Photovoltaic panel flatness detection method and system in photovoltaic construction
By using full-coverage scanning and data processing technology, a depth information map is generated and fitted to an ideal plane, which solves the problem of foreign object interference in the existing photovoltaic panel flatness detection and improves the detection accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-26
- Publication Date
- 2026-03-27
Smart Images
Figure CN121739936A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of photovoltaic installation, in particular to a photovoltaic panel flatness detection method and system in photovoltaic construction. BACKGROUND
[0002] The photovoltaic panel, also known as a solar panel or a photovoltaic module, is a core device for converting light energy into electrical energy. During the photovoltaic construction process, various mechanical operations (such as cutting, grinding, etc.) may exert uneven forces on the photovoltaic panel, resulting in mechanical stress. This stress may cause the surface of the photovoltaic panel to appear concave or convex, and the unevenness of the surface concave or convex phenomenon will further affect the photoelectric conversion efficiency of the photovoltaic panel, thereby reducing the performance of the photovoltaic panel. Therefore, it is necessary to detect the flatness of the surface of the photovoltaic panel.
[0003] However, the existing photovoltaic panel flatness detection method still has some shortcomings, specifically: the existing photovoltaic panel flatness detection method measures the flatness of the surface of the photovoltaic panel by using a laser scanner. During data processing, it is unable to distinguish between the debris attached to the surface of the photovoltaic panel, resulting in low accuracy of the detection results. SUMMARY
[0004] The purpose of the present application is to provide a photovoltaic panel flatness detection method and system in photovoltaic construction to solve the problems raised in the background.
[0005] To achieve the above purpose, the present application provides the following technical solutions: A photovoltaic panel flatness detection method in photovoltaic construction, comprising the following steps: S1, installing an irradiation device, a collection device, and a transmission device around the photovoltaic panel, the irradiation device projects a laser line bundle onto the surface of the photovoltaic panel, the transmission device drives the irradiation device to move along the photovoltaic panel, ensuring that the laser line scans every part of the photovoltaic panel, achieving full coverage, and the collection device synchronously captures images of the laser line; S2, generating depth information of the laser line at the target position according to the images of the laser line captured by the collection device and the polar constraint calculation, obtaining 3D point cloud data of the laser line at the target position according to the depth information, and performing filtering and noise reduction processing on the 3D point cloud data to remove abnormal data, then aligning and splicing all the point cloud data to the same coordinate system to form a complete point cloud data set; S3, fitting an ideal plane from the processed point cloud data set by least squares method as the reference for flatness calculation, calculating the distance from each measurement point to the reference plane to obtain the overall flatness, comparing the calculated flatness, maximum concave / convex height and other parameters with the preset tolerance threshold, and generating a report containing a two-dimensional / three-dimensional color mapping, a key data table and a detection conclusion according to the comparison result.
[0006] As a preferred scheme of the present application, the installation steps of the irradiation device and the acquisition device in S1 are as follows: using a marker pen or adopting a small dot sticker method to set up a marked control point around the photovoltaic panel, constructing a control network of the measurement area, fixing the acquisition device on the marked control point, adjusting the installation position and angle to ensure that the acquisition device can capture the whole photovoltaic panel, installing a conveyor belt in front of the photovoltaic panel, and then installing the irradiation device on the conveying device.
[0007] As a preferred scheme of the present application, the generation method of the depth information map in S2 is as follows: constructing an image generation model, obtaining a plurality of laser line projection image training sets projected on the photovoltaic panel under different phase shifts and different frequencies, training the image generation model using the laser line projection image training set, fixing the parameters after the training is stable, inputting the laser line image captured by the acquisition device into the image generation model, and calculating and processing the absolute phase shift map of the laser line image by the image generation model to obtain the depth information map of the laser line image according to the absolute phase shift map and the polar line constraint.
[0008] As a preferred scheme of the present application, the specific method for filtering and denoising the 3D point cloud data in S2 is as follows: removing the point cloud data of external dynamic interference from the point cloud data map, performing layered feature analysis on the point cloud data after removing the interference by using a multi-scale pyramid structure, extracting local density, curvature and normal change features at scales of 0.5 m, 1.0 m and 2.0 m respectively, generating a multi-dimensional feature matrix, performing importance evaluation on the feature matrix based on a random forest algorithm, screening parameters with a feature importance threshold higher than 0.7, dynamically generating an adaptive filtering rule, removing outlier noise points at a retention rate of 60% to 80% according to the adaptive rule, retaining potential effective point clouds, combining the motion trajectories of continuous frame point clouds, analyzing the displacement consistency of point clouds between adjacent frames by a dynamic time warping algorithm, marking non-consistent points as dynamic noise and removing them again, constructing a Gaussian mixture model based on local geometric features to finely remove residual noise at a retention rate of 85% to 95%, and performing topological connectivity detection on the filtered point cloud to complete the key structure points lost due to filtering by a region growing algorithm, and outputting the denoised point cloud data.
