Three-dimensional modeling method and system based on low-altitude point cloud

Abstract

The present invention relates to the technical field of abnormal data screening, and specifically to a three-dimensional modeling method and system based on low-altitude point clouds. First, initial three-dimensional point cloud data of the photovoltaic shooting area is acquired through non-contact, full-coverage scanning by a low-altitude drone. The high similarity between data points in the local area is compared to screen suspected fault points (including real fault points and interference points). For each suspected fault point, the height difference characteristics of the data points within its local range are analyzed, and the "abnormal isolation" is quantified to obtain an isolation degree value. The similarity and fluctuation stability of local height changes are captured to determine the fault degree value. The two are integrated to obtain fault parameters, and high-confidence final fault points are screened (interference points are eliminated). Finally, based on the location, quantity distribution, and fault parameters of the final fault point, the point cloud data is adaptively updated to optimize the three-dimensional model, preserving the overall structural information while accurately depicting the fault details.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal data screening, and in particular to a three-dimensional modeling method and system based on low-altitude point clouds. Background Art

[0002] As a core clean energy source, photovoltaic power generation continues to see rapid expansion. The operational and maintenance efficiency of photovoltaic power plants directly impacts power generation profitability. As key structures supporting photovoltaic modules, the condition monitoring of photovoltaic racks is crucial. In recent years, the use of low-altitude drones equipped with LiDAR or photogrammetry to acquire point cloud data has provided a new technical path for 3D modeling and fault detection of photovoltaic racks. This method reconstructs the spatial structure of photovoltaic arrays using high-precision 3D point cloud data, offering the advantages of non-contact, full coverage, and high efficiency.

[0003] During the 3D modeling process, photovoltaic bracket areas with defects or obvious faults need to be paid special attention to in order to build a more accurate 3D model for subsequent maintenance. Existing technologies usually screen these data points based on the numerical mutation characteristics of the height values ​​between data points. However, given that drones usually do not move completely along the established route during flight, but will have certain maneuvers and adjustments, some non-fault interference points will be generated, and the height values ​​of these interference data points will also mutate. Therefore, if the fault data points are screened solely based on the mutation of the height values ​​of the data points, the final 3D modeling effect will have low reliability and reduced accuracy. Summary of the Invention

[0004] In order to solve the technical problem that, since drones usually do not move completely along a predetermined route during flight, but rather undergo certain maneuvers and adjustments, some non-fault interference points will be generated, and the height values ​​of these interference data points will also undergo sudden changes, if the fault data points are screened based solely on the sudden changes in the height values ​​of the data points, the final 3D modeling effect will be less reliable and less accurate, the purpose of the present invention is to provide a 3D modeling method and system based on low-altitude point clouds. The technical solutions adopted are as follows:

[0005] A three-dimensional modeling method based on low-altitude point clouds, comprising:

[0006] Obtain initial three-dimensional point cloud data of the photovoltaic shooting area;

[0007] In the photovoltaic shooting area, analyzing the similarity between the height values ​​corresponding to the data points in different area blocks, thereby screening suspected fault data points;

[0008] Within the local range corresponding to each suspected fault data point, the difference characteristics between the height values ​​of the data points are analyzed to determine the isolation degree value of each suspected fault data point; within the local range corresponding to each suspected fault data point, the similarity characteristics and difference fluctuation characteristics of the changes between the height values ​​of the data points are analyzed to determine the fault degree value of each suspected fault data point;

[0009] The isolation degree value and fault degree value of each suspected fault data point are combined to obtain the fault parameter of each suspected fault data point, which is used to screen the final fault data point. The initial three-dimensional point cloud data is adaptively updated according to the position, quantity distribution and fault parameters of all final fault data points to obtain the updated three-dimensional point cloud data of the photovoltaic shooting area.

[0010] Furthermore, the method for obtaining the suspected fault data point includes:

[0011] In the photovoltaic shooting area, all data points are divided into regions based on two-dimensional plane coordinates to obtain multiple region blocks;

[0012] In each block, the mean of the height values ​​of all data points is used as the height representation value of each block. For any two blocks, the absolute value of the difference between the height representation values ​​of the two blocks is negatively correlated and normalized to obtain the height representation similarity between the two blocks.

[0013] Select any area block as the block to be tested, and use the area blocks whose high performance similarity value with the block to be tested is greater than or equal to a preset similarity threshold as similar blocks to the block to be tested;

[0014] Multiply the sum of the high-performance similarity values ​​between the block to be tested and all similar blocks by the number of similar blocks, and perform negative correlation mapping on the obtained product and normalize it to obtain the value as the suspected fault index of the block to be tested;

[0015] All the regional blocks are sorted in descending order according to the suspected fault index, and the data points in the first preset number of regional blocks are taken as suspected fault data points.

[0016] Furthermore, the method for obtaining the isolation degree value includes:

[0017] In a two-dimensional plane, with each suspected fault data point as the center, several data points are selected in each preset direction as the neighborhood data point sequence of each suspected fault data point;

[0018] In the sequence of neighboring data points in each preset direction of each suspected fault data point, the absolute value of the difference in height between each neighboring data point and the suspected fault data point is used as a height difference parameter, and the sum of all height difference parameters is used as a height significance indicator of each suspected fault data point in each preset direction;

[0019] Among the highly significant indicators corresponding to all preset directions of each suspected fault data point, the normalized value of the minimum highly significant indicator is used as the isolation degree value of each suspected fault data point.

