A method for accurately measuring the size of a photovoltaic module steel frame

Abstract

The present invention relates to the field of linear dimension measurement technology, and in particular to a method for accurately measuring the dimensions of a photovoltaic module steel frame. The method includes measuring the photovoltaic module steel frame and obtaining its original point cloud information, the point cloud information including normal vector, depth information, and three-dimensional coordinates; obtaining the frame attention feature value of each point cloud according to the normal vector change characteristics of the local point cloud; performing density clustering on the frame attention feature value, and obtaining the frame abnormality degree value of each point cloud according to the clustering result; correcting the frame abnormality degree value according to the depth information of each point cloud to obtain the final frame abnormality degree value of each point cloud; determining the abnormal point cloud information according to the final frame abnormality degree value of each point cloud, and correcting the abnormal point cloud information; and measuring the dimensions of the photovoltaic module steel frame based on the corrected point cloud information. The present invention reduces the shadow influence in the process of measuring the size of an object using three-dimensional point cloud technology, and improves the accuracy of the dimension measurement of the photovoltaic module steel frame.

Description

Technical Field

[0001] The invention relates to the technical field of linear dimension measurement, and in particular to a method for accurately measuring the dimensions of a photovoltaic component steel frame. Background Art

[0002] In the production and application of photovoltaic modules, the steel frame is one of the important components of photovoltaic modules. Its main functions include supporting, protecting and fixing photovoltaic panels. At the same time, the steel frame also affects the appearance, structural stability and performance of photovoltaic modules. Therefore, achieving accurate measurement of the size of the steel frame of photovoltaic modules can improve the quality of photovoltaic modules and enhance the competitiveness and acceptability of products.

[0003] Existing methods often use three-dimensional point cloud technology to measure the size of the steel frame of photovoltaic modules. During the measurement process, since some corners of the steel frame of the photovoltaic module have the characteristic of large curvature changes, this feature will lead to insufficient lighting or occlusion at the corners of the frame. Insufficient lighting or occlusion will lead to shadow coverage areas, resulting in a shadow effect in the three-dimensional point cloud measurement process. The shadow effect is that the shadow coverage area of ​​the laser point cloud hits the measured object, which will cause distortion of the point cloud data. The distorted point cloud data will present erroneous object measurement information, thereby greatly reducing the accuracy of measuring the object size using three-dimensional point cloud technology. Summary of the invention

[0004] In order to solve the technical problem of low accuracy in measuring the size of the steel frame of a photovoltaic module using the three-dimensional point cloud technology in the prior art, the purpose of the present invention is to provide a method for accurately measuring the size of the steel frame of a photovoltaic module. The technical solution adopted is as follows:

[0005] A method for accurately measuring the size of a photovoltaic module steel frame, the method comprising:

[0006] Step S1: measuring the steel frame of the photovoltaic module and obtaining its original point cloud information, wherein the point cloud information includes a normal vector, depth information, and three-dimensional coordinates;

[0007] Step S2: obtaining the border focus feature value of each point cloud according to the normal vector change feature of the local point cloud;

[0008] Step S3: performing density clustering on the border focus feature values, and obtaining the border abnormality value of each point cloud according to the clustering result;

[0009] Step S4: correcting the border anomaly value according to the depth information feature of each point cloud to obtain the final border anomaly value of each point cloud;

[0010] Step S5: determining abnormal point cloud information based on the final frame abnormality degree value of each point cloud, and correcting the abnormal point cloud information;

[0011] Step S6: Based on the corrected point cloud information, the dimensions of the steel frame of the photovoltaic module are measured.

[0012] Furthermore, the step S1 of measuring the steel frame of the photovoltaic module and obtaining its original point cloud information is specifically as follows:

[0013] Step S11: placing the laser scanner at an appropriate location, and adjusting its angle and direction as needed to ensure that it can cover the entire steel frame of the photovoltaic module;

[0014] Step S12: Start the laser scanner to emit a laser beam, and obtain the original point cloud data of the surface of the steel frame of the photovoltaic module by receiving the reflected light.

