Photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system

Abstract

The invention relates to the technical field of photovoltaic module deformation detection, in particular to a photovoltaic module deformation detection method for intelligent routing inspection of a mountain photovoltaic system, and the method comprises the steps: obtaining abnormal points in each cluster; determining a deformation characteristic value by evaluating the similarity of normal vectors between each abnormal point and all abnormal points in the neighborhood of the abnormal point and the similarity of descriptors; fitting each abnormal point and all abnormal points in a neighborhood of the abnormal point, and constructing a filtering window length of each abnormal point in each cluster by analyzing an error condition in a fitting process and combining with the deformation characteristic value; and carrying out filtering processing on all abnormal points in all the clusters based on the filtering window length so as to carry out deformation detection on the to-be-detected photovoltaic module. The invention aims to improve the precision of performing deformation evaluation on the photovoltaic module for intelligent routing inspection of the mountain photovoltaic system by using the three-dimensional point cloud data of the photovoltaic module.

Description

Technical Field

[0001] This application relates to the technical field of photovoltaic module deformation detection, and specifically relates to a method for detecting deformation of photovoltaic modules for intelligent inspection of mountain photovoltaic systems. Background Art

[0002] A mountain photovoltaic system refers to a photovoltaic power generation system built on mountainous terrain. The complexity of the mountainous terrain increases the maintenance difficulty of photovoltaic modules. Therefore, it is necessary to regularly detect the deformation of photovoltaic modules to timely discover problems such as aging and damage of the modules, and perform timely maintenance and replacement to avoid a reduction in the power generation efficiency of the entire photovoltaic system due to a decline in the performance of the modules.

[0003] Although the drone measurement technology equipped with a three-dimensional laser scanner can obtain the three-dimensional structure information of the object surface to evaluate the overall deformation of the object, due to the fact that the solar cells on the surface of photovoltaic modules usually use colors with low reflectivity such as black or dark blue, a large amount of noise data is easily generated in the three-dimensional point cloud data of the photovoltaic modules collected by the three-dimensional laser scanner, thereby reducing the accuracy of using the three-dimensional point cloud data of the photovoltaic modules to evaluate the overall deformation of the photovoltaic modules. Summary of the Invention

[0004] In order to solve the above technical problems, this application provides a method for detecting deformation of photovoltaic modules for intelligent inspection of mountain photovoltaic systems to solve the existing problems.

[0005] The method for detecting deformation of photovoltaic modules for intelligent inspection of mountain photovoltaic systems in this application adopts the following technical solutions: An embodiment of this application provides a method for detecting deformation of photovoltaic modules for intelligent inspection of mountain photovoltaic systems. The method includes the following steps: S1: Use laser scanning to obtain the three-dimensional point cloud of the photovoltaic module to be measured; S2: Determine the filtering window length by analyzing the distribution characteristics of the data points in the three-dimensional point cloud. Specifically: S201: Obtain the reflectivity values of all data points in the three-dimensional point cloud, and based on the reflectivity values, screen out the feature points from all data points, cluster all feature points, and obtain the abnormal points in each clustering cluster; S202: Obtain the normal vector and descriptor of each abnormal point. Within each clustering cluster, determine the deformation feature value of each abnormal point within each clustering cluster by evaluating the similarity of the normal vectors and the similarity of the descriptors between each abnormal point and all abnormal points in its neighborhood; S203: Fit each outlier and all outliers within its neighborhood. By analyzing the error situation during the fitting process, determine the noise interference degree of each outlier within each cluster, and combine with the deformation eigenvalue to determine the filtering window adjustment coefficient of each outlier within each cluster, so as to construct the filtering window length of each outlier within each cluster; S3: Based on the filtering window length, perform filtering processing on all outliers in all clusters to obtain the denoised three-dimensional point cloud of the photovoltaic module to be detected. Obtain the three-dimensional point cloud of the standard photovoltaic module, and detect the deformation of the photovoltaic module to be measured by comparing the difference between the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module.

[0006] Preferably, the process of screening the feature points is as follows: Take the reflectivity values of all data points in the three-dimensional point cloud as the input of the threshold segmentation algorithm, output the segmentation threshold, and take all data points with reflectivity values less than the segmentation threshold as feature points.

[0007] Preferably, the outliers in each cluster are the results obtained by using an outlier detection algorithm for all feature points in each cluster.