[0020] Furthermore, the method for obtaining the fault degree value includes:

[0021] In a two-dimensional plane, with each suspected fault data point as the center, in each preset direction, the height value change difference characteristics between the data points are compared to determine the height change data sequence of each suspected fault data point in each preset direction;

[0022] In each height change data sequence of each suspected fault data point, the height difference between each two adjacent data points is calculated as the height difference factor, the variance of all height difference factors is used as the height change fluctuation factor of each height change data sequence, and the sum of the height change fluctuation factors corresponding to all height change data sequences of each suspected fault data point is negatively correlated and normalized to obtain the value as the local height change consistency of each suspected fault data point;

[0023] Analyze the similarity characteristics of the height change data sequences between the preset directions of each suspected fault data point, and determine the local height change similarity value of each suspected fault data point;

[0024] The normalized value of the sum of the local height change consistency and the local height change similarity value of each suspected fault data point is used as the fault degree value of each suspected fault data point.

[0025] Furthermore, the method for obtaining the height change data sequence includes:

[0026] Taking each suspected fault data point as the starting point, an initial sequence is generated in each preset direction, and the data points are traversed in each preset direction. If the normalized value of the absolute difference between the height value of the traversed data point and the height value of the last data point in the initial sequence is less than the preset difference threshold, the traversed data point is incorporated into the initial sequence, otherwise the traversal is stopped.

[0027] After the traversal is completed, the initial sequence of each suspected fault data point in each preset direction is used as the height change data sequence.

[0028] Furthermore, the method for obtaining the local height change similarity value includes:

[0029] For any two preset directions, the height values ​​in the height change data sequence of each suspected fault data point in these two preset directions are processed using a linear interpolation algorithm to obtain two height value sequences of equal length. The normalized Pearson correlation coefficient of these two height value sequences is used as the height change similarity factor of each suspected fault data point in these two preset directions.

[0030] The average of the height change similarity factors between all preset directions of each suspected fault data point is used as the local height change similarity value of each suspected fault data point.

[0031] Furthermore, the method for obtaining the final fault data point includes:

[0032] The value after negative correlation mapping of the isolation degree value of each suspected fault data point is multiplied by the fault degree value of each suspected fault data point, and the obtained product is normalized as the fault parameter of each suspected fault data point;

[0033] Among all suspected fault data points, the suspected fault data points whose fault parameters are greater than the preset fault threshold are taken as the final fault data points.

[0034] Furthermore, the method for acquiring updated three-dimensional point cloud data includes:

[0035] All final fault data points are clustered based on the DBSCAN clustering algorithm and the three-dimensional coordinates of the final fault point to obtain all cluster clusters;

[0036] The area corresponding to each cluster in the initial three-dimensional point cloud data is used as the fault point cloud area;

[0037] Analyze the position between the fault point cloud areas and the number of final fault data points, and combine them with the numerical characteristics of the fault parameters of the final fault data points to obtain the fault severity index of each fault point cloud area;

[0038] The fault severity index of each fault point cloud area is used as the sampling density to input into the Poisson disk sampling algorithm to uniformly generate updated point cloud data for each fault point cloud area;

[0039] The updated point cloud data of each fault point cloud area is merged with the initial three-dimensional point cloud data of the non-fault point cloud area to obtain updated three-dimensional point cloud data of the photovoltaic shooting area.

[0040] Furthermore, the method for obtaining the fault severity indicator includes:

[0041] Obtain the centroid of each fault point cloud area, calculate the normalized Euclidean distance between the two-dimensional plane coordinates of the centroids of any two fault point cloud areas, and use this distance factor as the distance factor between the two fault point cloud areas.

[0042] Select any fault point cloud area as the test area, and perform negative correlation mapping on the sum of the distance factors between the test area and the remaining fault point cloud areas, and normalize the sum to obtain the fault severity factor of the test area.

[0043] The normalized value of the product of the total number of all final fault data points in the test area, the mean value of the fault parameters of all final fault data points, and the fault severity factor of the test area is used as the fault severity index of the test area.

[0044] A three-dimensional modeling system based on low-altitude point clouds includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, the steps of a three-dimensional modeling method based on low-altitude point clouds are implemented.

[0045] The present invention has the following beneficial effects:

[0046] A low-altitude drone is used to perform a non-contact, full-coverage scan of the photovoltaic imaging area, acquiring initial three-dimensional point cloud data. Within the photovoltaic imaging area, the similarities in height values ​​between data points at different locations are then compared to screen suspected faulty data points and preliminarily identify areas of height anomalies. These suspected faulty data points include both true faulty points and non-faulty interference points caused by changes in the drone's flight attitude or path deviation. Furthermore, because interference data points typically exhibit abrupt height differences on the photovoltaic panel, their height variations lack a clear regularity compared to surrounding data points and are typically isolated. True faulty data points, on the other hand, exhibit relatively consistent height variation trends within a local area due to the region's nature. Therefore, within the local area of ​​each suspected faulty data point, the height difference characteristics between the data point and the suspected faulty point are analyzed to determine an isolation degree value, which is used to quantify the data point's "abnormal isolation." Furthermore, the similarity and fluctuation characteristics of the height value variations of the data points are analyzed to determine a fault degree value. This captures the variation patterns of the height values ​​within the local area and quantifies the stability of the height fluctuation within the local area, helping to distinguish true faults from transient interference. Furthermore, the isolation degree and fault severity values ​​of suspected fault data points are combined to obtain fault parameters. The final fault data points are then filtered based on these parameters, effectively eliminating interference points caused by environmental interference with the drone, providing a high-purity set of fault points for subsequent 3D modeling. Finally, the point cloud data is adaptively updated based on the location, quantity distribution, and fault parameters of the final fault data points to optimize the 3D model. The final model retains the overall structural information of the PV array while accurately depicting the details of the fault area, providing highly reliable 3D data support for maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. 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 any creative work.