[0015] Furthermore, the step S2 obtains the border attention feature value of each point cloud according to the normal vector change feature of the local point cloud, specifically:

[0016] Step S21: define a local sphere range: for the i-th point cloud data, take the point cloud data as the sphere center and extend N data distances outward to form a sphere range of the point cloud data with N+1 as the radius; wherein N is a positive integer;

[0017] Step S22: define the local linear range and characteristic angle: for the i-th point cloud data, define the local linear range of the point cloud data as T ray segments along the coordinate axes of the three-dimensional coordinate system within the local sphere range with the point cloud data as the starting point; define the characteristic angle as the angle between the unit normal vector of the point cloud data and the unit normal vector of the nearest point cloud data;

[0018] Step S23: Calculate the predicted feature angle value: for the i-th point cloud data, the predicted feature angle value at the position of the i-th point cloud data is obtained by weighted fusion of the feature angles of each point cloud data in its T local linear ranges;

[0019] Step S24: Calculate the border attention feature value according to the predicted feature angle value: take the difference between the predicted feature angle value at the i-th point cloud data position and the actual feature angle of the i-th point cloud data as the border attention feature value of the i-th point cloud data.

[0020] Furthermore, the step S3 performs density clustering on the border attention feature values, and obtains the border abnormality degree value of each point cloud according to the clustering result, which is specifically:

[0021] Step S31: Arrange the border focus feature values ​​of each point cloud data from small to large into one-dimensional coordinates;

[0022] Step S32: Perform DBSCAN density clustering on the border focus feature values ​​of each point cloud data, select the cluster with the largest amount of data in the cluster as the normal cluster among multiple density clusters, and calculate the border abnormality degree value corresponding to each point cloud data outside the normal cluster.

[0023] Furthermore, the specific construction logic of the border anomaly degree value is: if the difference between the border attention feature value of the i-th point cloud data and the border attention feature value corresponding to each point cloud data in the normal cluster is greater, it means that the possibility that the i-th point cloud data is abnormal is greater, that is, the border anomaly degree value of the i-th point cloud data is greater.

[0024] Furthermore, the specific construction logic of the final border abnormality degree value is: if the depth information of the i-th point cloud data and the mean horizontal deviation of the depth information of other point cloud data within the local sphere of the i-th point cloud data are greater, it means that the abnormal probability of the i-th point cloud data is higher, then it is used as a confidence parameter and multiplied with the border abnormality degree value of the i-th point cloud data, and the border abnormality degree value of the i-th point cloud data is corrected to be increased to obtain the final border abnormality degree value of the i-th point cloud data.

[0025] Furthermore, in step S5, the abnormal point cloud information is determined by the final frame abnormality degree value of each point cloud, and the abnormal point cloud information is corrected as follows:

[0026] Step S51: offsetting the abnormal point cloud information of the abnormal point cloud to the geometric center of each point cloud data within the local sphere;

[0027] Step S52: Correct the abnormal point cloud coordinate information.

[0028] Furthermore, the step S51 offsets the abnormal point cloud information of the abnormal point cloud to the geometric center of each point cloud data within its local sphere by calculating the distance of the ith abnormal point cloud offset to the geometric center of each point cloud data within its local sphere; the specific construction logic of the offset distance is: for the ith abnormal point cloud data, if the final border abnormality degree value of the remaining point cloud data within the local sphere with the ith abnormal point cloud data as the center is smaller, the ith abnormal point cloud data is offset more to the geometric center position of all point cloud data within its local sphere.

[0029] Furthermore, the step S52 of correcting the abnormal point cloud coordinate information specifically includes: replacing the abnormal point cloud coordinate information with the coordinate information of the geometric center.