[0009] As a preferred scheme of the present application, the method for removing the dynamic interference point cloud data from the outside world comprises the following steps: acquiring a plurality of frames of first point cloud data from a point cloud data graph, acquiring a first distance corresponding to each scanning line according to the first point cloud data, determining whether the first distance corresponding to the first scanning line is greater than or equal to a preset distance threshold, and if so, regarding the plurality of frames of first point cloud data corresponding to the first scanning line as a plurality of frames of second point cloud data, determining the absolute position of the edge point of the point cloud of the interference object according to the absolute position of the acquisition device when the second point cloud data is acquired and the first distance corresponding to the first scanning line in the second point cloud data, determining whether the interference object contains a dynamic interference object, regarding the plurality of frames of second point cloud data in which the absolute position of the interference object changes with time as third point cloud data if the interference object contains a dynamic interference object, determining a second scanning line including the point cloud of the interference object according to the third point cloud data, and determining the first segment of continuous point cloud corresponding to each second scanning line as the point cloud of the dynamic interference object along the direction in which the elevation value increases, and removing the point cloud data determined as the point cloud of the dynamic interference object from the point cloud data graph.
[0010] As a preferred scheme of the present application, the first distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the object actually scanned, and the reference distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the photovoltaic panel.
[0011] As a preferred scheme of the present application, the method for determining the dynamic interference object comprises the following steps: if the absolute position of the interference object in at least two consecutive frames of second point cloud data changes with time, determining that the interference object is a dynamic interference object, and otherwise, determining that the interference object is a static interference object.
[0012] As a preferred scheme of the present application, the specific method for aligning and splicing all the point cloud data in S2 to the same coordinate system comprises the following steps: extracting the corner points or edge points in the point cloud as key feature points, iteratively optimizing the rotation matrix and the translation vector by minimizing the Euclidean distance between the feature point pairs, taking the key feature point coordinate data as a reference to perform coordinate conversion on the point cloud data and establish a coordinate system, and performing coordinate conversion on all the point cloud data according to the coordinate conversion parameters to complete the coordinate system conversion.
[0013] A photovoltaic panel flatness detection system in photovoltaic construction, comprising a control processor, an irradiation module and an acquisition module, characterized in that: the control processor stores a program model for implementing the photovoltaic panel flatness detection method in photovoltaic construction according to any one of claims 1-8, the irradiation module is various line laser profilers, and the acquisition module comprises various high-resolution industrial cameras.
[0014] Compared with the prior art, the present application has the following advantages: 1. In this invention, an irradiation device, a data acquisition device, and a transmission device are installed around a photovoltaic panel. The irradiation device projects a laser beam onto the surface of the photovoltaic panel, and the transmission device moves the irradiation device along the photovoltaic panel to ensure that the laser beam scans every part of the photovoltaic panel, achieving full coverage. The data acquisition device simultaneously captures images of the laser beam. Based on the images captured by the data acquisition device and epipolar constraints, a depth information map of the laser beam at the target location is calculated. Then, 3D point cloud data of the laser beam at the target location is obtained based on the depth information map. The 3D point cloud data is filtered and noise-reduced to remove abnormal data. All point cloud data are then aligned and stitched together in the same coordinate system to form a complete point cloud dataset. An ideal plane is fitted from the processed point cloud dataset using the least squares method as the benchmark for flatness calculation. The flatness of each measurement is then calculated. The distance from the measurement point to the reference plane is used to obtain the overall flatness. The calculated flatness, maximum indentation / bulge height, and other parameters are compared with preset tolerance thresholds. Based on the comparison results, a report containing two-dimensional / three-dimensional color mapping maps, key data tables, and inspection conclusions is generated. During data processing, interference point clouds in the point cloud data are first removed, and then the point cloud data is filtered and denoised. The efficient identification and suppression of dynamic noise is achieved by utilizing the spatiotemporal correlation of fusion. Combined with the consistency verification of the motion trajectory of multi-frame point clouds, the dynamic interference is accurately located, effectively reducing the interference of external debris on the inspection results. This solves the problem that existing photovoltaic panel flatness inspection methods, which measure the flatness of the photovoltaic panel surface using laser scanners, cannot distinguish debris attached to the photovoltaic panel surface during data processing, resulting in low accuracy of the inspection results. Attached Figure Description
[0015] Figure 1 This is a schematic diagram of the overall hardware structure of the present invention.
