud model corresponding to the photovoltaic park according to the multiple photovoltaic park pictures; obtaining the position information and tilt angle information of multiple polysilicon photovoltaic modules according to the three-dimensional point cloud model; taking the current position as the starting point, performing stratified random sampling on the park ground in the three-dimensional point cloud model to generate multiple alternative navigation points, wherein a first sampling density is adopted in the ground area within a preset distance from the target polysilicon photovoltaic module, and a second sampling density is adopted in the ground area not within the preset distance from the target polysilicon photovoltaic module, the first sampling density is greater than the second sampling density, and the target polysilicon photovoltaic module is any one of the multiple polysilicon photovoltaic modules; screening multiple target navigation points based on the multiple alternative navigation points, and generating a travel route based on the multiple target navigation points; obtaining the cleaning height of the cleaning robot, and generating a cleaning path by integrating the position information, the tilt angle information, and the cleaning height; cleaning the polysilicon photovoltaic modules according to the travel route and the cleaning path.

[0007] By adopting the above technical solutions, by performing multi-directional scanning on the photovoltaic park, obtaining pictures of the photovoltaic park from different perspectives, and then constructing a three-dimensional point cloud model based on these pictures of the photovoltaic park, the actual terrain of the photovoltaic park and the spatial distribution of polysilicon photovoltaic modules can be comprehensively and accurately reflected. Using this three-dimensional point cloud model, key information such as the position and tilt angle of each polysilicon photovoltaic module can be accurately obtained, providing a reliable data basis for subsequent steps such as navigation point generation and path planning. When generating alternative navigation points, a hierarchical random sampling strategy is adopted. According to the distance from the target polysilicon photovoltaic module, different sampling densities are used, which can reduce the computational complexity while ensuring navigation accuracy and improve the efficiency of navigation point generation. By screening the alternative navigation points to obtain the target navigation points and then generating a travel route based on the target navigation points, a navigation route with the optimal path can be obtained. At the same time, comprehensively considering the cleaning height of the cleaning robot, as well as the position and tilt angle of the polysilicon photovoltaic module, a matching cleaning path is generated, which can achieve precise and efficient cleaning operations. Finally, the cleaning robot autonomously cleans the polysilicon photovoltaic module according to the travel route and the cleaning path, which can significantly improve the cleaning efficiency, reduce labor costs, and ensure the cleaning effect and safety.

[0008] Optionally, starting from the current position, perform hierarchical random sampling on the park ground in the three-dimensional point cloud model to generate multiple alternative navigation points, specifically including: dividing the park ground in the three-dimensional point cloud model into multiple grid regions; determining the center point coordinates of each of the grid regions; calculating the distance between the center point coordinates of each of the grid regions and the target polysilicon photovoltaic module; marking the grid regions whose distance from the target polysilicon photovoltaic module is within a preset distance as the first grid regions, and marking the grid regions whose distance from the target polysilicon photovoltaic module is not within the preset distance as the second grid regions; performing random sampling on the first grid regions with the first sampling density, and performing random sampling on the second grid regions with the second sampling density to obtain multiple of the alternative navigation points.

[0009] By adopting the above technical solutions, by dividing the park ground in the three-dimensional point cloud model into multiple grid regions and calculating the distance between the center point coordinates of each grid region and the target polysilicon photovoltaic module, it can be quickly and accurately determined which grid regions are near the target polysilicon photovoltaic module and which are in the farther regions. By adjusting the magnitudes of the first sampling density and the second sampling density, the distribution and quantity of the navigation points can be flexibly controlled to adapt to different terrains and task requirements. This solution utilizes the idea of hierarchical sampling, fully considering the spatial distribution characteristics and value differences of the navigation points, and while improving the sampling efficiency, it also ensures the representativeness and effectiveness of the navigation points, providing high-quality alternative nodes for subsequent path planning and optimization.

[0010] Optionally, filtering multiple target navigation points based on the multiple alternative navigation points and generating a travel route based on the multiple target navigation points specifically includes: inputting the position information of the polysilicon photovoltaic module into a preset navigation range model to obtain a corresponding target range, where the preset navigation range model includes the corresponding relationship between the position information of the polysilicon photovoltaic module and the optimal navigation distance; determining the alternative navigation points within the target range as the target navigation points; calculating the distances and azimuth angles between the respective target navigation points; based on a preset navigation point selection strategy, determining a first navigation point and a second navigation point from the target navigation points, where the first navigation point and the second navigation point are any two target navigation points among the multiple target navigation points, and where the second navigation point is the target navigation point among the multiple target navigation points that is the closest to the first navigation point or has the smallest azimuth angle; connecting the first navigation point and the second navigation point to generate the travel route.

[0011] By adopting the above technical solution, introducing a preset navigation range model and establishing a mapping relationship between the position information of the polysilicon photovoltaic module and the corresponding optimal navigation distance can make full use of existing empirical data to quickly and accurately predict the optimal navigation range of each polysilicon photovoltaic module. Comparing the alternative navigation points with the predicted target range and screening out the alternative navigation points within the optimal navigation range as the target navigation points can automatically obtain a set of target navigation points, providing reliable input for subsequent path generation. By calculating the distances and azimuth angles between the target navigation points and automatically determining the starting point and the ending point according to the preset navigation point selection strategy, and then connecting the starting point and the ending point to generate a travel route, a travel route with the optimal starting point and ending point and covering all target navigation points can be obtained. This solution uses the method of machine learning to establish a preset navigation range model, which can automatically learn and optimize according to historical data and experience, continuously improving the prediction accuracy and adaptability.

[0012] Optionally, generating a cleaning path by synthesizing the position information, the tilt angle information, and the cleaning height specifically includes: determining the horizontal position coordinates of the cleaning robot at each cleaning position according to the position information and the travel route; calculating the tilt angle of the surface of the polysilicon photovoltaic module at each cleaning position according to the tilt angle information, and determining the cleaning angle of the cleaning brush of the cleaning robot according to the tilt angle, so that the cleaning angle is consistent with the normal direction of the surface of the polysilicon photovoltaic module; constructing a path planning graph based on graph theory, taking the positions of each polysilicon photovoltaic module as nodes of the path planning graph, taking the distance between each adjacent polysilicon photovoltaic module as the weight of the edge between each node in the path planning graph, and taking the horizontal position coordinates, the cleaning angle, and the cleaning height as the attributes of each node; using the Dijkstra algorithm to search for the shortest path in the path planning graph to obtain the cleaning path.

[0013] By adopting the above technical solution, by comprehensively utilizing the position information of the polysilicon photovoltaic module, the tilt angle information, and the cleaning height of the cleaning robot, the horizontal position coordinates of the cleaning robot at each cleaning position, the cleaning angle of the cleaning brush, and the cleaning height parameters can be accurately calculated, providing an accurate control basis for performing high-quality cleaning operations. Among them, by analyzing the tilt angle of the surface of the polysilicon photovoltaic module and dynamically adjusting the cleaning angle of the cleaning brush to be consistent with the normal direction of the component surface, the optimal cleaning angle can be achieved, the cleaning effect can be improved, and the wear on the component surface can be reduced. Using the graph theory algorithm, modeling the cleaning position and related parameter information as a path planning graph, and using the Dijkstra algorithm for shortest path search, a cleaning path with the shortest total distance and the optimal cleaning effect can be efficiently found among many possible cleaning paths. This solution makes full use of the environmental information obtained by multi-sensor fusion, and through refined modeling and intelligent optimization algorithms, realizes the automatic planning of the cleaning path and parameter control, which can significantly improve the accuracy, efficiency, and safety of the cleaning operation, and reduce the difficulty and labor intensity of manual operation. At the same time, the modeling method based on graph theory and the shortest path search algorithm have universality and scalability.