[0048] Figure 1 A flowchart of a method for three-dimensional modeling based on low-altitude point clouds provided by one embodiment of the present invention;

[0049] Figure 2 A flow chart of a method for obtaining a fault degree value provided by one embodiment of the present invention;

[0050] Figure 3 A system block diagram of a low-altitude point cloud-based 3D modeling system provided by one embodiment of the present invention;

[0051] Figure 4 A schematic diagram of the system structure of a low-altitude point cloud-based 3D modeling system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0052] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a low-altitude point cloud-based 3D modeling method and system proposed by the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0053] 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.

[0054] The following describes in detail a three-dimensional modeling method and system based on low-altitude point clouds provided by the present invention with reference to the accompanying drawings.

[0055] See also Figure 1 , which shows a method flow chart of a 3D modeling method based on low-altitude point clouds provided by one embodiment of the present invention, the method comprising the following steps:

[0056] Step S1: Acquire initial three-dimensional point cloud data of the photovoltaic shooting area.

[0057] First, a low-altitude multi-rotor drone (such as the DJI Phantom 4 RTK or a fixed-wing vertical take-off and landing drone) is equipped with a high-precision multi-line lidar with a scanning frequency exceeding 10 Hz and an angular resolution of less than 0.2°. This ensures the ability to capture the fine structures of the photovoltaic bracket (such as screws and connectors). A route such as a zigzag or grid-shaped route is planned for the drone, and spatial calibration is performed based on the checkerboard calibration method to ensure that the point cloud is aligned with the image space, thereby obtaining the initial 3D point cloud data of the photovoltaic shooting area, where each data point records the XYZ coordinates. The X and Y coordinates represent the two-dimensional plane coordinate information of the data point, and the Z coordinate represents the height value of each data point.

[0058] Step S2: In the photovoltaic shooting area, the similarities between the height values ​​corresponding to the data points in different area blocks are analyzed to screen out suspected fault data points.

[0059] During the three-dimensional modeling process, it is necessary to focus on the photovoltaic support areas with defects or obvious faults in order to build a more accurate three-dimensional model for subsequent maintenance. Therefore, in this embodiment of the present invention, it is necessary to screen out fault data points.

[0060] Given that a photovoltaic bracket is composed of multiple similar photovoltaic brackets to form an overall structure, forming a frame to support the panel; therefore, if the height similarity between a local area and other areas is smaller, and the number of similar areas is smaller, it means that the height value of the data point in the area has undergone a sudden change, and it is more likely to be a data point in the fault area; therefore, in this step, the similarity between the height values ​​corresponding to the data points of different area blocks can be analyzed in the photovoltaic shooting area, and suspected fault data points can be preliminarily screened out.

[0061] Preferably, in one embodiment of the present invention, the method for obtaining suspected fault data points includes:

[0062] In the photovoltaic imaging area, all data points are divided into regions based on the two-dimensional plane coordinates to obtain multiple regional blocks. That is, on the two-dimensional plane (only the X and Y coordinates of the data points in the three-dimensional point cloud data are considered), the photovoltaic imaging area is divided into several 3×3 regional blocks, and there is no overlap between the regional blocks.

[0063] In each area block, the mean of the height values ​​of all data points is used as the height performance value of each area block. The height performance value reflects the average height level of the data points in each area block. For any two area blocks, the absolute value of the difference between the height performance values ​​of the two area blocks is calculated. The smaller the absolute value of the difference here, the more consistent the average height level of the data points in the two area blocks, and the higher the similarity. Therefore, the absolute value of the difference is negatively correlated and normalized to achieve logical relationship correction, and the height performance similarity value between the two area blocks is obtained. The larger the height performance similarity value, the higher the height similarity of the data points in the two area blocks. The negative correlation mapping and normalization here can be performed using the formula ,in, It represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0064] Based on the above analysis, it can be known that when the height similarity between a certain local area and other areas is smaller, and the number of similar areas is smaller, it means that the height value of the data point in the area has mutated, and it needs to be screened out first for subsequent analysis. Therefore, any area block is selected as the block to be tested, and the area block with a height similarity value greater than or equal to the preset similarity threshold with the block to be tested is used as a similar block of the block to be tested; then the sum of the height similarity values ​​between the block to be tested and all similar blocks is multiplied by the number of similar blocks. The larger the product, the more area blocks have the same height value as the data point in the block to be tested, and the higher the similarity, the smaller the possibility of a sudden change in the height value of the data point in the block to be tested. Therefore, the obtained product is negatively correlated and normalized, and the value is used as the suspected fault index of the block to be tested. At this time, the suspected fault index realizes the logical relationship correction, and the larger the value, the greater the probability that the height value of the data point in the block to be tested has mutated. The negative correlation mapping and normalization processing here can be used using the formula ,in, It represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0065] Finally, all the regional blocks are sorted in descending order according to the suspected fault index, and the data points in the first preset number of regional blocks are used as suspected fault data points.