[0030] Furthermore, the step S6 is to measure the size of the steel frame of the photovoltaic module based on the corrected point cloud information as follows:

[0031] Step S61: fitting a three-dimensional image of the steel frame of the photovoltaic module to be tested using the corrected point cloud information;

[0032] Step S62: The measurer extracts the part of the steel frame of the photovoltaic module to be measured, and calculates the accurate size information of the part through the corresponding point cloud information of the part.

[0033] The present invention has the following beneficial effects: by analyzing the normal vector change characteristics of the local point cloud, the border attention feature value of each point cloud is obtained, and then the border attention feature value is density clustered, and the border abnormality degree value of each point cloud is obtained according to the clustering result, and then the border abnormality degree value is corrected by the depth information characteristics of each point cloud to obtain the final border abnormality degree value of each point cloud, and the abnormal point cloud information is determined by the final border abnormality degree value of each point cloud, so as to correct the abnormal point cloud information, and then the corrected point cloud information reflects more accurate size measurement information of the steel frame of the photovoltaic module. The present invention reduces the negative impact of the shadow effect on the process of measuring the size of the object by the three-dimensional point cloud technology, and improves the size measurement accuracy of the steel frame of the photovoltaic module. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A flow chart of a method for accurately measuring the size of a photovoltaic module steel frame provided by an embodiment of the present invention;

[0036] Figure 2 A flow chart of a method for obtaining original point cloud information of a photovoltaic module steel frame provided by an embodiment of the present invention;

[0037] Figure 3 A flow chart of a method for calculating the border attention feature value of each point cloud provided by an embodiment of the present invention;

[0038] Figure 4 A flow chart of a method for calculating the abnormality degree value of the border of each point cloud provided by an embodiment of the present invention;

[0039] Figure 5 A flow chart of a method for correcting abnormal point cloud information provided by an embodiment of the present invention;

[0040] Figure 6A flow chart of a method for measuring the size of a steel frame of a photovoltaic module based on corrected point cloud information provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for accurately measuring the size of a photovoltaic module steel frame proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the method for accurately measuring the size of a photovoltaic module steel frame provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1-Figure 6 , showing a flow chart of a method for accurately measuring the size of a photovoltaic module steel frame provided by an embodiment of the present invention;

[0045] A method for accurately measuring the size of a photovoltaic module steel frame, the method comprising:

[0046] Step S1: measuring the steel frame of the photovoltaic module and obtaining its original point cloud information, wherein the point cloud information includes a normal vector, depth information, and three-dimensional coordinates;

[0047] Specifically, the main purpose of the present invention is to correct the abnormal three-dimensional point cloud information caused by the shadow effect in the process of measuring the steel frame size of the photovoltaic module using the three-dimensional point cloud technology, so as to obtain more accurate measurement information of the steel frame size of the photovoltaic module. Therefore, the present invention first needs to obtain the original point cloud information used for measurement.

[0048] Specifically, the method for obtaining the original point cloud information of the steel frame of the photovoltaic module is as follows: Figure 2 As shown, it includes: step S11: placing the laser scanner at an appropriate position, and adjusting its angle and direction as needed to ensure that it can cover the entire photovoltaic module steel frame; step S12: starting the laser scanner to emit a laser beam, and obtaining point cloud data on the surface of the photovoltaic module steel frame by receiving reflected light.

[0049] Step S2: obtaining the border focus feature value of each point cloud according to the normal vector change feature of the local point cloud;

[0050] Specifically, a normal continuous point cloud area should show that the internal points have small normal vector change difference characteristics. Therefore, this step obtains the bounding box attention feature value of the local center point cloud based on the normal vector change characteristics of the local point cloud, thereby obtaining the bounding box attention feature value of each point cloud.

[0051] Specifically, the method for calculating the bounding box attention eigenvalue of each point cloud is as follows: Figure 3 As shown, it includes: step S21: defining a local sphere range: for the i-th point cloud data, taking the point cloud data as the sphere center, extending N data distances outward respectively, to form a sphere range of the point cloud data with N+1 as the radius; wherein N is a positive integer;

[0052] Specifically, N takes the empirical value of 9.