[0014] Optionally, constructing the three-dimensional point cloud model corresponding to the photovoltaic park according to the multiple photovoltaic park pictures specifically includes: extracting features from the first photovoltaic park picture to obtain first features, and extracting features from the second photovoltaic park picture to obtain second features, where the first photovoltaic park picture and the second photovoltaic park picture are any two photovoltaic park pictures among the multiple photovoltaic park pictures; performing feature point matching on the first features and the second features to obtain matching feature points; estimating the relative pose between the first photovoltaic park picture and the second photovoltaic park picture according to the matching feature points; triangulating the matching feature point pairs according to the relative pose to obtain three-dimensional point cloud data of the matching feature points; and constructing the three-dimensional point cloud model corresponding to the photovoltaic park according to the three-dimensional point cloud data.

[0015] By adopting the above technical solution, by extracting and matching features of photovoltaic park pictures from different perspectives, the corresponding relationship between pictures can be found. Then, using these corresponding relationships, the relative pose between the pictures can be estimated, that is, their relative position and angle in the three-dimensional space. Based on this pose information, the matching feature points can be triangulated to obtain their three-dimensional coordinates, thereby forming the three-dimensional point cloud model of the entire park. This method overcomes the problem of insufficient information in a single picture and can restore the complete three-dimensional structure of the park by integrating multi-perspective information, providing a reliable environmental model for subsequent cleaning path planning.

[0016] Optionally, performing feature point matching on the first features and the second features to obtain matching feature points specifically includes: calculating the Euclidean distance between the first features and the second features; judging the size relationship between the Euclidean distance and a preset distance threshold; and if it is determined that the Euclidean distance is less than the preset distance threshold, then the first features and the second features are used as the matching feature points.

[0017] By adopting the above technical solution, by calculating the Euclidean distance between two feature points and comparing it with a preset distance threshold, it can be judged whether the two feature points match. The advantage of this method is that the calculation is simple, the speed is fast, and it is easy to implement in real time. The Euclidean distance can effectively measure the similarity degree of feature points in the feature space. The closer the distance, the more similar the features, so that reliable matching point pairs can be screened out. At the same time, by setting a preset distance threshold, the matching accuracy and the tolerance for incorrect matching can be flexibly controlled.

[0018] Optionally, obtaining the position information of the polysilicon photovoltaic module according to the three-dimensional point cloud model specifically includes: performing clustering segmentation on the three-dimensional point cloud model to obtain a plurality of point cloud clusters; calculating the height and width of each of the point cloud clusters; determining a target point cloud cluster among the plurality of point cloud clusters, where the target point cloud cluster is a point cloud cluster with a height greater than a preset height threshold and a width greater than a preset width threshold among the plurality of point cloud clusters; obtaining the point cloud cluster information of the target point cloud cluster, and using the point cloud cluster information as the position information of the polysilicon photovoltaic module, where the point cloud cluster information includes the position, height, and width of the target point cloud cluster.

[0019] By adopting the above technical solution, first, the entire point cloud is subjected to clustering segmentation and divided into a plurality of non-overlapping point cloud clusters. This step can effectively distinguish different objects and regions in the point cloud. Then, the height and width of each point cloud cluster are calculated respectively. By comparing with the preset height threshold and the preset width threshold, the point cloud clusters that meet the characteristics of the photovoltaic module can be screened out. These point cloud clusters correspond to the real photovoltaic modules, and their position, height, and width information can approximately represent the spatial attributes of the actual modules. This method cleverly utilizes the prior knowledge of the size of the photovoltaic module and realizes the positioning and extraction of the module through the clustering and statistical analysis of the point cloud, avoiding complex semantic segmentation or object recognition steps. By making full use of the structural characteristics of the photovoltaic module, the point cloud information is fully mined, enabling the robot to accurately perceive the position of the cleaning object and laying a solid foundation for the execution of the cleaning task in terms of environmental perception.

[0020] In a second aspect of the present application, a path determination device for a polysilicon photovoltaic cleaning robot is provided. The device includes: a scanning and photographing module, a model construction module, a position information determination module, and a cleaning path determination module; the scanning and photographing module is configured to perform multi-directional scanning on a photovoltaic park to obtain multiple photovoltaic park pictures from different perspectives; the model construction module is configured to construct a three-dimensional point cloud model corresponding to the photovoltaic park according to the multiple photovoltaic park pictures; the position information determination module is configured to obtain the position information and tilt angle information of multiple polysilicon photovoltaic modules according to the three-dimensional point cloud model; the cleaning path determination module is configured to take the current position as a starting point, perform stratified random sampling on the ground of the park in the three-dimensional point cloud model to generate multiple alternative navigation points, wherein a first sampling density is adopted for the ground area within a preset distance from the target polysilicon photovoltaic module, and a second sampling density is adopted for the ground area not within the preset distance from the target polysilicon photovoltaic module, the first sampling density is greater than the second sampling density, and the target polysilicon photovoltaic module is any one of the multiple polysilicon photovoltaic modules; the cleaning path determination module is further configured to screen out multiple target navigation points based on the multiple alternative navigation points, and generate a travel route based on the multiple target navigation points; the cleaning path determination module is further configured to obtain the cleaning height of the cleaning robot, and generate a cleaning path by integrating the position information, the tilt angle information, and the cleaning height; the cleaning path determination module is further configured to clean the polysilicon photovoltaic module according to the travel route and the cleaning path.

[0021] In a third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of the above.

[0022] In a fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, and when the instructions are executed, the method as described in any one of the above is executed.

[0023] In summary, one or more technical solutions provided by the present application have at least the following technical effects or advantages: 1. By scanning the photovoltaic park from multiple angles to obtain images and constructing a three-dimensional point cloud model of the park based on the images, the three-dimensional point cloud model of the park can be accurately reconstructed. Then, the positions of polysilicon photovoltaic modules are extracted from the three-dimensional point cloud model, and target navigation points are randomly sampled on the ground of the park to explore feasible cleaning paths within the global scope. Next, the target navigation points are screened according to the positions of the polysilicon photovoltaic modules to obtain target navigation points with appropriate distances from the polysilicon photovoltaic modules, ensuring the safety and efficiency of the cleaning process. Finally, the optimal cleaning path is generated based on the target navigation points and the robot is controlled to perform the cleaning, realizing the automation and intelligence of the cleaning process. Through the organic combination of technologies such as machine vision, point cloud processing, and path planning, this series of steps enables the cleaning robot to autonomously adapt to the environments of different photovoltaic parks and complete the cleaning task with the optimal path, greatly improving the cleaning efficiency and quality of photovoltaic modules.

[0024] 2. By extracting and matching features from photovoltaic park pictures with different perspectives, the corresponding relationships between the pictures can be found. Then, the relative poses between the pictures are estimated using these corresponding relationships, that is, their relative positions and angles in the three-dimensional space. Based on this pose information, the triangulation measurement of the matched feature points can be performed to obtain their three-dimensional coordinates, thus forming the three-dimensional point cloud model of the entire park. This method overcomes the problem of insufficient information in a single picture and can restore the complete three-dimensional structure of the park by integrating multi-perspective information, providing a reliable environmental model for subsequent cleaning path planning.