[0066] It should be noted that, in this embodiment of the present invention, the preset similarity threshold is 0.7, and the preset number is set to one-third of the total number of regional blocks and rounded up. The specific values ​​can be adjusted according to the implementation scenario and are not limited here.

[0067] Step S3: Within the local range corresponding to each suspected fault data point, analyze the difference characteristics between the height values ​​of the data points to determine the isolation degree value of each suspected fault data point; within the local range corresponding to each suspected fault data point, analyze the similarity characteristics and difference fluctuation characteristics of the changes between the height values ​​of the data points to determine the fault degree value of each suspected fault data point.

[0068] Because low-altitude drones usually do not move completely along a predetermined path during flight, but will have certain maneuvers and adjustments; therefore, low-altitude drones may scan the same ground point from slightly different angles or positions, and the scanning data at different angles or positions may produce some additional errors, which will also appear as sudden changes in height values. Therefore, the suspected fault data points obtained in the aforementioned step S2, in addition to the data points with height changes caused by real photovoltaic panel failures or defects, will also include these interference data points caused by the drone itself. If the interference data points and the real fault data points are not distinguished, the final point cloud data will not be able to truly reflect the actual geometric shape of the photovoltaic bracket when performing three-dimensional modeling, thereby increasing unnecessary interference and errors.

[0069] Interference data points are usually caused by factors such as the external environment (such as wind, climate change) or system errors affecting the drone; their variation range is large, irregular, and usually isolated in space; while fault data points are usually caused by structural problems (such as deformation and damage of the bracket), which may manifest as mutations in local areas and have certain similarities and regularities, and this change is reasonable in the actual structure of the photovoltaic bracket.

[0070] Interference data points appear on the photovoltaic panel as: forming an abrupt height difference with other data points, and there is no obvious regularity in the height changes between these interference data points and the surrounding data points; that is, they are significantly different from the normal surface characteristics of the photovoltaic panel. For example, the surface of a normal photovoltaic panel is often relatively smooth or uniform, while interference data points may show abrupt jumps or fluctuations, and the interference data points are usually isolated and have no direct spatial correlation with the surrounding data points. Therefore, in an embodiment of the present invention, the difference characteristics between the height values ​​of the data points can first be analyzed within the local range corresponding to each suspected fault data point to determine the isolation degree value of each suspected fault data point.

[0071] Preferably, in one embodiment of the present invention, the method for obtaining the isolation degree value includes:

[0072] Since the photovoltaic architecture has directionality, such as arrangement in rows and columns, in a two-dimensional plane (that is, only considering the X and Y coordinates of the data points), with each suspected fault data point as the center, several data points are selected in each preset direction as a neighborhood data point sequence for each suspected fault data point. In the embodiment of the present invention, the preset directions are specifically set to eight neighborhood directions, and the number of selected data points is set to 8. The selection of the specific directions and the setting of the number of data points can be set according to the implementation scenario and are not limited here.

[0073] Then, in the sequence of neighborhood data points in each preset direction of each suspected fault data point, the absolute value of the difference in height between each neighborhood data point and the suspected fault data point is used as the height difference parameter. The larger the height difference parameter, the more prominent the height value of the suspected fault data point is compared with the neighborhood data point. The sum of all height difference parameters in each preset direction is used as the height significance index of each suspected fault data point in each preset direction. Similarly, the larger the height significance index, the greater the difference in height between the suspected fault data point and other data points in the preset direction, and the more obvious the isolation of the suspected fault data point.

[0074] Finally, among the height significance indicators corresponding to all preset directions of each suspected fault data point, the minimum height significance indicator is normalized to the value of the isolation degree of each suspected fault data point. The larger the minimum height significance indicator, the more significant the height abruptness of the suspected fault data point, the higher the isolation degree, and the more likely it is an interference data point. Normalization is a technical means well known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0075] The above steps can analyze the isolated state of suspected fault data points. If the photovoltaic panel has faults such as cracks, stains, wear or deformation, the point cloud around the fault area will show a relatively regular height mutation; that is, the data points in the fault area of ​​the photovoltaic panel usually have a consistent height change trend, forming a relatively regular change pattern; this is because the surface of the photovoltaic panel is usually uniform. For example, if there is some kind of pollution or corrosion, the height change in the point cloud may be along a certain direction or form a certain geometric shape; therefore, within the local range corresponding to each suspected fault data point, the change similarity characteristics and difference fluctuation characteristics between the height values ​​of the data points can be analyzed to determine the fault degree value of each suspected fault data point.

[0076] Preferably, a method for obtaining a fault severity value in one embodiment of the present invention includes:

[0077] See also Figure 2 , which shows a flow chart of a method for obtaining a fault degree value in one embodiment of the present invention, the method comprising the following steps:

[0078] Step S301: In a two-dimensional plane, with each suspected fault data point as the center, in each preset direction, compare the change difference characteristics of the height values ​​between the data points to determine the height change data sequence of each suspected fault data point in each preset direction.