[0008] Preferably, the metric distance in the clustering algorithm used in the process of clustering all feature points is the Euclidean distance between the three-dimensional coordinates of the feature points.

[0009] Preferably, the method for determining the deformation eigenvalue of each outlier within each cluster is as follows: In each cluster, calculate the mean value of the normal vector similarity and the mean value of the descriptor similarity between each outlier and all outliers within its neighborhood, and denote them as the first mean value and the second mean value. Take the reciprocal of the sum of the first mean value and the second mean value, and denote it as the deformation eigenvalue of each outlier within each cluster.

[0010] Preferably, the noise interference degree of each outlier within each cluster is the residual value during the surface fitting process of each outlier within each cluster and all outliers within its neighborhood.

[0011] Preferably, the filtering window adjustment coefficient of each outlier within each cluster is the ratio of the noise interference degree of each outlier within each cluster to the deformation eigenvalue.

[0012] Preferably, the expression of the filtering window length of each outlier within each cluster is: ; where represents the filtering window length of the j-th outlier in the i-th cluster; represents the normalized value of the filtering window adjustment coefficient of the j-th outlier in the i-th cluster; 、 respectively represent a preset first value and a preset second value; f( ) represents a custom function for outputting the even number closest to the function input value.

[0013] Preferably, the method for obtaining the denoised three-dimensional point cloud of the photovoltaic module to be detected is as follows: Take all the abnormal points in all the clustering clusters as the input of the filtering algorithm. Among them, take the filtering window length of each abnormal point as the size of the filtering window in the filtering algorithm, output all the denoised abnormal points, and denote them as corrected points. The three-dimensional point cloud composed of all the corrected points and all the data points that have not been filtered is used as the denoised three-dimensional point cloud of the photovoltaic module to be detected.

[0014] Preferably, the deformation detection of the photovoltaic module to be measured includes: Use the irregular triangular mesh method to construct the three-dimensional models of the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module respectively, and denote them as the denoised three-dimensional model and the standard three-dimensional model. Use the interpolation calculation method to calculate the interpolation of each pair of corresponding meshes between the denoised three-dimensional model and the standard three-dimensional model. Take the average value of the interpolation of all pairs of corresponding meshes as the deformation degree of the photovoltaic module to be measured.

[0015] This application has at least the following beneficial effects: This application screens out abnormal data points through the reflectivity values of the data points in the three-dimensional point cloud data. Compared with directly using the point cloud filtering algorithm to filter all the data in the collected three-dimensional point cloud data, it can effectively avoid the over-smoothing of the normal point cloud data on the surface of the battery cells in the photovoltaic module to be detected that is not affected by noise, thereby reducing its impact on the subsequent evaluation of the deformation of the battery cells on the surface of the photovoltaic module; further, by analyzing the local spatial distribution eigenvalue of the abnormal points, a suitable filtering window is constructed for each abnormal point, and the selection of the window size in the filtering algorithm is optimized based on the filtering window, so as to more accurately perform denoising processing on the local area where the abnormal points are located. While effectively filtering out the noise data in each battery cell area on the surface of the photovoltaic module to be detected in the three-dimensional point cloud data, it retains the detailed features of the point cloud data in the battery cell deformation area as much as possible, thereby improving the accuracy of the subsequent evaluation of the deformation of the battery cells on the surface of the photovoltaic module; further, based on the denoised three-dimensional point cloud data, the deformation detection of the photovoltaic module to be detected is realized. Compared with the existing methods, it can more comprehensively evaluate the deformation of the overall photovoltaic module and effectively reduce the impact of the noise data in the collected three-dimensional point cloud data of the photovoltaic module on the overall deformation evaluation of the photovoltaic module, thereby improving the accuracy of using the three-dimensional point cloud data of the photovoltaic module to evaluate the deformation of the photovoltaic modules in the intelligent inspection of the mountain photovoltaic system. Description of the Drawings

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a flowchart of the steps of a method for detecting deformation of a photovoltaic module in intelligent inspection of a mountain photovoltaic system provided by an embodiment of the present application; Figure 2 It is a schematic diagram of the process of obtaining the filtering window length provided by an embodiment of the present application. Detailed implementation manners

[0018] To further elaborate on the technical means and effects adopted by the present application to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a method for detecting deformation of a photovoltaic module in intelligent inspection of a mountain photovoltaic system proposed according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs.