[0025] 3. By calculating the Euclidean distance between two feature points and comparing it with a preset distance threshold, it can be determined whether the two feature points match. The advantage of this method is that the calculation is simple, the speed is fast, and it is easy to implement in real time. The Euclidean distance can effectively measure the similarity degree of feature points in the feature space. The closer the distance, the more similar the features, so reliable matching point pairs can be screened out. At the same time, by setting the preset distance threshold, the matching accuracy and the tolerance for incorrect matches can be flexibly controlled. Brief Description of the Drawings

[0026] Figure 1 is a schematic flowchart of a method for determining a path of a polysilicon photovoltaic cleaning robot disclosed in an embodiment of the present application; Figure 2 is a schematic module diagram of a device for determining a path of a polysilicon photovoltaic cleaning robot disclosed in an embodiment of the present application; Figure 3 is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application.

[0027] Explanation of the accompanying drawings: 201, scanning and shooting module; 202, model building module; 203, position information determination module; 204, cleaning path determination module; 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0028] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0029] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0030] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0031] The present application provides a method for determining the path of a polysilicon photovoltaic cleaning robot, referring to Figure 1 , Figure 1 1 is a flow chart of a method for determining a path of a polysilicon photovoltaic cleaning robot provided in an embodiment of the present application. The method is applied to a cleaning robot, and the cleaning robot can execute a path determination program of the polysilicon cleaning robot. The method includes steps S101 to S107, and the steps are as follows: Step S101: perform a multi-directional scan of the photovoltaic park to obtain a plurality of photovoltaic park pictures from different viewing angles.

[0032] In step S101, the cleaning robot performs a multi-directional scan of the stair environment to obtain multiple photovoltaic park pictures from different perspectives. This process can be achieved by a camera mounted on the cleaning robot or by a camera built into the photovoltaic park.

[0033] Step S102: Construct a three-dimensional point cloud model corresponding to the photovoltaic park according to multiple photovoltaic park pictures.

[0034] In step S102, to construct a three-dimensional point cloud model corresponding to the photovoltaic park according to multiple photovoltaic park pictures, it specifically includes: extracting features from the first photovoltaic park picture to obtain the first feature, and extracting features from the second photovoltaic park picture to obtain the second feature, where the first photovoltaic park picture and the second photovoltaic park picture are any two of the multiple photovoltaic park pictures; matching the feature points of the first feature and the second feature to obtain the matching feature points; estimating the relative pose between the first photovoltaic park picture and the second photovoltaic park picture according to the matching feature points; triangulating the pair of matching feature points according to the relative pose to obtain the three-dimensional point cloud data of the matching feature points; and constructing a three-dimensional point cloud model corresponding to the photovoltaic park according to the three-dimensional point cloud data.

[0035] In a possible implementation manner, to match the feature points of the first feature and the second feature to obtain the matching feature points, it specifically includes: calculating the Euclidean distance between the first feature and the second feature; judging the magnitude relationship between the Euclidean distance and a preset distance threshold; if it is determined that the Euclidean distance is less than the preset distance threshold, then the first feature and the second feature are used as the matching feature points.

[0036] Specifically, the cleaning robot randomly selects any two photovoltaic park pictures from the multiple obtained photovoltaic park pictures, namely the first photovoltaic park picture and the second photovoltaic park picture. Then, feature extraction is performed on these two photovoltaic park pictures respectively. The purpose of feature extraction is to find relatively stable and easily recognizable points or regions in the pictures, which are called key points. The feature extraction algorithm used in this application is SIFT. The cleaning robot will apply the SIFT feature extraction algorithm to the two photovoltaic park pictures respectively to obtain a series of feature points and their descriptors for each, which are called the first feature and the second feature respectively. The next step is to match the two sets of features (the first feature and the second feature). The cleaning robot will calculate the Euclidean distance between each feature point in the first feature and all feature points in the second feature respectively, and judge whether two feature points can be matched by comparing the magnitude of the Euclidean distance with the preset distance threshold. If the calculated Euclidean distance is less than the preset distance threshold, it is considered that these two feature points are the projections of the same physical point in the two photovoltaic park pictures and can be regarded as matching feature points. By this method, the cleaning robot will find a series of paired matching feature points in the two photovoltaic park pictures.

[0037] The embodiments of this application take any two photovoltaic park pictures as examples to illustrate the process of feature extraction and matching. However, in actual applications, the cleaning robot needs to process all the scanned photovoltaic park pictures, rather than just selecting two. The cleaning robot will repeat the above matching process until all the photovoltaic park pictures are matched. Through this pairwise matching method, the cleaning robot finally calculates the matching feature points between all the photovoltaic park pictures.

[0038] These matching feature points carry the connection information between the shooting perspectives of all the photovoltaic park pictures. Based on the matching feature points, the cleaning robot can estimate the relative pose between two photovoltaic park pictures through matrix transformation, that is, the relative rotation and translation amounts between the two shooting perspectives. According to the relative pose, triangulation measurement is performed on the matching feature point pairs, and the three-dimensional spatial coordinates of these matching feature points can be calculated to obtain their three-dimensional point cloud data.

[0039] For example, the cleaning robot obtains two photovoltaic park pictures, denoted as Picture A and Picture B respectively. The cleaning robot performs feature extraction and matching on the two pictures and finds a set of matching feature point pairs. For example, the feature points a1, a2, a3 in Picture A correspond one-to-one with the feature points b1, b2, b3 in Picture B, forming three pairs of matching feature points (a1, b1), (a2, b2), (a3, b3). These matching feature points carry the connection information between the two shooting perspectives and reflect the corresponding relationship between Picture A and Picture B.

[0040] The cleaning robot uses the matching feature points to estimate the relative pose between the two pictures. The relative pose includes the relative rotation matrix R and the relative translation vector t, which represent the rotation and translation amounts of the shooting perspective of Picture B relative to the shooting perspective of Picture A. The cleaning robot can calculate the values of R and t by solving the fundamental matrix or the essential matrix. For example, through matrix operations, the cleaning robot estimates that the shooting perspective of Picture B relative to Picture A rotates 10 degrees around the X-axis, 20 degrees around the Y-axis, 30 degrees around the Z-axis, and at the same time translates 0.5 meters in the X direction, 0.8 meters in the Y direction, and 1.2 meters in the Z direction.

[0041] The cleaning robot performs triangulation measurement on the matched feature points according to the estimated relative pose \(R\) and \(t\), and calculates their three-dimensional spatial coordinates. The triangulation measurement projects the pixel coordinates of the matched feature points back into the three-dimensional space and solves the disparity equation to obtain the depth information and three-dimensional coordinates of the matched feature points. For example, for the matched feature point pair \((a1, b1)\), the cleaning robot calculates the coordinates of this feature point in the three-dimensional space as \((X1, Y1, Z1)\) through triangulation measurement based on the pixel coordinates of \(a1\) and \(b1\), combined with the internal parameter matrix and relative pose matrix of the camera. Similarly, the cleaning robot performs triangulation measurement on the matched feature point pairs such as \((a2, b2)\), \((a3, b3)\), etc., and obtains their coordinates \((X2, Y2, Z2)\), \((X3, Y3, Z3)\) in the three-dimensional space. The cleaning robot aggregates the three-dimensional coordinates of all the matched feature points together to generate a three-dimensional point cloud data. Each three-dimensional point corresponds to a matched feature point and carries the spatial position information of this point. By performing the above triangulation process on all the matched feature points, the cleaning robot can finally obtain a three-dimensional point cloud model.

[0042] Step S103: Obtain the position information of the polysilicon photovoltaic module according to the three-dimensional point cloud model.