[0079] Faulty areas (such as surface depressions, cracks, etc.) usually cause the surrounding height data to present relatively regular height variation characteristics. Therefore, in this embodiment of the present invention, each suspected fault data point can be taken as the center, and then the height variation characteristics of the surrounding data points can be analyzed to first obtain a height variation data sequence to provide a basis for subsequent analysis.

[0080] Therefore, in the two-dimensional plane (only considering the X and Y coordinates of the data points in the three-dimensional point cloud data), each suspected fault data point is used as the starting point, and an initial sequence is corresponding to each preset direction. The data points are traversed in each preset direction. If the normalized value of the difference between the height value of the traversed data point and the height value of the last data point in the initial sequence is less than the preset difference threshold, the traversed data point is incorporated into the initial sequence, otherwise the traversal is stopped.

[0081] After the traversal is completed, the initial sequence of each suspected fault data point in each preset direction is used as the height change data sequence.

[0082] In the embodiment of the present invention, the multiple preset directions are specifically set as eight-neighborhood directions. Here, an example is given to illustrate the acquisition of a height variation data sequence: for a suspected fault data point a, with the suspected fault data point a as the starting point, in any preset direction, initially, the initial sequence in the preset direction only contains the suspected fault data point a. Then, the absolute value of the difference in height between the suspected fault data point a and the first data point b in the preset direction is calculated. If the normalized value of the absolute value of the difference is less than a preset difference threshold, it indicates that the heights of data point b and the suspected fault data point are relatively consistent and may be affected by the same factor. Therefore, data point b is added to the initial sequence. Then, traversal continues along the preset direction, and the absolute value of the difference in height between data point b and data point c (the traversed data point) is calculated. If the normalized value of the absolute value of the difference is greater than or equal to the preset difference threshold, it indicates that the height consistency of data point c is low with the data points in the initial sequence. Therefore, traversal in the preset direction is stopped, thereby obtaining the height variation data sequence in the preset direction (consisting of the suspected fault data point a and data point b).

[0083] It should be noted that the preset difference threshold is 0.7, and the specific value can be adjusted according to the implementation scenario and is not limited here.

[0084] Step S302: In each height change data sequence of each suspected fault data point, analyze the difference fluctuation characteristics of the height values ​​between adjacent data points, and determine the local height change consistency of each suspected fault data point.

[0085] In each height change data sequence of each suspected fault data point, the difference in height values ​​between each two adjacent data points is calculated as the height difference factor. The height difference factor directly reflects the severity of the local height change. The larger the value, the more severe the change. In view of the fact that the variance can reflect the degree of discreteness between a set of data, the variance of all height difference factors in each height change data sequence is used as the height change fluctuation factor of each height change data sequence. The smaller the height change fluctuation factor, the more concentrated the difference distribution between the height values ​​of the data points in this height change data sequence is, and the higher the consistency is. Finally, the sum of the height change fluctuation factors corresponding to all height change data sequences of each suspected fault data point is negatively correlated and normalized as the local height change consistency of each suspected fault data point. The greater the local height change consistency, the more consistent the height value of the suspected fault data point within the local range has a more consistent height change trend, and therefore the possibility of it being in the real fault area will also be higher. The negative correlation mapping and normalization processing here can be performed using the formula ,in, It represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0086] Step S303: analyzing the variation similarity features between the height variation data sequences between the preset directions of each suspected fault data point, and determining the local height variation similarity value of each suspected fault data point.

[0087] In step S302, the consistent pattern of height changes of each suspected fault data point in the local range of each preset direction is analyzed. Since the influence of the fault area will cause similar height change patterns of data points in each direction, in this sub-step, the consistent pattern of height changes between different preset directions can be further analyzed.

[0088] For any two preset directions, first use the linear interpolation algorithm to process the height values ​​of the height change data sequence of each suspected fault data point in the two preset directions to obtain two height value sequences of equal length; if the height change patterns of the data points in the two preset directions are consistent, then the linear correlation will be more obvious, so calculate the Pearson correlation coefficient of the height value sequence in the two preset directions. If it is closer to 1, the degree of consistency of the change is higher, so the value after normalization of the Pearson correlation coefficient is used as the height change similarity factor of each suspected fault data point in the two preset directions. At this time, there is a height change similarity factor between each two preset directions, and the larger the value, the greater the degree of consistency of the height value change of the data points in the two preset directions. Since the Pearson correlation coefficient may be positive or negative, the normalization here adopts function.

[0089] Finally, the mean of the height change similarity factors between all preset directions of each suspected fault data point is taken as the local height change similarity value of each suspected fault data point. The larger the local height change similarity value, the stronger the regularity of the height change of each suspected fault data point in multiple preset directions, and therefore the more likely it is that it belongs to the data point in the real fault area.

[0090] It should be noted that the linear interpolation algorithm is a well-known technology and the specific process will not be described here in detail.

[0091] Step S304: The local height change consistency and the local height change similarity value of each suspected fault data point are integrated to determine the fault degree value of each suspected fault data point.