[0020] The following specifically describes the specific solution of a method for detecting deformation of a photovoltaic module in intelligent inspection of a mountain photovoltaic system provided by the present application with reference to the accompanying drawings.

[0021] A method for detecting deformation of a photovoltaic module in intelligent inspection of a mountain photovoltaic system provided by an embodiment of the present application. Specifically, a method for detecting deformation of a photovoltaic module in intelligent inspection of a mountain photovoltaic system is provided as follows. Please refer to Figure 1 The method includes the following steps: S1: Obtain the three-dimensional point cloud of the photovoltaic module to be measured by laser scanning.

[0022] In this embodiment, a drone equipped with a three-dimensional laser scanner is used to inspect the photovoltaic modules in the mountain photovoltaic system to achieve regular deformation detection of the photovoltaic modules.

[0023] Use a drone equipped with a 3D laser scanner to collect the 3D point cloud of the photovoltaic module to be measured, and convert the collected 3D point cloud into the coordinate system of the laser scanner. Among them, the laser emission center of the 3D laser scanner is used as the origin, the starting direction of the scanning plane is used as the positive direction of the X-axis, the direction perpendicular to the scanning plane and upward is used as the Y-axis direction, and the positive direction of the Z-axis is determined by the principle of the right-hand coordinate system.

[0024] Among them, the principle of the right-hand coordinate system is a formula technique, and its specific steps will not be elaborated here.

[0025] S2: Determine the filtering window length by analyzing the distribution characteristics of the data points in the 3D point cloud.

[0026] To reduce the influence of noise data in the collected 3D point cloud, the 3D point cloud is usually smoothed using a point cloud filtering algorithm, such as: median filtering algorithm, mean filtering algorithm, or Gaussian filtering algorithm, etc. However, the existing data filtering methods usually process all the data points in the collected 3D point cloud, which easily causes the normal point cloud data that is not affected by noise in the solar cells on the surface of the photovoltaic module to be over-smoothed, thereby affecting the accuracy when evaluating the deformation of the solar cells on the surface of the photovoltaic module subsequently. Therefore, to avoid this situation, the following processing is carried out: S201: Obtain the reflectivity values of all the data points in the 3D point cloud, and based on the reflectivity values, screen out the feature points from all the data points, cluster all the feature points, and perform anomaly detection on all the feature points in each clustering cluster to obtain the anomaly points in each clustering cluster.

[0027] (1) Since the color of the solar cells on the surface of the photovoltaic module is usually black or dark blue, the solar cells on the surface of the photovoltaic module have a lower reflectivity compared to the junction boxes and metal frames on the surface. Therefore, obtain the reflectivity values of all the data points in the 3D point cloud, use the reflectivity values of all the data points in the 3D point cloud as the input of the threshold segmentation algorithm, output the segmentation threshold, and use all the data points with reflectivity values less than the segmentation threshold as the feature points, which are used to represent all the data points corresponding to all the solar cells on the surface of the photovoltaic module to be measured in the 3D point cloud.

[0028] It should be supplemented that in this embodiment, the maximum inter-class variance is used to classify the data points. In the actual application process, the implementer can also select other threshold segmentation methods according to the specific situation. Regarding the selection of the threshold segmentation method, this embodiment does not make special restrictions.

[0029] Among them, the maximum inter-class variance is a well-known technology, and its specific principle will not be elaborated here.

[0030] (2) Further, since the cells of the photovoltaic module are usually separated by other junction boxes or metal frames, the point cloud data corresponding to each cell on the three-dimensional point cloud of the photovoltaic module to be measured is relatively aggregated, while the point cloud data corresponding to different cells is relatively discrete. Therefore, all feature points are used as the input of the clustering algorithm. Among them, the distance between the three-dimensional coordinates of the feature points is used as the metric distance in the clustering algorithm, and the number of cells on the surface of the photovoltaic module to be measured is used as the number of clustering clusters, and multiple clustering clusters are output to represent different cells on the photovoltaic module.

[0031] It should be understood that in the actual application process, common clustering algorithms include k-means clustering algorithm, hierarchical clustering algorithm, mean shift clustering algorithm and other clustering algorithms. In this embodiment, the k-means clustering algorithm is selected to cluster the data points. As other implementation manners, the implementer can also select according to the specific situation by himself / herself, and this embodiment does not make special restrictions.