[0043] In step S103, to obtain the position information of the polysilicon photovoltaic module according to the three-dimensional point cloud model, it specifically includes: performing clustering segmentation on the three-dimensional point cloud model to obtain multiple point cloud clusters; calculating the height and width of each point cloud cluster; determining the target point cloud cluster among the multiple point cloud clusters, where the target point cloud cluster is a point cloud cluster with a height greater than a preset height threshold and a width greater than a preset width threshold among the multiple point cloud clusters; obtaining the point cloud cluster information of the target point cloud cluster, and using the point cloud cluster information as the position information of the polysilicon photovoltaic module, where the point cloud cluster information includes the position, height, and width of the target point cloud cluster.

[0044] Specifically, the cleaning robot performs clustering segmentation on the entire three-dimensional point cloud model. The purpose of clustering segmentation is to divide the points in the three-dimensional point cloud model into multiple subsets according to certain similarity criteria, and each subset is called a point cloud cluster. The K-means clustering based on distance is adopted in this application. The steps of clustering segmentation are described below: The cleaning robot first randomly selects \(K\) points as the initial clustering centers, then traverses each point in the three-dimensional point cloud model, calculates its Euclidean distance to each clustering center, and classifies it into the category of the nearest clustering center. Next, the cleaning robot recalculates the geometric center of each category and updates it as the new clustering center. The above process is continuously iterated until the clustering centers no longer change significantly or reach the maximum number of iterations. Finally, the three-dimensional point cloud model is divided into \(K\) point cloud clusters, and each point cloud cluster represents an independent geometric entity.

[0045] After the segmentation is completed, the cleaning robot analyzes each point cloud cluster and calculates its height and width. Specifically, for each point cloud cluster, the highest point and the lowest point in the three-dimensional space are found, and the height difference between the two is the height of the point cloud cluster. Similarly, the leftmost point and the rightmost point of the point cloud cluster in the horizontal direction are found, and the distance between the two is the width of the point cloud cluster. In this way, the cleaning robot can obtain the height and width of each point cloud cluster. The cleaning robot determines the target point cloud clusters from all the point cloud clusters according to a preset height threshold and a preset width threshold. The target point cloud clusters refer to the point cloud clusters with a height greater than the preset height threshold and a width greater than the preset width threshold. These point cloud clusters correspond to the polysilicon photovoltaic modules in the photovoltaic park. The preset height threshold and the preset width threshold can be set according to the size and passing ability of the cleaning robot, and this application does not limit this.

[0046] Finally, the cleaning robot extracts the information of the target point cloud clusters as the position information of the polysilicon photovoltaic modules. Specifically, the cleaning robot calculates the centroid coordinates of each target point cloud cluster as the position of the polysilicon photovoltaic module. At the same time, the height and width information of the target point cloud clusters are also extracted, and together with the position, they form the complete position information, including the position, height, and width of the polysilicon photovoltaic module.

[0047] For example, assume that the cleaning robot constructs a three-dimensional point cloud model of the photovoltaic park. This three-dimensional point cloud model contains 10,000 points. The cleaning robot first uses the K-means clustering algorithm to perform clustering segmentation on it, dividing the point cloud into 50 point cloud clusters. Then, the cleaning robot calculates the height and width of each point cloud cluster and finds that 5 of them have a height exceeding 0.2 meters and 3 of them have a width exceeding 0.5 meters. The cleaning robot determines these 8 point cloud clusters as the target point cloud clusters, extracts their centroid coordinates as the positions of the polysilicon photovoltaic modules, and at the same time records their heights and widths. Finally, the cleaning robot obtains the position information of 8 polysilicon photovoltaic modules, and the position information of each polysilicon photovoltaic module includes its position in the three-dimensional space, as well as its height and width.

[0048] Step S104: Starting from the current position, perform stratified random sampling on the ground of the park in the three-dimensional point cloud model to generate multiple alternative navigation points. Among them, the first sampling density is used in the ground area within a preset distance from the target polysilicon photovoltaic module, and the second sampling density is used in the ground area not within the preset distance from the target polysilicon photovoltaic module. The first sampling density is greater than the second sampling density, and the target polysilicon photovoltaic module is any one of the multiple polysilicon photovoltaic modules.

[0049] In step S104, starting from the current position, perform stratified random sampling on the park ground in the 3D point cloud model to generate multiple alternative navigation points, specifically including: dividing the park ground in the 3D point cloud model into multiple grid regions; determining the center point coordinates of each grid region; calculating the distance between the center point coordinates of each grid region and the target polysilicon photovoltaic module; marking the grid regions whose distance from the target polysilicon photovoltaic module is within the preset distance as the first grid regions, and marking the grid regions whose distance from the target polysilicon photovoltaic module is not within the preset distance as the second grid regions; performing random sampling on the first grid regions with the first sampling density and performing random sampling on the second grid regions with the second sampling density.

[0050] Specifically, the cleaning robot divides the park ground in the 3D point cloud model into multiple grid regions. The size of the grid regions can be set according to actual needs. For example, the park ground can be divided into 1m×1m grids, or the grid size can be determined according to the size of the polysilicon photovoltaic module to ensure that at least one complete polysilicon photovoltaic module is included in each grid. This application does not make any limitations in this regard.

[0051] After dividing the grids, the cleaning robot determines the center point coordinates of each grid region. The center point coordinates can be obtained by calculating the average value of the boundary point coordinates of the grid region. For example, for a rectangular grid composed of four vertices (x 1 , y 1 ), (x 2 , y 2 ), (x 3 , y 3 ), (x 4 , y 4 ), its center point coordinates can be expressed as ((x 1 + x 2 + x 3 + x 4 ) / 4, (y 1 + y 2 + y 3 + y 4) / 4). Next, the cleaning robot calculates the distances between the center point coordinates of each grid area and the target polysilicon photovoltaic module. The target polysilicon photovoltaic module can be any one of multiple polysilicon photovoltaic modules, and its position information can be obtained from step S103. The distance can be calculated using the Euclidean distance formula, that is, the straight-line distance between two points. After calculating the distances between the center point coordinates of each grid area and the target polysilicon photovoltaic module, the cleaning robot marks the grid areas within the preset distance as the first grid areas, and marks the grid areas not within the preset distance as the second grid areas. The preset distance can be set according to actual needs. For example, the preset distance can be set to 10m, that is, the grid areas within 10m of the target polysilicon photovoltaic module are marked as the first grid areas, and the remaining grid areas are marked as the second grid areas. In the specific implementation process, when performing stratified random sampling on the park ground in the three-dimensional point cloud model, in fact, it radiates centered on the target polysilicon photovoltaic module, and divides the park ground into two areas according to the distance from the target polysilicon photovoltaic module: the first grid area and the second grid area. If there is an overlap in the radiation areas of multiple polysilicon photovoltaic modules, that is, a certain area may be the first grid area corresponding to one polysilicon photovoltaic module and the second grid area corresponding to another polysilicon photovoltaic module at the same time, then this area is preferentially determined as the first grid area.

[0052] Finally, the cleaning robot performs random sampling on the first grid area and the second grid area respectively with different sampling densities. Specifically, the first sampling density is used for the first grid area, and the second sampling density is used for the second grid area, and the first sampling density is greater than the second sampling density. The setting of the sampling density can be adjusted according to actual needs. For example, the first sampling density can be set to 10 sampling points per square meter, and the second sampling density can be set to 2 sampling points per square meter. Within each grid area, the cleaning robot randomly generates a specified number of sampling points and uses these sampling points as alternative navigation points.