[0053] Step S22: define the local linear range and characteristic angle: for the i-th point cloud data, define the local linear range of the point cloud data as T ray segments along the coordinate axes of the three-dimensional coordinate system within the local sphere range with the point cloud data as the starting point; define the characteristic angle as the angle between the unit normal vector of the point cloud data and the unit normal vector of the nearest point cloud data;

[0054] It should be noted that if there are multiple nearest point cloud data with the same distance, you can select any one; the local linear range of the present invention is specifically: a total of six ray segments along the positive and negative directions of the x-axis, y-axis, and z-axis within its local spherical range are the local linear range of the point cloud data.

[0055] Step S23: Calculate the predicted feature angle value: for the i-th point cloud data, the predicted feature angle value at the position of the i-th point cloud data is obtained by weighted fusion of the feature angles of each point cloud data in its T local linear ranges;

[0056] It should be noted that in the process of obtaining the predicted feature angle at the i-th point cloud data position by weighted fusion of the feature angles of the point cloud data in different local linear ranges, the greater the deviation of the feature angle of a certain point cloud data from the mean level of the feature angles of all point cloud data in its local linear range, the greater the weight factor. Therefore, according to the above logic, the predicted feature angle weight factor is constructed as follows:

[0057]

[0058] In the formula, Represents the predicted feature angle weight factor of the xth point cloud data within the jth local linear range of the i-th point cloud data; Represents the characteristic angle of the xth point cloud data within the jth local linear range of the i-th point cloud data; It represents the mean of the characteristic angles of all point cloud data within the jth local linear range of the i-th point cloud data; It represents the maximum characteristic angle corresponding to the i-th point cloud data among all point cloud data in the j-th local linear range; It represents the minimum characteristic angle corresponding to the i-th point cloud data among all point cloud data in the j-th local linear range;

[0059] Furthermore, in the process of obtaining the predicted feature angle at the position of the i-th point cloud data by weighted fusion of the feature angles of the point cloud data in different local linear ranges, the farther the distance of a certain point cloud data is, the lower the trustworthiness of the calculated predicted feature angle value is; at the same time, the greater the deviation of the feature angle of a certain point cloud data from the mean level of the feature angles of all point cloud data in its local linear range, that is, the greater the predicted feature angle weight factor is, the lower the trustworthiness of the calculated predicted feature angle value is. Therefore, the predicted feature angle is constructed according to the above logic as follows:

[0060]

[0061] In the formula, Represents the predicted feature angle value of the i-th point cloud data; represents the number of all local linear ranges, ; Represents the total number of all point cloud data collected within the jth local linear range of the i-th point cloud data; Represents the characteristic angle of the xth point cloud data within the jth local linear range of the i-th point cloud data; Represents the characteristic angle weight factor of the xth point cloud data within the jth local linear range of the i-th point cloud data; Represents the distance value from the xth point cloud data to the i-th point cloud data within the j-th local linear range.

[0062] Step S24: Calculate the border attention feature value according to the predicted feature angle value: take the difference between the predicted feature angle value at the i-th point cloud data position and the actual feature angle of the i-th point cloud data as the border attention feature value of the i-th point cloud data.

[0063] Specifically, the calculation formula of the border attention feature value is:

[0064]

[0065] In the formula, Represents the bounding box attention feature value of the i-th point cloud data; Represents the predicted feature angle value of the i-th point cloud data; Represents the actual characteristic angle value of the i-th point cloud data.

[0066] It should be noted that the larger the border attention eigenvalue is, the less the normal vector information of the i-th point cloud data satisfies the characteristic of stable and uniform change of the normal vector information within the local sphere, and the greater the possibility of abnormality of the i-th point cloud data is.