[0016] In the diagram: 1-Control processor, 2-Illumination module, 3-Acquisition module. Detailed Implementation
[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0018] To facilitate understanding of the present invention, a more comprehensive description of the invention will be given below with reference to the accompanying drawings, and several embodiments of the invention will be provided. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the invention will be more thorough and complete.
[0019] For examples, please refer to Figure 1 The present invention provides a technical solution: A method for detecting the flatness of photovoltaic panels during photovoltaic construction includes the following steps: S1, the irradiation device, the acquisition device and the transmission device are installed around the photovoltaic panel. The irradiation device projects the laser beam onto the surface of the photovoltaic panel. The transmission device will drive the irradiation device to move along the photovoltaic panel to ensure that the laser beam scans every part of the photovoltaic panel to achieve full coverage. The acquisition device will simultaneously capture and photograph the image of the laser beam. S2. Based on the image of the laser line captured by the acquisition device and the epipolar constraint, a depth information map of the laser line at the target position is calculated and generated. Then, 3D point cloud data of the laser line at the target position is obtained based on the depth information map. The 3D point cloud data is filtered and noise-reduced to remove abnormal data. Finally, all point cloud data are aligned and stitched together in the same coordinate system to form a complete point cloud dataset. S3 uses the least squares method to fit an ideal plane from the processed point cloud dataset as the benchmark for flatness calculation. It calculates the distance from each measurement point to the benchmark plane to obtain the overall flatness. The calculated flatness, maximum depression / bulge height and other parameters are compared with the preset tolerance threshold, and a report containing two-dimensional / three-dimensional color mapping map, key data table and detection conclusion is generated based on the comparison results.
[0020] Furthermore, in S1, the installation steps for the irradiation equipment and the acquisition equipment are as follows: use a marker pen or small dot stickers to set up control points around the photovoltaic panel to construct a control network for the measurement area. Fix the acquisition equipment on the control points, adjust the installation position and angle to ensure that the acquisition equipment can capture the entire photovoltaic panel, install a conveyor belt in front of the photovoltaic panel, and then install the irradiation equipment on the conveyor belt.
[0021] Furthermore, the method for generating the depth information map in S2 is as follows: an image generation model is constructed, a training set of laser line projection images projected onto the photovoltaic panel at different phase shifts and frequencies is obtained, the image generation model is trained using the laser line projection image training set, and after the training is stable, the parameters are fixed, the laser line images captured by the acquisition device are input into the image generation model, the image generation model calculates and processes to obtain the absolute phase shift map of the laser line image, and the depth information map of the laser line image is calculated based on the absolute phase shift map and epipolar constraints.
[0022] Furthermore, the specific method for filtering and denoising 3D point cloud data in S2 is as follows: Point cloud data with external dynamic interference removed is processed using a multi-scale pyramid structure to perform hierarchical feature analysis on the interference-free point cloud data. Local density, curvature, and normal variation features are extracted at scales of 0.5m, 1.0m, and 2.0m, generating a multi-dimensional feature matrix. The importance of the feature matrix is evaluated based on the random forest algorithm, and parameters with an importance threshold higher than 0.7 are selected. Adaptive filtering rules are dynamically generated, and outlier noise points are quickly removed with a retention rate of 60%–80% according to the adaptive rules, preserving potentially effective point clouds. Combining the motion trajectory of continuous frame point cloud data, the displacement consistency of point clouds between adjacent frames is analyzed using a dynamic time warping algorithm. Inconsistent points are marked as dynamic noise and removed a second time. A Gaussian mixture model is constructed based on local geometric features, and residual noise is finely filtered out with a retention rate of 85%–95%. Topological connectivity is detected on the filtered point cloud, and key structural points lost due to filtering are completed using a region growing algorithm. The denoised point cloud data is then output.