[0053] For example, assume that in the three-dimensional point cloud model of a certain photovoltaic park, the park ground is divided into 100 1m×1m grid areas, the center point coordinates of the target polysilicon photovoltaic module are (5, 5), the preset distance is 3m, the first sampling density is 8 sampling points per square meter, and the second sampling density is 2 sampling points per square meter. For the grid area with center point coordinates (4, 4), its distance from the target polysilicon photovoltaic module is sqrt((4 - 5) 2 +(4 - 5) 2) = sqrt(2) ≈ 1.414 m, which is less than the preset distance of 3 m. Therefore, this grid area is marked as the first grid area, and the cleaning robot randomly generates 8 sampling points in this grid area as alternative navigation points. For the grid area with the center point coordinates of (10, 10), its distance from the target polysilicon photovoltaic module is sqrt((10 - 5) 2 +(10 - 5) 2 ) = sqrt(50) ≈ 7.071 m, which is greater than the preset distance of 3 m. Therefore, this grid area is marked as the second grid area, and the cleaning robot randomly generates 2 sampling points in this grid area as alternative navigation points.

[0054] Step S105: Based on multiple alternative navigation points, multiple target navigation points are selected, and based on the multiple target navigation points, a travel route is generated.

[0055] In step S105, based on multiple alternative navigation points, multiple target navigation points are selected, and based on the multiple target navigation points, a travel route is generated, which specifically includes: inputting the position information of the polysilicon photovoltaic module into a preset navigation range model to obtain the corresponding target range. The preset navigation range model includes the corresponding relationship between the position information of the polysilicon photovoltaic module and the optimal navigation distance; determining the alternative navigation points within the target range as target navigation points; calculating the distances and azimuth angles between the respective target navigation points; based on a preset navigation point selection strategy, determining a first navigation point and a second navigation point from the target navigation points. The first navigation point and the second navigation point are any two target navigation points among the multiple target navigation points, where the second navigation point is the target navigation point among the multiple target navigation points that is the closest to the first navigation point or has the smallest azimuth angle; connecting the first navigation point and the second navigation point to generate a travel route.

[0056] Specifically, the cleaning robot inputs the position information of the polysilicon photovoltaic modules into a preset navigation range model, which contains the corresponding relationship between the position information of the polysilicon photovoltaic modules and the optimal navigation distance. Through this model, the cleaning robot can obtain the target navigation range corresponding to each polysilicon photovoltaic module. Then, the cleaning robot compares all the alternative navigation points with the target navigation ranges of each polysilicon photovoltaic module, and determines the alternative navigation points within the target navigation range as the target navigation points. This can ensure that when the cleaning robot cleans each polysilicon photovoltaic module, it can select the most suitable navigation point, neither being too far from the module resulting in poor cleaning effect nor being too close to the module resulting in a collision risk. Next, the cleaning robot calculates the distances and azimuth angles between all the target navigation points. The distance can be calculated by the Euclidean distance formula, and the azimuth angle can be calculated by the arctangent function. The calculation results can form a distance matrix and an azimuth angle matrix for subsequent navigation point selection and path generation. After determining the target navigation points, the cleaning robot selects appropriate navigation points from them as the starting point and the ending point of the travel route of the cleaning robot. Here, a preset navigation point selection strategy is adopted. Specifically, the cleaning robot first randomly selects a target navigation point as the first navigation point, and then among the remaining target navigation points, selects the navigation point with the shortest distance or the smallest azimuth angle from the first navigation point as the second navigation point. This can ensure that the travel route between the first navigation point and the second navigation point is as short as possible, and at the same time is as consistent as possible with the arrangement direction of the polysilicon photovoltaic modules, reducing unnecessary turning and detouring. Finally, the cleaning robot connects the first navigation point and the second navigation point to generate the travel route of the cleaning robot.

[0057] For example, assume that there are 100 polysilicon photovoltaic modules in a photovoltaic park. Through the previous steps, the cleaning robot obtains 500 alternative navigation points. After screening by the preset navigation range model, 200 target navigation points are finally obtained. The cleaning robot randomly selects the target navigation point numbered 87 as the first navigation point, and then calculates its distances and azimuth angles from the other 199 target navigation points, and finds that the target navigation point numbered 92 has the shortest distance and the smallest azimuth angle, so it is taken as the second navigation point. The cleaning robot connects these two navigation points to generate a travel route with a length of 15 meters and basically parallel to the arrangement direction of the polysilicon photovoltaic modules. The cleaning robot can clean the polysilicon photovoltaic modules near the two navigation points numbered 87 and 92 along this route.

[0058] The construction process of the preset navigation range model is as follows: First, use lidar and GPS sensors to collect a large amount of terrain data of the photovoltaic park, including the position coordinates, height, and tilt angle information of polysilicon photovoltaic modules. Then, perform preprocessing operations such as denoising, filtering, and interpolation on the collected data to improve the data quality and reliability. Extract features related to the navigation range from the preprocessed data, such as the length, width, height, tilt angle, and spacing of polysilicon photovoltaic modules, and use these features as input variables for the preset navigation range model. According to expert experience and actual tests, label the navigation range for the collected data. Specifically, for each polysilicon photovoltaic module, label an optimal navigation range centered on it, and the navigation points within this navigation range can ensure the cleaning effect and safety of the cleaning robot. The size of the navigation range can be dynamically adjusted according to factors such as the size of the module, tilt angle, and surrounding environment. According to the labeled data, select a suitable machine learning model to fit the corresponding relationship between the position information of polysilicon photovoltaic modules and the corresponding optimal navigation distance. The machine learning model can be a decision tree, support vector machine, or neural network. Divide the labeled data into a training set and a test set, use the training set to train the selected model, and improve the fitting ability and generalization ability of the model by adjusting the hyperparameters, feature selection, regularization, etc. of the model. During the training process, it is necessary to evaluate and optimize the performance of the model to ensure the accuracy and robustness of the model. Finally, use the test set to evaluate the trained model and calculate performance indicators such as the accuracy, recall rate, and F1 score of the model. Deploy the trained model to the cleaning robot to provide real-time navigation range prediction services. When new position information of polysilicon photovoltaic modules is input, the preset navigation range model can quickly predict its corresponding optimal navigation distance and generate a navigation range.

[0059] For example, assume that a photovoltaic park covers an area of 1 square kilometer and 10,000 polysilicon photovoltaic modules are installed in it. The three-dimensional point cloud data of these modules are collected through lidar and GPS. Each point cloud contains information such as position coordinates, height, and normal vector. After preprocessing and feature extraction, feature vectors such as the length, width, height, tilt angle, and density of surrounding obstacles of each module are obtained. Then, experts, based on experience and on-site tests, label the optimal navigation ranges for 1,000 randomly selected modules, obtaining 1,000 sample data. Next, a support vector regression model is selected to train these 1,000 sample data, and the optimal model hyperparameters are selected through grid search and cross-validation. The trained model achieves 95% accuracy and 90% recall rate on the test set, proving its good performance and generalization ability. Finally, this model is deployed to the cleaning robot to predict the position information of the remaining 9,000 modules, obtaining their optimal navigation distances and navigation ranges. The cleaning robot can use these navigation ranges to plan the travel route and cleaning tasks.

[0060] Step S106: Obtain the cleaning height of the cleaning robot, and generate a cleaning path by integrating the position information, tilt angle information, and cleaning height.