[0032] Among them, the k-means clustering algorithm is a well-known technology, and its specific clustering principle will not be elaborated here.

[0033] (3) Further, since there is a large spatial distribution difference between the normal data points and the noise data points in the three-dimensional point cloud of the photovoltaic module, the three-dimensional coordinates of all feature points in each clustering cluster are used as the input of the anomaly detection algorithm, and the anomaly points among all feature points in each clustering cluster are output.

[0034] Among them, in this embodiment, the LOF anomaly detection algorithm is used to detect anomalies in all feature points in each clustering cluster. In the actual application process, as other implementation manners, the implementer can also use other anomaly detection methods such as the isolation forest algorithm. Regarding the selection of the anomaly detection method, this embodiment does not make special restrictions. The LOF anomaly detection algorithm is a well-known technology, and its specific principle will not be elaborated here.

[0035] So far, by analyzing the spatial distribution characteristics of all data points in the three-dimensional point cloud, the anomaly points have been screened out from all data points.

[0036] S202: Obtain the normal vector and descriptor of each anomaly point. Within each clustering cluster, by evaluating the similarity of the normal vectors and the similarity of the descriptors between each anomaly point and all anomaly points in its neighborhood, the deformation eigenvalue of each anomaly point within each clustering cluster is determined.

[0037] Under normal circumstances, the point cloud data of the cells on the surface of the photovoltaic module is regularly distributed in space, and the distance between points is relatively uniform. However, when the cell undergoes a large deformation, the distribution of the point cloud data in the deformed area becomes irregular, the distance between points may change significantly, and the distribution density of the point cloud data in the deformed area may also change. For example, if the deformation causes local depression or protrusion on the cell surface, it will increase or decrease the density of the point cloud data in this area, forming a contrast with the density of the point cloud data in the normal area, so that the abnormal data points in the three-dimensional point cloud data of the photovoltaic module extracted by the anomaly detection algorithm not only include noise data, but may also include some point cloud data with the characteristics of cell deformation.

[0038] Therefore, in this embodiment, a larger filtering window is assigned to the noise data in the obtained abnormal data point set, so that it can be effectively filtered by the point cloud filtering algorithm, while a smaller filtering window is assigned to the point cloud data in the abnormal data point set that has the characteristics of cell deformation and is less affected by noise, so as to effectively retain the detailed characteristics of the point cloud data in the cell deformation area in the three-dimensional point cloud data of the photovoltaic module to be measured.

[0039] Under normal circumstances, the cell on the surface of the photovoltaic module is a plane, so that the different data points in the point cloud data in the area where the normal cell is located have relatively consistent normal vectors and relatively similar local spatial distribution characteristics. When the cell is deformed, the deformed area usually forms an irregular curved surface, so that the normal vectors and local spatial distribution characteristics of the different data points in the point cloud data in the deformed area have large differences.

[0040] Based on the above analysis, the normal vector and descriptor of each abnormal point are obtained respectively. In this embodiment, the descriptor is the FPFH descriptor. In actual application, as other implementation methods, the implementer can also use the PPF descriptor or the PFH descriptor, and this embodiment does not make special restrictions.

[0041] Among them, the acquisition of the normal vector and FPFH descriptor of the point cloud data are both well-known technologies, and their specific principles will not be elaborated here.

[0042] In each clustering cluster, the mean value of the normal vector similarity and the mean value of the descriptor similarity between each abnormal point and all abnormal points in its neighborhood are calculated respectively, and are denoted as the first mean value and the second mean value. The reciprocal of the sum of the first mean value and the second mean value is denoted as the deformation characteristic value of each abnormal point in each clustering cluster, which is used to evaluate whether the local surface area of the cell position where the abnormal point is located has deformation characteristics.

[0043] It should be noted that there are many methods to measure the similarity between vectors and descriptors. In this embodiment, cosine similarity is used as the calculation method for similarity. In actual application processes, as other implementation manners, implementers can also use other methods to measure similarity, such as the reciprocal of the Euclidean distance. Regarding the selection of the method for measuring similarity, this embodiment does not make special restrictions.

[0044] Among them, the calculation method of cosine similarity is a well-known technology, and its specific calculation process will not be elaborated here.