[0023] Furthermore, the method for removing point cloud data of external dynamic interference objects is as follows: First, multiple frames of first point cloud data are acquired from the point cloud data map. Based on the first point cloud data, a first distance corresponding to each scan line is obtained. If the difference between the first distance corresponding to the first scan line and the reference distance is greater than or equal to a preset distance threshold, then the multiple frames of first point cloud data corresponding to the first scan line are used as multiple frames of second point cloud data. Based on the absolute position when the acquisition device acquires the second point cloud data and the first distance corresponding to the first scan line in the second point cloud data, the absolute position of the edge points of the point cloud of the interference object is determined. It is then determined whether the interference object contains dynamic interference objects. If it does, the multiple frames of second point cloud data whose absolute position changes over time are used as third point cloud data. Based on the third point cloud data, a second scan line including the point cloud of the interference object is determined. Taking the edge line of the photovoltaic panel as the starting point, along the direction of increasing elevation value, the first continuous point cloud segment corresponding to each second scan line is determined as the point cloud of the dynamic interference object. The point cloud data determined to be dynamic interference objects is removed from the point cloud data map.
[0024] Furthermore, the first distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the actual scanned object, and the reference distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the photovoltaic panel. In the third point cloud data, the scan line corresponding to the first distance whose difference from the reference distance is greater than or equal to the preset distance threshold is determined as the second scan line.
[0025] Furthermore, the method for determining dynamic interference objects is as follows: if the absolute position of the interference object in the second point cloud data of at least two consecutive frames changes over time, then the interference object is determined to be a dynamic interference object; otherwise, it is a static interference object.
[0026] Furthermore, the specific method for aligning and stitching all point cloud data in S2 to the same coordinate system is as follows: extract corner points or edge points in the point cloud as key feature points, iteratively optimize the rotation matrix and translation vector by minimizing the Euclidean distance between feature point pairs, perform coordinate transformation on the point cloud data based on the coordinate data of the key feature points and establish a coordinate system, and perform coordinate transformation on all point cloud data according to the coordinate transformation parameters to complete the coordinate system transformation.
[0027] A photovoltaic panel flatness detection system for photovoltaic construction includes a control processor 1, an irradiation module 2, and a data acquisition module 3. The control processor 1 stores a program model that implements a photovoltaic panel flatness detection method for photovoltaic construction as described in any one of claims 1-8. The irradiation module 2 is a line laser profilometer, and the data acquisition module 3 includes a high-resolution industrial camera.
[0028] Specific Implementation Cases Mark control points are set around the photovoltaic panel using markers or small dot stickers to construct a control network for the measurement area. The acquisition equipment is then fixedly installed on the marked control points, and the installation position and angle are adjusted to ensure that the acquisition equipment can capture the entire photovoltaic panel. A conveyor belt is installed in front of the photovoltaic panel, and the irradiation equipment is then installed on the conveyor belt. The irradiation equipment projects a laser beam onto the surface of the photovoltaic panel, and the conveyor belt will move the irradiation equipment along the photovoltaic panel to ensure that the laser line scans every part of the photovoltaic panel to achieve full coverage. The acquisition equipment will simultaneously capture and photograph the image of the laser line. An image generation model is constructed, and a training set of laser line projection images projected onto a photovoltaic panel at different phase shifts and frequencies is obtained. The image generation model is trained using the laser line projection image training set. After the training is stable, the parameters are fixed, and the laser line images captured by the acquisition device are input into the image generation model. The image generation model calculates and processes the laser line images to obtain the absolute phase shift map of the laser line images. The depth information map of the laser line images is calculated based on the absolute phase shift map and epipolar constraints. Then, based on the depth information map, obtain the 3D point cloud data of the laser line at the target position. From the point cloud data map, obtain the collected first point cloud data in multiple frames. Based on the first point cloud data, obtain the first distance corresponding to each scan line. If the difference between the first distance corresponding to the first scan line and the reference distance is greater than or equal to the preset distance threshold, then the multiple frames of first point cloud data corresponding to the first scan line are used as multiple frames of second point cloud data. The first distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the actual scanned object. The reference distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the photovoltaic panel. Based on the absolute position of the second point cloud data collected by the acquisition device and the first distance corresponding to the first scan line in the second point cloud data, the absolute position of the edge points of the point cloud of the interfering object is determined. If the absolute position of