[0061] In step S106, generating a cleaning path by integrating the position information, tilt angle information, and cleaning height specifically includes: determining the horizontal position coordinates of the cleaning robot at each cleaning position according to the position information and travel route; calculating the tilt angle of the surface of the polysilicon photovoltaic module at each cleaning position according to the tilt angle information, and determining the cleaning angle of the cleaning brush of the cleaning robot according to the tilt angle so that the cleaning angle is consistent with the normal direction of the surface of the polysilicon photovoltaic module; constructing a path planning graph based on graph theory, taking the positions of each polysilicon photovoltaic module as the nodes of the path planning graph, taking the distances between each adjacent polysilicon photovoltaic module as the weights of the edges between each node in the path planning graph, and taking the horizontal position coordinates, cleaning angle, and cleaning height as the attributes of each node; using the Dijkstra algorithm to search for the shortest path in the path planning graph to obtain the cleaning path.

[0062] Specifically, the cleaning robot determines the horizontal position coordinates of the cleaning robot at each cleaning position according to the position information and the travel route. The cleaning position refers to the position where the cleaning robot needs to stop and operate during the cleaning process, and they are usually located directly below or on the side of the polysilicon photovoltaic module. The cleaning robot can calculate the horizontal position coordinates of the cleaning robot at each cleaning position in the three-dimensional point cloud model based on the position information of the polysilicon photovoltaic module and in combination with the travel route. Then, the cleaning robot calculates the tilt angle of the surface of the polysilicon photovoltaic module at each cleaning position according to the tilt angle information, and determines the cleaning angle of the cleaning brush of the cleaning robot according to the tilt angle. The tilt angle of the surface of the polysilicon photovoltaic module can be measured by a three-dimensional point cloud model or a sensor. The cleaning angle refers to the angle between the rotation axis of the cleaning brush and the horizontal plane, which determines the tilt range of the surface of the polysilicon photovoltaic module that the cleaning brush can effectively clean. In order to achieve the best cleaning effect, the cleaning angle should be consistent with the normal direction of the surface of the polysilicon photovoltaic module. The cleaning robot can calculate the cleaning angle using trigonometric functions according to the tilt angle.

[0063] Then, the cleaning robot constructs a path planning graph based on graph theory, takes the positions of each polysilicon photovoltaic module as the nodes of the path planning graph, takes the distances between each adjacent polysilicon photovoltaic module as the weights of the edges between each node in the path planning graph, and takes the horizontal position coordinates, cleaning angle, and cleaning height as the attributes of each node. The path planning graph is a mathematical model used to describe and optimize paths, which consists of nodes and edges. Nodes represent key positions or states in the path, and edges represent transfer relationships or costs between nodes. In this step, the cleaning robot abstracts the position of each polysilicon photovoltaic module as a node and constructs edges according to the distances between them to form a complete path planning graph. The attributes of the nodes include horizontal position coordinates, cleaning angle, and cleaning height, and these attributes determine the specific operation parameters of the cleaning robot at this node.

[0064] Finally, the cleaning robot uses the Dijkstra algorithm to search for the shortest path in the path planning graph to obtain the cleaning path. The Dijkstra algorithm is a classic graph search algorithm that can find the shortest path from the starting point to the ending point in a weighted directed graph. In this step, the cleaning robot sets the starting point as the current position of the cleaning robot, sets the ending point as the position of the last polysilicon photovoltaic module, and then uses the Dijkstra algorithm to search for the shortest path from the starting point to the ending point in the path planning graph. The shortest path is the cleaning path, which consists of a series of nodes and edges. Each node corresponds to the position of a polysilicon photovoltaic module, and each edge represents the movement trajectory of the cleaning robot between two adjacent polysilicon photovoltaic modules.

[0065] For example, assume that there are 5 polysilicon photovoltaic modules in a photovoltaic park, and their position coordinates are (1, 1), (2, 3), (4, 2), (3, 5), (5, 4) respectively, and the tilt angles are 30°, 45°, 60°, 30°, 45° respectively. The current position of the cleaning robot is (0, 0), and the cleaning height is 1.5 m. First, according to the position information and the travel route, the cleaning robot calculates the horizontal position coordinates of the cleaning robot at each cleaning position. For example, directly below the polysilicon photovoltaic module (1, 1), the horizontal position coordinates of the cleaning robot are (1, 1). Next, based on the tilt angle information, the cleaning robot calculates the tilt angle of the surface of the polysilicon photovoltaic module at each cleaning position and determines the cleaning angle of the cleaning brush. For example, at the polysilicon photovoltaic module (1, 1), the tilt angle is 30°, so the cleaning angle should also be 30°. Then, the cleaning robot constructs a path planning graph, takes the positions of the 5 polysilicon photovoltaic modules as nodes, and numbers them 0, 1, 2, 3, 4 in the order of node input. The distance between node 0 and node 1 is sqrt((2 - 1) 2 +(3 - 1) 2 ) ≈ 2.24, the distance between node 1 and node 2 is sqrt((4 - 2) 2 +(2 - 3) 2 ) ≈ 2.24, and so on. The cleaning robot calculates the distances between all nodes as the weights of the edges. The attributes of node 0 are (1, 1, 30°, 1.5), the attributes of node 1 are (2, 3, 45°, 1.5), and so on. The cleaning robot takes the horizontal position coordinates, cleaning angle, and cleaning height as the attributes of each node. Finally, starting from node 0 and ending at node 4, the cleaning robot uses the Dijkstra algorithm to search for the shortest path in the path planning graph.

[0066] Step S107: Clean the polysilicon photovoltaic modules according to the travel route and the cleaning path.

[0067] In step S107, the cleaning robot sends the travel route and the cleaning path to the cleaning robot as the navigation basis for its autonomous operation. After the cleaning robot determines the travel route and path information, it will autonomously plan the specific trajectories of travel and cleaning according to its own positioning and perception capabilities. Next, the cleaning robot starts to move along the travel route. During the movement, the cleaning robot will continuously monitor its own position and posture and make real-time corrections and adjustments according to the route information. When the cleaning robot reaches the first cleaning position, it will start to clean the corresponding polysilicon photovoltaic module according to the cleaning path information. The cleaning robot will adjust the cleaning angle and cleaning height of the cleaning brush according to the attribute information of the nodes to adapt to the inclination of the surface of the polysilicon photovoltaic module. After completing the cleaning of the first polysilicon photovoltaic module, the cleaning robot will continue to clean other polysilicon photovoltaic modules in sequence according to the travel route. During the transfer to the next cleaning position, the cleaning robot will navigate according to the travel route.

[0068] Referring to Figure 2 , the present application also provides a path determination device for a polysilicon photovoltaic cleaning robot. The device is a cleaning robot, and the cleaning robot includes a scanning and photographing module 201, a model construction module 202, a position information determination module 203, and a cleaning path determination module 204; the scanning and photographing module 201 is used to perform multi-directional scanning on the photovoltaic park to obtain multiple photovoltaic park pictures with different perspectives; the model construction module 202 is used to construct a three-dimensional point cloud model corresponding to the photovoltaic park according to the multiple photovoltaic park pictures; the position information determination module 203 is used to obtain the position information and inclination angle information of multiple polysilicon photovoltaic modules according to the three-dimensional point cloud model; the cleaning path determination module 204 is used to perform stratified random sampling on the ground of the park in the three-dimensional point cloud model with the current position as the starting point to generate multiple alternative navigation points, wherein, the first sampling density is adopted in the ground area within a preset distance from the target polysilicon photovoltaic module, and the second sampling density is adopted in the ground area not within the preset distance from the target polysilicon photovoltaic module, the first sampling density is greater than the second sampling density, and the target polysilicon photovoltaic module is any one of the multiple polysilicon photovoltaic modules; the cleaning path determination module 204 is further used to screen and obtain multiple target navigation points based on the multiple alternative navigation points, and generate a travel route based on the multiple target navigation points; the cleaning path determination module 204 is further used to obtain the cleaning height of the cleaning robot, and generate a cleaning path by integrating the position information, the inclination angle information, and the cleaning height; the cleaning path determination module 204 is further used to clean the polysilicon photovoltaic module according to the travel route and the cleaning path.