[0067] Step S3: performing density clustering on the border focus feature values, and obtaining the border abnormality value of each point cloud according to the clustering result;

[0068] Specifically, the method for calculating the abnormality value of the border of each point cloud is as follows: Figure 4 As shown, it includes: step S31: arranging the border attention feature values ​​of each point cloud data from small to large into a one-dimensional coordinate; step S32: performing DBSCAN density clustering on the border attention feature values ​​of each point cloud data, selecting the cluster with the largest amount of data in multiple density clusters as the normal cluster, and calculating the border abnormality degree value corresponding to each point cloud data outside the normal cluster.

[0069] It should be noted that the minimum number of sample points for density clustering is 8, and the density radius is 4. Since normal point cloud data accounts for the majority, and abnormal point cloud data accounts for only a minority, the cluster with the largest amount of data in multiple density clusters is taken as the normal cluster. If the border attention feature value of the ith point cloud data (if there is no special explanation, the "ith point cloud data" hereafter refers to the ith point cloud data in the remaining point cloud data except the point cloud data in the normal cluster) is different from the border attention feature value corresponding to each point cloud data in the normal cluster, the more abnormal the border attention feature value of the ith point cloud data is, the greater the possibility that the ith point cloud data is abnormal. Therefore, according to the above logic, the border abnormality degree value is constructed as follows:

[0070]

[0071] In the formula, Indicates the abnormality value of the border of the i-th point cloud data; Represents the total number of point cloud data in the normal cluster; Represents the bounding box attention feature value of the i-th point cloud data; Represents the bounding box attention feature value of the y-th point cloud data in the normal cluster; Represents the normalization function.

[0072] Step S4: correcting the border anomaly value according to the depth information feature of each point cloud to obtain the final border anomaly value of each point cloud;

[0073] It should be noted that, since the border anomaly degree value of each point cloud obtained by only analyzing the normal vector change characteristics of each point cloud can only reflect the shape characteristics of the point cloud, but cannot analyze the complete point cloud coordinate information, this step further combines the depth feature information of the point cloud to correct the border anomaly degree value of each point cloud. Specifically, for the i-th point cloud data, if the depth information of the i-th point cloud data (the depth information of the point cloud data, i.e., the distance between the emission point of the laser scanner and the collection point, can be directly obtained from the point cloud data), if the depth information of the k-th point cloud data and the depth information of other point cloud data within the local sphere of the k-th point cloud data have a greater mean horizontal deviation, it means that the probability of abnormality of the k-th point cloud data is higher, and it is used as a confidence parameter and multiplied with the border anomaly degree value of the k-th point cloud data to increase the border anomaly degree value of the k-th point cloud data; otherwise, it is reduced. Therefore, according to the above logic, the final border anomaly degree value is constructed as follows:

[0074] In the formula, Represents the final border abnormality value of the i-th point cloud; Indicates the abnormality value of the border of the i-th point cloud data; Represents the depth information of the i-th point cloud data, that is, the distance between the laser scanner emission point and the collection point; Represents the total number of all point cloud data collected by the i-th point cloud data within the j-th local linear range; Represents the depth information of the xth point cloud data within the jth local linear range of the i-th point cloud data; Represents the difference between the depth information of the i-th point cloud data and the mean level of the depth information of the rest of the point cloud data within the local sphere of the i-th point cloud data.

[0075] Step S5: determining abnormal point cloud information based on the final frame abnormality degree value of each point cloud, and correcting the abnormal point cloud information;

[0076] It should be noted that the abnormal point cloud data caused by the shadow effect is obtained in the aforementioned steps, and then the abnormal point cloud information of the abnormal point cloud data is adjusted to the mean level of the point cloud information within the local sphere.