[0045] From the deformation characteristic values of each outlier within each clustering cluster, it can be understood that if the similarity of the normal vectors between an outlier and the outliers within its neighborhood is greater, it indicates that the outlier and the outliers within its neighborhood have similar geometric features. The distribution characteristics of the data points corresponding to the deformation of the photovoltaic module are usually relatively similar, while the distribution characteristics of the data points corresponding to noise are relatively chaotic and irregular. Therefore, the larger the first mean value, the more likely it is that the distribution characteristics of these outliers are caused by the deformation of the photovoltaic module. The greater the similarity of the descriptors between an outlier and the outliers within its neighborhood, that is, the larger the second mean value, it indicates that the outlier and the outliers within its neighborhood have similar local shape features. At this time, the local distribution characteristics of the outlier are more likely to be caused by the deformation of the photovoltaic module, and the obtained deformation characteristic value is also larger, indicating that the local surface area of the position of the cell where the outlier is located is more likely to have deformed; On the contrary, if the similarity of the normal vectors between an outlier and the outliers within its neighborhood is smaller, that is, the first mean value is smaller, it indicates that the geometric feature differences between the outlier and the outliers within its neighborhood are greater, and the distribution characteristics of these outliers are more likely to be caused by noise interference. The smaller the similarity of the descriptors between an outlier and the outliers within its neighborhood, that is, the smaller the second mean value, it indicates that the local shape feature differences between the outlier and the outliers within its neighborhood are greater. At this time, the local distribution characteristics of the outlier are more likely to be caused by external noise, and the obtained deformation characteristic value is smaller, indicating that the possibility of deformation of the local surface area of the position of the cell where the outlier is located is smaller, and the possibility of being affected by noise is greater.

[0046] So far, by analyzing the local geometric features of the outliers, the deformation characteristic values of the outliers have been obtained.

[0047] S203: Fit each outlier and all the outliers within its neighborhood, determine the noise interference degree of each outlier within each clustering cluster by analyzing the error situation during the fitting process, and combine the deformation characteristic value to determine the filtering window adjustment coefficient of each outlier within each clustering cluster, so as to construct the filtering window length of each outlier within each clustering cluster.

[0048] Noise usually causes a deviation in the spatial position of a data point in the collected point cloud data, causing it to deviate from its true position. The greater the influence of noise on a data point in the point cloud data within the area of the solar cell, the greater the deviation of the actual position of the solar cell where the data point is located from its true position, that is, the greater the deviation of the data point from the surface fitted by the point cloud data within its local area.

[0049] Based on the above analysis, in each clustering cluster, surface fitting is performed on each outlier and all outliers within its neighborhood. The residual value during the fitting process is recorded as the degree of noise interference of each outlier in each clustering cluster. The greater the residual value, that is, the greater the degree of noise interference, the greater the deviation of the actual position of the solar cell where the outlier is located from its true position, and the greater the degree of noise interference on the outlier. Then, the larger the filtering window of the outlier in the point cloud filtering algorithm should be to effectively filter out the noise contained in the data points.

[0050] It should be noted that there are many commonly used surface fitting algorithms. In this embodiment, a point cloud surface fitting algorithm based on the moving least squares method is used to perform surface fitting on outliers and all outliers within their neighborhoods. In actual application processes, as other implementation methods, implementers can also use other methods such as the random sample consensus algorithm (RANSAC). Regarding the selection of the surface fitting algorithm, this embodiment does not make special restrictions.

[0051] Among them, both the point cloud surface fitting algorithm based on the moving least squares method and the calculation of the residual value are well-known technologies, and their specific processes will not be elaborated here.

[0052] Furthermore, in each clustering cluster, the ratio of the degree of noise interference of each outlier to the deformation eigenvalue is recorded as the filtering window adjustment coefficient of each outlier within each clustering cluster, which is used to adjust the size of the filtering window of the outlier in the subsequent point cloud filtering algorithm. The larger the filtering window adjustment coefficient, the greater the degree of noise interference on the outlier, and the larger the filtering window of the outlier.

[0053] Normalize the filtering window adjustment coefficients of all outliers in each clustering cluster to obtain the normalized value of the filtering window coefficient of each outlier within each clustering cluster, which is used as the size of the filtering window of the outlier in the point cloud filtering algorithm when using the point cloud filtering algorithm to filter all outliers in the three-dimensional point cloud. Specifically: The filtering window length of the j-th outlier in the i-th clustering cluster The expression is: ; In the formula, represents the normalized value of the filtering window adjustment coefficient of the j-th outlier in the i-th clustering cluster; 、 respectively represent a preset first value and a preset second value; f( ) represents a custom function for outputting the even number closest to the function input value.