[0069] In a possible implementation, the cleaning path determination module 204 takes the current position as the starting point and performs stratified random sampling on the park ground in the 3D point cloud model to generate multiple alternative navigation points, specifically including: the cleaning path determination module 204 divides the park ground in the 3D point cloud model into multiple grid regions; the cleaning path determination module 204 determines the center point coordinates of each grid region; the cleaning path determination module 204 calculates the distance between the center point coordinates of each grid region and the target polysilicon photovoltaic module; the cleaning path determination module 204 marks the grid regions within a preset distance from the target polysilicon photovoltaic module as the first grid regions, and marks the grid regions not within the preset distance from the target polysilicon photovoltaic module as the second grid regions; the cleaning path determination module 204 performs random sampling on the first grid regions with a first sampling density and performs random sampling on the second grid regions with a second sampling density to obtain multiple alternative navigation points.

[0070] In a possible implementation, the cleaning path determination module 204 filters out multiple target navigation points based on the multiple alternative navigation points, and generates a travel route based on the multiple target navigation points, specifically including: the cleaning path determination module 204 inputs the position information of the polysilicon photovoltaic module into a preset navigation range model to obtain a corresponding target range, and the preset navigation range model includes the corresponding relationship between the position information of the polysilicon photovoltaic module and the optimal navigation distance; the cleaning path determination module 204 determines the alternative navigation points within the target range as the target navigation points; calculates the distance and azimuth angle between each target navigation point; the cleaning path determination module 204 determines a first navigation point and a second navigation point from the target navigation points based on a preset navigation point selection strategy, where the first navigation point and the second navigation point are any two target navigation points among the multiple target navigation points, and the second navigation point is the target navigation point among the multiple target navigation points that is the closest to the first navigation point or has the smallest azimuth angle; the cleaning path determination module 204 connects the first navigation point and the second navigation point to generate a travel route.

[0071] In a possible implementation, the cleaning path determination module 204 synthesizes the position information, tilt angle information, and cleaning height to generate a cleaning path, which specifically includes: the cleaning path determination module 204 determines the horizontal position coordinates of the cleaning robot at each cleaning position according to the position information and the traveling route; the cleaning path determination module 204 calculates the tilt angle of the surface of the polysilicon photovoltaic module at each cleaning position according to the tilt angle information, and determines the cleaning angle of the cleaning brush of the cleaning robot according to the tilt angle, so that the cleaning angle is consistent with the normal direction of the surface of the polysilicon photovoltaic module; the cleaning path determination module 204 constructs a path planning graph based on graph theory, uses the positions of the polysilicon photovoltaic modules as the nodes of the path planning graph, uses the distances between adjacent polysilicon photovoltaic modules as the weights of the edges between the nodes in the path planning graph, and uses the horizontal position coordinates, cleaning angle, and cleaning height as the attributes of each node; the cleaning path determination module 204 uses the Dijkstra algorithm to search for the shortest path in the path planning graph to obtain the cleaning path.

[0072] In a possible implementation, the model construction module 202 constructs a three-dimensional point cloud model corresponding to the photovoltaic park according to multiple photovoltaic park pictures, which specifically includes: the model construction module 202 extracts features from the first photovoltaic park picture to obtain the first feature, and extracts features from the second photovoltaic park picture to obtain the second feature, where the first photovoltaic park picture and the second photovoltaic park picture are any two photovoltaic park pictures among the multiple photovoltaic park pictures; the model construction module 202 performs feature point matching on the first feature and the second feature to obtain matching feature points; the model construction module 202 estimates the relative pose between the first photovoltaic park picture and the second photovoltaic park picture according to the matching feature points; the model construction module 202 triangulates the matching feature point pairs according to the relative pose to obtain the three-dimensional point cloud data of the matching feature points; the model construction module 202 constructs a three-dimensional point cloud model corresponding to the photovoltaic park according to the three-dimensional point cloud data.

[0073] In a possible implementation, the model construction module 202 performs feature point matching on the first feature and the second feature to obtain matching feature points, which specifically includes: the model construction module 202 calculates the Euclidean distance between the first feature and the second feature; the model construction module 202 judges the magnitude relationship between the Euclidean distance and a preset distance threshold; if the model construction module 202 determines that the Euclidean distance is less than the preset distance threshold, the first feature and the second feature are used as matching feature points.

[0074] In a possible implementation, the location information determination module 203 obtains the location information of the polysilicon photovoltaic module according to the three-dimensional point cloud model, specifically including: the location information determination module 203 performs clustering segmentation on the three-dimensional point cloud model to obtain a plurality of point cloud clusters; the location information determination module 203 calculates the height and width of each point cloud cluster; the location information determination module 203 determines a target point cloud cluster among the plurality of point cloud clusters, where the target point cloud cluster is a point cloud cluster with a height greater than a preset height threshold and a width greater than a preset width threshold among the plurality of point cloud clusters; the location information determination module 203 obtains the point cloud cluster information of the target point cloud cluster and uses the point cloud cluster information as the location information of the polysilicon photovoltaic module, and the point cloud cluster information includes the location, height, and width of the target point cloud cluster.

[0075] It should be noted that: when the device provided in the above embodiment realizes its functions, only the above division of each functional module is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be seen in the method embodiment, which will not be repeated here.

[0076] This application also provides an electronic device. Refer to Figure 3 , Figure 3 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of this application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0077] Among them, the communication bus 302 is used to realize the connection and communication between these components.

[0078] Among them, the user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may further include a standard wired interface and a wireless interface.

[0079] Among them, the network interface 304 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface).

[0080] Among them, the processor 301 may include one or more processing cores. The processor 301 connects various parts within the entire server through various interfaces and lines, and executes various functions of the server and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling the data stored in the memory 305. Optionally, the processor 301 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 301 may integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 301 and may be implemented separately by a single chip.

[0081] Among them, the memory 305 may include random access memory (RAM) and may also include read-only memory. Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store the data involved in the above-mentioned various method embodiments. Optionally, the memory 305 may also be at least one storage device located far from the aforementioned processor 301. Refer to Figure 3 , the memory 305, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a path determination method of a polysilicon photovoltaic cleaning robot.

[0082] In Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for the user to input and obtain the data input by the user; and the processor 301 can be used to call the application program stored in the memory 305 for a method of determining the path of a polysilicon photovoltaic cleaning robot. When executed by one or more processors 301, the electronic device 300 is caused to execute one or more of the methods as described in the foregoing embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0083] The present application also provides a computer-readable storage medium storing instructions. When executed by one or more processors 301, the electronic device 300 is caused to execute one or more of the methods as described in the foregoing embodiments.

[0084] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0085] In several implementation manners provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some service interfaces. The indirect couplings or communication connections of the devices or units can be in electrical or other forms.

[0086] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087] In addition, in each embodiment of the present application, the functional units can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0088] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of this application. And the aforementioned memory includes: various media such as USB flash drives, mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0089] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. After considering the specification and the disclosure of the practical truth, those skilled in the art will easily think of other implementation schemes of the present disclosure.