[0092] Based on the above analysis, it can be seen that the local height change consistency and local height change similarity value of each suspected fault data point are positively correlated with the possibility that the suspected fault data point belongs to the real fault area. Therefore, here, the sum of the local height change consistency and local height change similarity value of each suspected fault data point is normalized and used as the fault degree value of each suspected fault data point. The larger the fault degree value, the more regular the height mutation of the data points around the suspected fault data point. The more likely the suspected fault data point is a data point in the real fault area. Normalization is a technical means well known to those skilled in the art. The normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0093] Step S4: Combine the isolation degree value and the fault degree value of each suspected fault data point to obtain the fault parameter of each suspected fault data point for screening the final fault data point; adaptively update the initial three-dimensional point cloud data according to the position, quantity distribution and fault parameters between all final fault data points to obtain updated three-dimensional point cloud data of the photovoltaic shooting area.

[0094] Based on the analysis and calculation in step S3, the isolation degree value and fault degree value of each suspected fault data point can be obtained. The former is used to measure the possibility that the suspected fault data point is an interference data point, and the latter is used to measure the probability that the suspected fault data point is a real fault data point. Therefore, the two are combined to determine which suspected fault data points are the real final fault data points.

[0095] Preferably, in one embodiment of the present invention, the method for obtaining the final fault data point includes:

[0096] The larger the isolation degree value of a suspected fault data point, the more likely it is to be an interference data point; the larger the fault degree value of a suspected fault data point, the more likely it is to be a real fault data point. Therefore, the isolation degree value of each suspected fault data point is negatively correlated with the isolation degree value to achieve logical relationship correction, and then multiplied by the isolation degree value of each suspected fault data point. The obtained product is normalized and used as the fault parameter of each suspected fault data point. At this time, the larger the fault parameter, the greater the probability that the suspected fault data point is not a real fault data point. The negative correlation mapping and normalization processing here can be performed using the formula ,in, It represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0097] Finally, among all the suspected fault data points, the suspected fault data points whose fault parameters are greater than the preset fault threshold are taken as the final fault data points.

[0098] It should be noted that the preset fault threshold in this embodiment of the present invention is 0.65, and the specific value can be adjusted according to the implementation scenario and is not limited here.

[0099] At this point, the actual fault data points can be effectively distinguished from the interference data points, resulting in all final fault data points. Given that areas with more severe faults (such as deep dents and structural fractures) have a greater impact on the stability and power generation efficiency of the photovoltaic mounts, a more refined data model is required to facilitate more accurate positioning and maintenance. Therefore, in this embodiment of the present invention, the fault severity is quantified based on the position, quantity distribution, and fault parameters of all final fault data points, thereby adaptively updating the initial 3D point cloud data to obtain updated 3D point cloud data for the photovoltaic capture area.

[0100] Preferably, in one embodiment of the present invention, the method for acquiring updated three-dimensional point cloud data includes:

[0101] The DBSCAN clustering algorithm can automatically identify clusters of any shape, such as irregular fault areas. Therefore, all final fault data points are clustered based on the DBSCAN clustering algorithm and the 3D coordinates of the final fault points. All clusters are obtained, and the area corresponding to each cluster in the initial 3D point cloud data is used as the fault point cloud area. The default radius for the clustering process is set to 0.1 meters, and the minimum number of points is set to 4. The specific values ​​can be adjusted according to the implementation scenario and are not limited here.

[0102] Photovoltaic panel failures usually manifest as local hot spots, damaged photovoltaic cells, or poor contact. The larger the fault area, the more severe the damage, which may have affected a larger area of ​​photovoltaic cells or conductive paths, resulting in a decrease or loss of power output. In addition, a local failure of a cell may cause surrounding cells to be affected, resulting in a regional failure.

[0103] Therefore, the centroid of each fault point cloud area is obtained, and the normalized value of the Euclidean distance between the two-dimensional plane coordinates (X coordinate and Y coordinate) of the centroids of any two fault point cloud areas is calculated as the distance factor between the two fault point cloud areas. Then, for the convenience of explanation and illustration, any fault point cloud area is selected as the test area. The smaller the distance factor, the closer the distance between the two fault point cloud areas. Therefore, the sum of the distance factors between the test area and the other fault point cloud areas is negatively correlated and normalized to achieve logical relationship correction and obtain the fault severity factor of the test area. At this time, the larger the fault severity factor, the closer the distance between the test area and the other fault point cloud areas is, which means that these faults may belong to the failure of the same component, or they may affect each other, indicating that the risk of fault spread is higher and the fault severity is greater. The negative correlation mapping and normalization here can be performed using the formula ,in, It represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0104] The area of ​​the fault region can be characterized by the number of final fault data points therein, so the product of the total number of all final fault data points in the test area, the mean of the fault parameters of all final fault data points, and the fault severity factor of the test area is normalized to obtain the value used as the fault severity index of the test area. The fault severity index integrates multi-dimensional data. The larger its value is, the larger the coverage of the test area is, the greater the average intensity of the fault (characterized by the mean of the fault parameters of the final fault data points), and the more obvious the superposition effect of adjacent faults is, and thus the greater the fault severity of the test area is. Normalization is a technical means well known to those skilled in the art. The normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0105] At this point, the fault severity index of each fault point cloud area can be obtained. Finally, the fault severity index of each fault point cloud area is input into the Poisson disk sampling algorithm as the sampling density, so as to evenly generate updated point cloud data for each fault point cloud area; the updated point cloud data of each fault point cloud area is merged with the initial 3D point cloud data of the non-fault point cloud area (the remaining area in the initial 3D point cloud data except the fault point cloud area) to obtain the updated 3D point cloud data of the photovoltaic shooting area.