[0077] Specifically, the method for correcting abnormal point cloud information is as follows: Figure 5 As shown, it includes: step S51: offsetting the abnormal point cloud information of the abnormal point cloud to the geometric center of each point cloud data within its local sphere;

[0078] Specifically, for the ith abnormal point cloud data, if the final border abnormality value of the remaining point cloud data within the local sphere with the ith abnormal point cloud data as the center is smaller, the ith abnormal point cloud data will be offset more to the geometric center position of all point cloud data within its local sphere (it should be noted here that the ith abnormal point cloud data itself is not necessarily the geometric center of the point cloud data within the local sphere with the ith abnormal point cloud data as the center, because point cloud data may not be acquired at every position); otherwise, the adjustment is less, thereby realizing the correction of the abnormal point cloud coordinate information in the process of constructing the three-dimensional image of the object to be measured by the laser point cloud. Therefore, according to the above logic, the distance of the abnormal point cloud data offset to the geometric center of each point cloud data within its local sphere is as follows:

[0079]

[0080] In the formula, Indicates the distance that the i-th abnormal point cloud data is offset from the geometric center of each point cloud data within its local sphere; Represents the distance value from the i-th abnormal point cloud data to the geometric center of each point cloud data within its local sphere; Represents the total number of all point cloud data within the local sphere with the i-th abnormal point cloud data as the sphere center; Represents the final border anomaly feature value of the i-th point cloud data within the local sphere with the i-th abnormal point cloud data as the sphere center; It represents the mean of the final border anomaly feature values ​​of all point cloud data within the local sphere with the i-th abnormal point cloud data as the center.

[0081] Step S52: Correct the abnormal point cloud coordinate information.

[0082] Specifically, the coordinate information of the abnormal point cloud is replaced by the coordinate information of the geometric center, so as to realize the correction of the coordinate information of the abnormal point cloud in the process of constructing the three-dimensional image of the object to be measured by the laser point cloud.

[0083] Step S6: Based on the corrected point cloud information, the dimensions of the steel frame of the photovoltaic module are measured.

[0084] Specifically, the method for measuring the size of the steel frame of the photovoltaic module based on the corrected point cloud information is as follows: Figure 6 As shown, it includes: step S61: using the corrected point cloud information to fit the three-dimensional image of the steel frame of the photovoltaic component to be tested; step S62: the measurer extracts the part of the steel frame of the photovoltaic component to be tested, and calculates the accurate size information of the part through the corresponding point cloud information of the part.

[0085] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for accurately measuring the size of a photovoltaic module steel frame, characterized in that: The method comprises: Step S1: measuring the steel frame of the photovoltaic module and obtaining its original point cloud information, wherein the point cloud information includes a normal vector, depth information, and three-dimensional coordinates; Step S2: obtaining the border focus feature value of each point cloud according to the normal vector change feature of the local point cloud; Step S3: performing density clustering on the border focus feature values, and obtaining the border abnormality value of each point cloud according to the clustering result; Step S4: correcting the border anomaly value according to the depth information feature of each point cloud to obtain the final border anomaly value of each point cloud; Step S5: determining abnormal point cloud information based on the final frame abnormality degree value of each point cloud, and correcting the abnormal point cloud information; Step S6: Based on the corrected point cloud information, measure the size of the steel frame of the photovoltaic module; The step S3 performs density clustering on the border attention feature values, and obtains the border abnormality degree value of each point cloud according to the clustering result, which is specifically: Step S31: Arrange the border focus feature values ​​of each point cloud data from small to large into one-dimensional coordinates; Step S32: performing DBSCAN density clustering on the border attention feature values ​​of each point cloud data, selecting the cluster with the largest amount of data in the cluster among multiple density clusters as the normal cluster, and calculating the border abnormality degree value corresponding to each point cloud data outside the normal cluster; The specific construction logic of the border anomaly degree value is: if the difference between the border attention feature value of the i-th point cloud data and the border attention feature value corresponding to each point cloud data in the normal cluster is greater, it means that the possibility of the i-th point cloud data being abnormal is greater, that is, the border anomaly degree value of the i-th point cloud data is greater; The specific construction logic of the final border anomaly degree value is: if the depth information of the i-th point cloud data and the depth information of other point cloud data within the local sphere of the i-th point cloud data have a larger mean horizontal deviation, it means that the probability of anomaly of the i-th point cloud data is higher, and it is used as a confidence parameter and multiplied with the border anomaly degree value of the i-th point cloud data, and the border anomaly degree value of the i-th point cloud data is corrected to be increased, so as to obtain the final border anomaly degree value of the i-th point cloud data; The step S2 obtains the border focus feature value of each point cloud according to the normal vector change feature of the local point cloud, specifically: Step S21: define a local sphere range: for the i-th point cloud data, take the point cloud data as the sphere center and extend N data distances outward to form a sphere range of the point cloud data with N+1 as the radius; wherein N is a positive integer; Step S22: define the local linear range and characteristic angle: for the i-th point cloud data, define the local linear range of the point cloud data as T ray segments along the coordinate axes of the three-dimensional coordinate system within the local sphere range with the point cloud data as the starting point; define the characteristic angle as the angle between the unit normal vector of the point cloud data and the unit normal vector of the nearest point cloud data; Step S23: Calculate the predicted feature angle value: for the i-th point cloud data, the predicted feature angle value at the position of the i-th point cloud data is obtained by weighted fusion of the feature angles of each point cloud data in its T local linear ranges; Step S24: Calculate the border attention feature value according to the predicted feature angle value: take the difference between the predicted feature angle value at the i-th point cloud data position and the actual feature angle of the i-th point cloud data as the border attention feature value of the i-th point cloud data.