[0054] It should be noted that 、 are used to determine the upper and lower limits of the filter window size in the filtering algorithm. In this embodiment 、 are 6 and 3 respectively. In this embodiment, a relatively small lower limit = 3 can filter out noise while retaining the deformation details, and the upper limit = 6 can limit the window size from being too large to avoid losing the true change trend of the data due to over-filtering. Regarding 、 The values of can be set by the implementer himself, and there is no special limitation in this embodiment.

[0055] So far, by analyzing the local fitting degree of the abnormal points and combining the deformation eigenvalue in S202, the filter window length of the abnormal points is obtained, optimizing the denoising process in the subsequent process.

[0056] Preferably, the schematic diagram of the process for obtaining the filter window length provided in this embodiment is as shown in Figure 2 shown.

[0057] S3: Based on the filter window length, filter all the abnormal points in all the clustering clusters to obtain the denoised three-dimensional point cloud of the photovoltaic module to be detected, obtain the three-dimensional point cloud of the standard photovoltaic module, and detect the deformation of the photovoltaic module to be detected by comparing the difference between the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module.

[0058] Use the filtering algorithm to filter all the abnormal points in all the clustering clusters in the three-dimensional point cloud. Among them, the filter window length of each abnormal point is used as the size of the filter window in the filtering algorithm to filter the abnormal points. The filtered abnormal points are recorded as corrected points. The three-dimensional point cloud composed of all the corrected points and all the data points that have not been filtered is used as the denoised three-dimensional point cloud of the photovoltaic module to be detected, and is used as the denoised three-dimensional point cloud of the photovoltaic module to be detected.

[0059] It should be noted that there are many commonly used filtering algorithms. In this embodiment, the mean filtering algorithm is used to denoise the abnormal points. In the actual application process, as other implementation manners, the implementer can also use other filtering algorithms such as median filtering or Gaussian filtering algorithms according to the specific situation. Regarding the selection of the filtering algorithm, there is no special limitation in this embodiment.

[0060] Among them, the median filtering algorithm is a well-known technology, and its specific denoising principle will not be elaborated here.

[0061] ​Obtain the three-dimensional point cloud of a standard photovoltaic module with the same specifications as the photovoltaic module to be tested and passing standard certification. Use the irregular triangular mesh method to construct the three-dimensional models of the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module respectively, and denote them as the denoised three-dimensional model and the standard three-dimensional model. Unify the coordinate systems of these two three-dimensional models, and use the interpolation calculation method with the three-dimensional model of the standard photovoltaic module as the reference to calculate the interpolation of each pair of corresponding meshes between the denoised three-dimensional model and the standard three-dimensional model. Take the mean value of the interpolations of all pairs of corresponding meshes as the deformation degree of the photovoltaic module to be tested.

[0062] It should be noted that the interpolation calculation method can be used to measure the difference between the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module, that is, by calculating the interpolation of the corresponding meshes between the two three-dimensional point clouds to measure the difference between the three-dimensional point clouds, so as to compare the difference between the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module and judge the degree of deformation of the photovoltaic module to be tested.

[0063] It is further supplemented that there are many commonly used interpolation calculation methods. In this embodiment, the local polynomial interpolation method (LP) is adopted. In the actual application process, as other implementation manners, the implementer can also adopt other interpolation calculation methods such as the nearest neighbor interpolation method (NeN) or the moving average interpolation method (MA). Regarding the selection of the interpolation calculation method, no special limitation is made in this embodiment.

[0064] Among them, the irregular triangular mesh method and the local polynomial interpolation method (LP) are both well-known technologies, and their specific principle processes will not be elaborated here.

[0065] So far, in this embodiment, the three-dimensional point cloud of the photovoltaic module is obtained by a laser scanner, the local spatial distribution characteristics of the three-dimensional point cloud are further analyzed, the selection of the filtering window in the filtering algorithm is optimized, the three-dimensional point cloud is more accurately denoised, and thus the accuracy of the deformation evaluation of the photovoltaic module in the intelligent inspection of the mountain photovoltaic system using the three-dimensional point cloud data of the photovoltaic module is improved.