[0090] This application aims to cover any variations, uses, or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include the common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for determining a path of a polysilicon photovoltaic cleaning robot, characterized in that: The method comprises: Scan the photovoltaic park from multiple angles to obtain multiple pictures of the photovoltaic park from different perspectives; Constructing a three-dimensional point cloud model corresponding to the photovoltaic park according to the plurality of photovoltaic park pictures; According to the three-dimensional point cloud model, position information and tilt angle information of multiple polycrystalline silicon photovoltaic modules are obtained; Taking the current position as the starting point, performing layered random sampling on the ground of the park in the three-dimensional point cloud model to generate multiple candidate navigation points, wherein a first sampling density is used in a ground area within a preset distance from a target polycrystalline silicon photovoltaic module, and a second sampling density is used in a ground area not within the preset distance from the target polycrystalline silicon photovoltaic module, the first sampling density is greater than the second sampling density, and the target polycrystalline silicon photovoltaic module is any one of the multiple polycrystalline silicon photovoltaic modules; Based on the multiple candidate navigation points, multiple target navigation points are screened and obtained, and based on the multiple target navigation points, a travel route is generated; Acquire the cleaning height of the cleaning robot, and generate a cleaning path by integrating the position information, the tilt angle information and the cleaning height; The polycrystalline silicon photovoltaic module is cleaned according to the traveling route and the cleaning path.

2. The method according to claim 1, characterized in that The method of taking the current position as the starting point and performing layered random sampling on the park ground in the three-dimensional point cloud model to generate multiple candidate navigation points specifically includes: Dividing the park ground in the three-dimensional point cloud model into a plurality of grid areas; Determine the center point coordinates of each of the grid areas; Calculating the distance between the center point coordinates of each of the grid areas and the target polycrystalline silicon photovoltaic module; Marking a grid area within a preset distance from the target polycrystalline silicon photovoltaic module as a first grid area, and marking a grid area not within the preset distance from the target polycrystalline silicon photovoltaic module as a second grid area; The first grid area is randomly sampled using the first sampling density, and the second grid area is randomly sampled using the second sampling density to obtain the plurality of candidate navigation points.

3. The method according to claim 1, characterized in that The method of screening a plurality of target navigation points based on the plurality of candidate navigation points, and generating a travel route based on the plurality of target navigation points, specifically includes: Inputting the location information of the polycrystalline silicon photovoltaic module into a preset navigation range model to obtain a corresponding target range, wherein the preset navigation range model includes a corresponding relationship between the location information of the polycrystalline silicon photovoltaic module and an optimal navigation distance; Determine the candidate navigation point within the target range as the target navigation point; Calculating the distance and azimuth between each of the target navigation points; Based on a preset navigation point selection strategy, determine a first navigation point and a second navigation point from the target navigation points, wherein the first navigation point and the second navigation point are any two target navigation points among the plurality of target navigation points, wherein the second navigation point is the target navigation point that is closest to the first navigation point or has the smallest azimuth angle among the plurality of target navigation points; The first navigation point and the second navigation point are connected to generate the travel route.

4. The method according to claim 1, characterized in that The step of generating a cleaning path by integrating the position information, the tilt angle information and the cleaning height specifically includes: Determining the horizontal position coordinates of the cleaning robot at each cleaning position according to the position information and the travel route; According to the inclination angle information, the inclination angle of the surface of the polycrystalline silicon photovoltaic module at each cleaning position is calculated, and the cleaning angle of the cleaning brush of the cleaning robot is determined according to the inclination angle, so that the cleaning angle is consistent with the normal direction of the surface of the polycrystalline silicon photovoltaic module; Constructing a path planning graph based on graph theory, and using the positions of each polycrystalline silicon photovoltaic module as a node of the path planning graph, using the distance between each adjacent polycrystalline silicon photovoltaic module as the weight of the edge between each node in the path planning graph, and using the horizontal position coordinate, the cleaning angle, and the cleaning height as attributes of each node; The Dijkstra algorithm is used to search for the shortest path in the path planning graph to obtain the cleaning path.

5. The method according to claim 1, characterized in that The constructing a three-dimensional point cloud model corresponding to the photovoltaic park according to the plurality of photovoltaic park pictures specifically includes: Performing feature extraction on the first photovoltaic park picture to obtain a first feature, and performing feature extraction on the second photovoltaic park picture to obtain a second feature, wherein the first photovoltaic park picture and the second photovoltaic park picture are any two photovoltaic park pictures from the multiple photovoltaic park pictures; Performing feature point matching on the first feature and the second feature to obtain matching feature points; estimating a relative position between the first photovoltaic park image and the second photovoltaic park image according to the matching feature points; triangulate the matching feature point pairs according to the relative position and posture to obtain three-dimensional point cloud data of the matching feature points; A three-dimensional point cloud model corresponding to the photovoltaic park is constructed according to the three-dimensional point cloud data.

6. The method according to claim 5, characterized in that The performing feature point matching on the first feature and the second feature to obtain matching feature points specifically includes: Calculating the Euclidean distance between the first feature and the second feature; Determine the relationship between the Euclidean distance and a preset distance threshold; If it is determined that the Euclidean distance is less than the preset distance threshold, the first feature and the second feature are used as the matching feature points.

7. The method according to claim 1, characterized in that The step of obtaining the position information of the polycrystalline silicon photovoltaic module according to the three-dimensional point cloud model specifically includes: Performing clustering segmentation on the three-dimensional point cloud model to obtain a plurality of point cloud clusters; Calculate the height and width of each point cloud cluster; Determine a target point cloud cluster among the plurality of point cloud clusters, wherein the target point cloud cluster is a point cloud cluster whose height is greater than a preset height threshold and whose width is greater than a preset width threshold among the plurality of point cloud clusters; The point cloud cluster information of the target point cloud cluster is acquired, and the point cloud cluster information is used as the position information of the polycrystalline silicon photovoltaic module, wherein the point cloud cluster information includes the position, height and width of the target point cloud cluster.

8. A path determination device for a polysilicon photovoltaic cleaning robot, characterized in that: The device comprises a scanning and photographing module (201), a model building module (202), a position information determining module (203), and a cleaning path determining module (204); The scanning and shooting module (201) is used to perform multi-directional scanning on the photovoltaic park to obtain multiple pictures of the photovoltaic park from different viewing angles; The model building module (202) is used to build a three-dimensional point cloud model corresponding to the photovoltaic park based on the plurality of photovoltaic park pictures; The position information determination module (203) is used to obtain the position information and tilt angle information of multiple polycrystalline silicon photovoltaic modules according to the three-dimensional point cloud model; The cleaning path determination module (204) is used to perform layered random sampling on the park ground in the three-dimensional point cloud model with the current position as the starting point to generate multiple candidate navigation points, wherein a first sampling density is used in a ground area within a preset distance from a target polycrystalline silicon photovoltaic component, and a second sampling density is used in a ground area not within the preset distance from the target polycrystalline silicon photovoltaic component, the first sampling density is greater than the second sampling density, and the target polycrystalline silicon photovoltaic component is any one of the multiple polycrystalline silicon photovoltaic components; The cleaning path determination module (204) is further used to screen and obtain a plurality of target navigation points based on the plurality of candidate navigation points, and generate a travel route based on the plurality of target navigation points; The cleaning path determination module (204) is further used to obtain the cleaning height of the cleaning robot and generate a cleaning path by integrating the position information, the tilt angle information and the cleaning height; The cleaning path determination module (204) is also used to clean the polycrystalline silicon photovoltaic module according to the travel route and the cleaning path.

9. An electronic device, characterized in that: The electronic device (300) comprises a processor (301), a memory (305), a user interface (303) and a network interface (304), wherein the memory (305) is used to store instructions, the user interface (303) and the network interface (304) are used to communicate with other devices, and the processor (301) is used to execute the instructions stored in the memory (305) so that the electronic device (300) executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is performed.

Images