2. The method for accurately measuring the size of a photovoltaic module steel frame according to claim 1, characterized in that: The step S1 of measuring the photovoltaic module steel frame and obtaining its original point cloud information is specifically as follows: Step S11: placing the laser scanner at an appropriate location, and adjusting its angle and direction as needed to ensure that it can cover the entire steel frame of the photovoltaic module; Step S12: Start the laser scanner to emit a laser beam, and obtain the original point cloud data of the surface of the steel frame of the photovoltaic module by receiving the reflected light.

3. The method for accurately measuring the size of a photovoltaic module steel frame according to claim 1, characterized in that: The step S5 determines the abnormal point cloud information based on the final frame abnormality value of each point cloud, and corrects the abnormal point cloud information specifically as follows: Step S51: offsetting the abnormal point cloud information of the abnormal point cloud to the geometric center of each point cloud data within the local sphere; Step S52: Correct the abnormal point cloud coordinate information.

4. The method for accurately measuring the size of a photovoltaic module steel frame according to claim 3, characterized in that: The step S51 offsets the abnormal point cloud information of the abnormal point cloud to the geometric center of each point cloud data within its local sphere, specifically by calculating the offset distance of the i-th abnormal point cloud to the geometric center of each point cloud data within its local sphere; the specific construction logic of the offset distance is: for the i-th abnormal point cloud data, if the final border abnormality degree value of the remaining point cloud data within the local sphere with the i-th abnormal point cloud data as the center is smaller, the i-th abnormal point cloud data is offset more to the geometric center position of all point cloud data within its local sphere.

5. The method for accurately measuring the size of a photovoltaic module steel frame according to claim 3, characterized in that: The step S52 of correcting the abnormal point cloud coordinate information specifically includes: replacing the abnormal point cloud coordinate information with the coordinate information of the geometric center.

6. The method for accurately measuring the size of a photovoltaic module steel frame according to claim 1, characterized in that: The step S6 of measuring the size of the steel frame of the photovoltaic module based on the corrected point cloud information is specifically as follows: Step S61: fitting a three-dimensional image of the steel frame of the photovoltaic module to be tested using the corrected point cloud information; Step S62: The measurer extracts the part of the steel frame of the photovoltaic module to be measured, and calculates the accurate size information of the part through the corresponding point cloud information of the part.

Images