[0066] It should be noted that: the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification is made. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some implementation manners, multi-task processing and parallel processing are also possible or may be beneficial.

[0067] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0068] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it; modifying the technical solutions described in the foregoing embodiments, or equivalently replacing some of the technical features therein, does not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. A photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic systems, characterized in that: The method comprises the following steps: S1: Use laser scanning to obtain the three-dimensional point cloud of the photovoltaic module to be tested; S2: Determine the filter window length by analyzing the distribution characteristics of data points in the 3D point cloud, specifically: S201: Obtaining reflectivity values ​​of all data points in the three-dimensional point cloud, and based on the reflectivity values, screening feature points from all data points, clustering all feature points, and obtaining abnormal points in each cluster; S202: Obtain the normal vector and descriptor of each outlier point, and determine the deformation feature value of each outlier point in each cluster by evaluating the similarity of the normal vector between each outlier point and all outlier points in its neighborhood and the similarity of the descriptor in each cluster; S203: Fitting each outlier point and all outlier points in its neighborhood, determining the noise interference degree of each outlier point in each cluster by analyzing the error in the fitting process, and determining the filter window adjustment coefficient of each outlier point in each cluster in combination with the deformation eigenvalue, so as to construct the filter window length of each outlier point in each cluster; S3: Filter all abnormal points in all clusters based on the filter window length to obtain a denoised three-dimensional point cloud of the photovoltaic component to be tested, obtain a three-dimensional point cloud of the standard photovoltaic component, and perform deformation detection on the photovoltaic component to be tested by comparing the difference between the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic component.

2. The photovoltaic module deformation detection method for intelligent inspection of a mountain photovoltaic system according to claim 1, characterized in that: The screening process of the feature points is as follows: The reflectivity values ​​of all data points in the three-dimensional point cloud are used as the input of the threshold segmentation algorithm, the segmentation threshold is output, and all data points with reflectivity values ​​less than the segmentation threshold are used as feature points.

3. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The abnormal points in each cluster are the results obtained by using an abnormality detection algorithm on all the feature points in each cluster.

4. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The metric distance in the clustering algorithm used in the process of clustering all feature points is the Euclidean distance between the three-dimensional coordinates of the feature points.

5. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The method for determining the deformation feature value of each abnormal point in each cluster is as follows: In each cluster, the mean of the normal vector similarity between each outlier point and all outlier points in its neighborhood, as well as the mean of the descriptor similarity, are calculated and recorded as the first mean and the second mean. The inverse of the sum of the first mean and the second mean is taken and recorded as the deformation feature value of each outlier point in each cluster.

6. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The noise interference degree of each outlier in each cluster is the residual value of each outlier in each cluster and all outliers in its neighborhood during the surface fitting process.

7. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The filter window adjustment coefficient of each abnormal point in each cluster is the ratio of the noise interference degree of each abnormal point in each cluster to the deformation characteristic value.

8. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The expression of the filter window length of each outlier in each cluster is: ; In the formula, Represents the filter window length of the jth outlier in cluster i; Represents the normalized value of the filter window adjustment coefficient of the jth outlier in cluster i; , represent the preset first value and the preset second value respectively; f( ) represents a custom function, which is used to output an even number closest to the function input value.

9. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The method for obtaining the denoised three-dimensional point cloud of the photovoltaic module to be detected is: All abnormal points in all clusters are used as the input of the filtering algorithm, where the filtering window length of each abnormal point is used as the size of the filtering window in the filtering algorithm. All abnormal points after denoising are output and recorded as correction points. The three-dimensional point cloud composed of all correction points and all data points that have not been filtered is used as the denoised three-dimensional point cloud of the photovoltaic module to be detected.

10. The photovoltaic module deformation detection method for intelligent inspection of mountain photovoltaic system according to claim 1, characterized in that: The deformation detection of the photovoltaic module to be tested comprises: The irregular triangular mesh method is used to construct the three-dimensional models of the denoised three-dimensional point cloud and the three-dimensional point cloud of the standard photovoltaic module, respectively, and they are recorded as the denoised three-dimensional model and the standard three-dimensional model. The interpolation calculation method is used to calculate the interpolation of each pair of grids with the same name between the denoised three-dimensional model and the standard three-dimensional model, and the mean of the interpolation of all pairs of grids with the same name is taken as the deformation degree of the photovoltaic module to be tested.

Images