the interfering object in the second point cloud data of at least two consecutive frames changes over time, the interfering object is determined to be a dynamic interfering object; otherwise, it is a static interfering object. The multiple frames of second point cloud data in which the absolute position of the interfering object changes over time are used as the third point cloud data. Based on the third point cloud data, the second scan line of the point cloud including the interfering object is determined. The scan line in the third point cloud data corresponding to the first distance whose difference from the reference distance is greater than or equal to the preset distance threshold is determined as the second scan line. Taking the edge line of the photovoltaic panel as the starting point, along the direction of increasing elevation value, the first segment of continuous point cloud corresponding to each second scan line is determined as the point cloud of the dynamic interfering object. The point cloud data determined to be the dynamic interfering object is removed from the point cloud data map. Point cloud data with external dynamic interference removed is analyzed using a multi-scale pyramid structure. Local density, curvature, and normal variation features are extracted at scales of 0.5m, 1.0m, and 2.0m to generate a multi-dimensional feature matrix. The feature matrix is evaluated for importance based on a random forest algorithm, and parameters with an importance threshold higher than 0.7 are selected. Adaptive filtering rules are dynamically generated, and outlier noise points are quickly removed with a retention rate of 60%–80% according to the adaptive rules, retaining potentially effective point clouds. Combining the motion trajectory of continuous frame point cloud data, the displacement consistency of point clouds between adjacent frames is analyzed using a dynamic time warping algorithm. Inconsistent points are marked as dynamic noise and removed a second time. A Gaussian mixture model is constructed based on local geometric features to finely filter residual noise with a retention rate of 85%–95%. Topological connectivity is detected on the filtered point cloud, and key structural points lost due to filtering are filled in using a region growing algorithm. The denoised point cloud data is then output. Corner or edge points in the point cloud are extracted as key feature points. By minimizing the Euclidean distance between feature point pairs, the rotation matrix and translation vector are iteratively optimized. The point cloud data is transformed and a coordinate system is established based on the coordinate data of the key feature points. All point cloud data are transformed according to the coordinate transformation parameters to complete the coordinate system transformation. The above steps are repeated to align and stitch all the processed point cloud data into the same coordinate system to form a complete point cloud dataset. An ideal plane is fitted from the processed point cloud dataset using the least squares method and used as the benchmark for flatness calculation. The distance from each measurement point to the benchmark plane is calculated to obtain the overall flatness. The calculated flatness, maximum depression / bulge height, and other parameters are compared with preset tolerance thresholds, and a report containing two-dimensional / three-dimensional color mapping maps, key data tables, and detection conclusions is generated based on the comparison results.
[0029] All standard parts used in this application can be purchased from the market. The specific connection methods of each part adopt conventional methods such as bolts, rivets, and welding that are mature in the prior art. The machinery, parts and equipment adopt conventional models in the prior art and are also general components, which are common knowledge in this field.
[0030] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for detecting the flatness of photovoltaic panels during photovoltaic construction, characterized in that, Includes the following steps: S1, the irradiation device, the acquisition device and the transmission device are installed around the photovoltaic panel. The irradiation device projects the laser beam onto the surface of the photovoltaic panel. The transmission device will drive the irradiation device to move along the photovoltaic panel to ensure that the laser beam scans every part of the photovoltaic panel to achieve full coverage. The acquisition device will simultaneously capture and photograph the image of the laser beam. S2. Based on the image of the laser line captured by the acquisition device and the epipolar constraint, a depth information map of the laser line at the target position is calculated and generated. Then, 3D point cloud data of the laser line at the target position is obtained based on the depth information map. The 3D point cloud data is filtered and noise-reduced to remove abnormal data. Finally, all point cloud data are aligned and stitched together in the same coordinate system to form a complete point cloud dataset. S3 uses the least squares method to fit an ideal plane from the processed point cloud dataset as the benchmark for flatness calculation. It calculates the distance from each measurement point to the benchmark plane to obtain the overall flatness. The calculated flatness, maximum depression / bulge height and other parameters are compared with the preset tolerance threshold, and a report containing two-dimensional / three-dimensional color mapping map, key data table and detection conclusion is generated based on the comparison results.
2. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 1, characterized in that: The installation steps of the irradiation device and the acquisition device in S1 are as follows: use a marker pen or small dot stickers to set up control points around the photovoltaic panel to build a control network for the measurement area. Fix the acquisition device on the control points, adjust the installation position and angle to ensure that the acquisition device can capture the entire photovoltaic panel. Install a conveyor belt in front of the photovoltaic panel and then install the irradiation device on the conveyor belt.
3. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 1, characterized in that: The method for generating the depth information map in S2 is as follows: an image generation model is constructed, a training set of laser line projection images projected onto a photovoltaic panel at different phase shifts and frequencies is obtained, the image generation model is trained using the laser line projection image training set, and after the training is stable, the parameters are fixed, the laser line images captured by the acquisition device are input into the image generation model, the image generation model calculates and processes to obtain the absolute phase shift map of the laser line image, and the depth information map of the laser line image is calculated based on the absolute phase shift map and epipolar constraints.
4. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 1, characterized in that: The specific method for filtering and denoising 3D point cloud data in S2 is as follows: Point cloud data with external dynamic interference removed is processed. A multi-scale pyramid structure is used to perform hierarchical feature analysis on the interference-free point cloud data. Local density, curvature, and normal variation features are extracted at scales of 0.5m, 1.0m, and 2.0m, respectively, generating a multi-dimensional feature matrix. The importance of the feature matrix is evaluated based on the random forest algorithm, and parameters with a feature importance threshold higher than 0.7 are selected. Adaptive filtering rules are dynamically generated. Outlier noise points are quickly removed with a retention rate of 60%–80% according to the adaptive rules, preserving potentially effective point clouds. Combining the motion trajectory of continuous frame point cloud data, the displacement consistency of point clouds between adjacent frames is analyzed using a dynamic time warping algorithm. Inconsistent points are marked as dynamic noise and removed a second time. A Gaussian mixture model is constructed based on local geometric features, and residual noise is finely filtered out with a retention rate of 85%–95%. Topological connectivity detection is performed on the filtered point cloud. Key structural points lost due to filtering are completed using a region growing algorithm, and the denoised point cloud data is output.
5. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 4, characterized in that: The method for removing point cloud data of external dynamic interference objects is as follows: First, multiple frames of first point cloud data are acquired from the point cloud data map. Based on the first point cloud data, a first distance corresponding to each scan line is obtained. If the difference between the first distance corresponding to the first scan line and the reference distance is greater than or equal to a preset distance threshold, then the multiple frames of first point cloud data corresponding to the first scan line are used as multiple frames of second point cloud data. Based on the absolute position of the acquisition device when acquiring the second point cloud data and the first distance corresponding to the first scan line in the second point cloud data, the absolute position of the edge points of the point cloud of the interference object is determined. It is then determined whether the interference object contains dynamic interference objects. If it does, the multiple frames of second point cloud data whose absolute position changes over time are used as third point cloud data. Based on the third point cloud data, a second scan line including the point cloud of the interference object is determined. Taking the edge line of the photovoltaic panel as the starting point, along the direction of increasing elevation value, the first continuous point cloud segment corresponding to each second scan line is determined as the point cloud of the dynamic interference object. The point cloud data determined to be dynamic interference objects is removed from the point cloud data map.
6. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 5, characterized in that: The first distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the actual scanned object. The reference distance is the distance from the motion trajectory line of the irradiation device to the edge point of the point cloud of the photovoltaic panel. In the third point cloud data, the scan line corresponding to the first distance whose difference from the reference distance is greater than or equal to the preset distance threshold is determined as the second scan line.
7. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 5, characterized in that: The method for determining dynamic interference objects is as follows: if the absolute position of the interference object in the second point cloud data of at least two consecutive frames changes over time, then the interference object is determined to be a dynamic interference object; otherwise, it is a static interference object.
8. The method for detecting the flatness of photovoltaic panels during photovoltaic construction according to claim 1, characterized in that: The specific method for aligning and stitching all point cloud data in S2 into the same coordinate system is as follows: extract corner points or edge points in the point cloud as key feature points, iteratively optimize the rotation matrix and translation vector by minimizing the Euclidean distance between feature point pairs, perform coordinate transformation on the point cloud data based on the coordinate data of the key feature points and establish a coordinate system, and perform coordinate transformation on all point cloud data according to the coordinate transformation parameters to complete the coordinate system transformation.
9. A photovoltaic panel flatness detection system during photovoltaic construction, comprising a control processor (1), an irradiation module (2), and a data acquisition module (3), characterized in that: The control processor (1) stores a program model for implementing a photovoltaic panel flatness detection method in photovoltaic construction as described in any one of claims 1-8. The irradiation module (2) is a line laser profilometer, and the acquisition module (3) includes a variety of high-resolution industrial cameras.