[0106] It should be noted that the DBSCAN clustering algorithm and the Poisson disk sampling algorithm are both well-known technologies, and the specific processes are not described here in detail.

[0107] Updating the 3D point cloud data at this time can achieve high-density sampling and detail repair of the fault area, so 3D modeling by updating the 3D point cloud data can improve accuracy and reliability.

[0108] In summary, a low-altitude drone is used to perform a non-contact, full-coverage scan of the photovoltaic imaging area, acquiring initial three-dimensional point cloud data. Within the photovoltaic imaging area, the similarities in height values ​​between data points at different locations are then compared to screen suspected faulty data points and preliminarily identify areas of height anomalies. These suspected faulty data points include both true faulty points and non-faulty interference points caused by changes in the drone's flight attitude or path deviation. Furthermore, because interference data points typically exhibit abrupt height differences on the photovoltaic panel, their height variations lack a clear regularity compared to surrounding data points and are typically isolated. True faulty data points, on the other hand, exhibit relatively consistent height variation trends within a local area due to their regional nature. Therefore, within the local area of ​​each suspected faulty data point, the height difference characteristics between the data point and the suspected faulty point are analyzed to determine an isolation degree value, which is used to quantify the data point's "abnormal isolation." Furthermore, the similarity and fluctuation characteristics of the height value variations of the data points are analyzed to determine a fault degree value. This captures the variation patterns of the height values ​​within the local area and quantifies the stability of the height fluctuation within the local area, helping to distinguish true faults from transient interference. Furthermore, the isolation degree and fault severity values ​​of suspected fault data points are combined to obtain fault parameters. The final fault data points are then filtered based on these parameters, effectively eliminating interference points caused by environmental interference with the drone, providing a high-purity set of fault points for subsequent 3D modeling. Finally, the point cloud data is adaptively updated based on the location, quantity distribution, and fault parameters of the final fault data points to optimize the 3D model. The final model retains the overall structural information of the PV array while accurately depicting the details of the fault area, providing highly reliable 3D data support for maintenance.

[0109] The present invention also provides a three-dimensional modeling system based on low-altitude point clouds. Figure 3 , which shows a system block diagram, including a data acquisition module 401, used to implement step S1 in the above method embodiment; a suspected fault data point screening module 402, used to implement step S2 in the above method embodiment; a fault feature analysis module 403, used to implement step S3 in the above method embodiment; and a three-dimensional point cloud data updating module 404, used to implement step S4 in the above method embodiment.

[0110] It should be noted that the system provided in the above embodiment is merely an example of the division of the aforementioned functional modules. In actual applications, the aforementioned functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the low-altitude point cloud-based 3D modeling system and the low-altitude point cloud-based 3D modeling method embodiment provided in the above embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0111] See also Figure 4 , which shows a system structure diagram of a low-altitude point cloud-based three-dimensional modeling system provided by an embodiment of the present invention, including a processor 500, a memory 501, a bus 502 and a communication interface 503, wherein the processor 500, the communication interface 503 and the memory 501 are connected via the bus 502; wherein the memory 501 may include a high-speed random access memory, the bus 502 may be an ISA bus, a PCI bus or an EISA bus, etc., and the processor 500 may be an integrated circuit chip with signal processing capabilities; the memory 501 stores at least one instruction, at least one program, a code set or an instruction set, and when the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor, the steps in a three-dimensional modeling method based on low-altitude point clouds are implemented.

[0112] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0113] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A 3D modeling method based on low-altitude point cloud, characterized in that: The method comprises: Obtain initial three-dimensional point cloud data of the photovoltaic shooting area; In the photovoltaic shooting area, analyzing the similarity between the height values ​​corresponding to the data points in different area blocks, thereby screening suspected fault data points; Within the local range corresponding to each suspected fault data point, the difference characteristics between the height values ​​of the data points are analyzed to determine the isolation degree value of each suspected fault data point; within the local range corresponding to each suspected fault data point, the similarity characteristics and difference fluctuation characteristics of the changes between the height values ​​of the data points are analyzed to determine the fault degree value of each suspected fault data point; The isolation degree value and fault degree value of each suspected fault data point are combined to obtain the fault parameter of each suspected fault data point, which is used to screen the final fault data point. The initial three-dimensional point cloud data is adaptively updated according to the position, quantity distribution and fault parameters of all final fault data points to obtain the updated three-dimensional point cloud data of the photovoltaic shooting area.

2. The three-dimensional modeling method based on low-altitude point cloud according to claim 1, characterized in that: The method for obtaining the suspected fault data point includes: In the photovoltaic shooting area, all data points are divided into regions based on two-dimensional plane coordinates to obtain multiple region blocks; In each block, the mean of the height values ​​of all data points is used as the height representation value of each block. For any two blocks, the absolute value of the difference between the height representation values ​​of the two blocks is negatively correlated and normalized to obtain the height representation similarity between the two blocks. Select any area block as the block to be tested, and use the area blocks whose high performance similarity value with the block to be tested is greater than or equal to a preset similarity threshold as similar blocks to the block to be tested; Multiply the sum of the high-performance similarity values ​​between the block to be tested and all similar blocks by the number of similar blocks, and perform negative correlation mapping on the obtained product and normalize it to obtain the value as the suspected fault index of the block to be tested; All the regional blocks are sorted in descending order according to the suspected fault index, and the data points in the first preset number of regional blocks are taken as suspected fault data points.

3. The three-dimensional modeling method based on low-altitude point cloud according to claim 1, characterized in that: The method for obtaining the isolation degree value includes: In a two-dimensional plane, with each suspected fault data point as the center, several data points are selected in each preset direction as the neighborhood data point sequence of each suspected fault data point; In the sequence of neighboring data points in each preset direction of each suspected fault data point, the absolute value of the difference in height between each neighboring data point and the suspected fault data point is used as a height difference parameter, and the sum of all height difference parameters is used as a height significance indicator of each suspected fault data point in each preset direction; Among the highly significant indicators corresponding to all preset directions of each suspected fault data point, the normalized value of the minimum highly significant indicator is used as the isolation degree value of each suspected fault data point.

4. The 3D modeling method based on low-altitude point cloud according to claim 1, characterized in that: The method for obtaining the fault degree value includes: In a two-dimensional plane, with each suspected fault data point as the center, in each preset direction, the height value change difference characteristics between the data points are compared to determine the height change data sequence of each suspected fault data point in each preset direction; In each height change data sequence of each suspected fault data point, the height difference between each two adjacent data points is calculated as the height difference factor, the variance of all height difference factors is used as the height change fluctuation factor of each height change data sequence, and the sum of the height change fluctuation factors corresponding to all height change data sequences of each suspected fault data point is negatively correlated and normalized to obtain the value as the local height change consistency of each suspected fault data point; Analyze the similarity characteristics of the height change data sequences between the preset directions of each suspected fault data point, and determine the local height change similarity value of each suspected fault data point; The normalized value of the sum of the local height change consistency and the local height change similarity value of each suspected fault data point is used as the fault degree value of each suspected fault data point.

5. The three-dimensional modeling method based on low-altitude point cloud according to claim 4, characterized in that: The method for obtaining the height change data sequence includes: Taking each suspected fault data point as the starting point, an initial sequence is generated in each preset direction, and the data points are traversed in each preset direction. If the normalized value of the absolute difference between the height value of the traversed data point and the height value of the last data point in the initial sequence is less than the preset difference threshold, the traversed data point is incorporated into the initial sequence, otherwise the traversal is stopped. After the traversal is completed, the initial sequence of each suspected fault data point in each preset direction is used as the height change data sequence.

6. The three-dimensional modeling method based on low-altitude point cloud according to claim 4, characterized in that: The method for obtaining the local height change similarity value includes: For any two preset directions, the height values ​​in the height change data sequence of each suspected fault data point in these two preset directions are processed using a linear interpolation algorithm to obtain two height value sequences of equal length. The normalized Pearson correlation coefficient of these two height value sequences is used as the height change similarity factor of each suspected fault data point in these two preset directions. The average of the height change similarity factors between all preset directions of each suspected fault data point is used as the local height change similarity value of each suspected fault data point.

7. The three-dimensional modeling method based on low-altitude point cloud according to claim 1, characterized in that: The method for obtaining the final fault data point includes: The value after negative correlation mapping of the isolation degree value of each suspected fault data point is multiplied by the fault degree value of each suspected fault data point, and the obtained product is normalized as the fault parameter of each suspected fault data point; Among all suspected fault data points, the suspected fault data points whose fault parameters are greater than the preset fault threshold are taken as the final fault data points.

8. The three-dimensional modeling method based on low-altitude point cloud according to claim 1, characterized in that: The method for acquiring updated three-dimensional point cloud data includes: All final fault data points are clustered based on the DBSCAN clustering algorithm and the three-dimensional coordinates of the final fault point to obtain all cluster clusters; The area corresponding to each cluster in the initial three-dimensional point cloud data is used as the fault point cloud area; Analyze the position between the fault point cloud areas and the number of final fault data points, and combine them with the numerical characteristics of the fault parameters of the final fault data points to obtain the fault severity index of each fault point cloud area; The fault severity index of each fault point cloud area is used as the sampling density to input into the Poisson disk sampling algorithm to uniformly generate updated point cloud data for each fault point cloud area; The updated point cloud data of each fault point cloud area is merged with the initial three-dimensional point cloud data of the non-fault point cloud area to obtain updated three-dimensional point cloud data of the photovoltaic shooting area.

9. The three-dimensional modeling method based on low-altitude point cloud according to claim 8, characterized in that: The method for obtaining the fault severity indicator includes: Obtain the centroid of each fault point cloud area, calculate the normalized Euclidean distance between the two-dimensional plane coordinates of the centroids of any two fault point cloud areas, and use this distance factor as the distance factor between the two fault point cloud areas. Select any fault point cloud area as the test area, and perform negative correlation mapping on the sum of the distance factors between the test area and the remaining fault point cloud areas, and normalize the sum to obtain the fault severity factor of the test area. The normalized value of the product of the total number of all final fault data points in the test area, the mean value of the fault parameters of all final fault data points, and the fault severity factor of the test area is used as the fault severity index of the test area.

10. A 3D modeling system based on low-altitude point cloud, characterized in that: The invention comprises a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, the steps of a three-dimensional modeling method based on a low-altitude point cloud are implemented as described in any one of claims 1 to 9.

Images