Modeling method and device of photovoltaic power station, electronic equipment and storage medium

Abstract

The invention provides a modeling method and device for a photovoltaic power station, electronic equipment and a storage medium, and belongs to the technical field of computer vision, and the method comprises the steps: obtaining remote sensing images of the photovoltaic power station at a plurality of visual angles; generating sparse point clouds based on the remote sensing images of the multiple visual angles; three-dimensional reconstruction is conducted on the sparse point cloud through an improved Gaussian splashing processing mode, a three-dimensional model of the photovoltaic power station is generated, and the improved Gaussian splashing processing mode comprises the steps that on the basis of a self-adaptive Gaussian point control strategy, through self-adaptive adjustment of Gaussian points, geometric and visual features of the photovoltaic power station can be captured more accurately, and the three-dimensional model of the photovoltaic power station is generated. The precision of the three-dimensional model of the photovoltaic power station is improved, the complexity and calculation cost of the model are effectively controlled, unnecessary calculation resource consumption is reduced, and the modeling efficiency of the photovoltaic power station is improved, so that the modeling quality and efficiency of the photovoltaic power station are improved.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, specifically to a modeling method, apparatus, electronic device, and storage medium for photovoltaic power plants. Background Technology

[0002] With the continuous advancement of technology, the planning, design, maintenance strategies, performance analysis, environmental impact assessment, and compliance with regulatory requirements of photovoltaic power plants are gradually becoming more intelligent and refined. Against this backdrop, 3D reconstruction technology, with its intuitive and comprehensive visual presentation, can support the full lifecycle management of photovoltaic power plants, thereby improving their economic and social benefits.

[0003] In related technologies, neural network-based 3D rendering techniques, such as Neural Radiance Fields (NeRF), are commonly used to generate 3D images by learning the radiance distribution of scenes in photovoltaic power plants; alternatively, lidar point cloud modeling technology is used to model photovoltaic power plants.

[0004] However, for 3D modeling of real-world photovoltaic power plants, the sheer size of the scene presents several challenges. Firstly, using LiDAR for point cloud modeling is complex due to its data processing requirements and high computational resource demands. Furthermore, objects with low or high reflectivity surfaces may cause data loss or distortion, resulting in inaccurate modeling. Secondly, NeRF-based modeling methods place high demands on computational resources and training time. This leads to limitations in high-precision photovoltaic power plant modeling, including high computational resource requirements, complex data processing, and long training times, ultimately reducing the quality and efficiency of photovoltaic power plant modeling. Summary of the Invention

[0005] In view of this, it is necessary to provide a modeling method, apparatus, electronic equipment and storage medium for photovoltaic power plants to solve the technical problems of low modeling quality and efficiency of photovoltaic power plants in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a modeling method for photovoltaic power plants, comprising: Based on the remote sensing images from the multiple perspectives, a sparse point cloud is generated; An improved Gaussian splashing processing method is used to perform three-dimensional reconstruction of the sparse point cloud to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

[0007] In one possible implementation, an improved Gaussian splashing method is used to perform three-dimensional reconstruction of the sparse point cloud to generate a three-dimensional model of the photovoltaic power station, including: The sparse point cloud is reconstructed in three dimensions using a Gaussian splashing method to generate a reconstructed region. Based on the density of the reconstructed Gaussian points in the reconstructed region, the number and properties of the reconstructed Gaussian points are adaptively adjusted to generate a three-dimensional model of the photovoltaic power station.

[0008] In one possible implementation, the step of adaptively adjusting the number and properties of the reconstructed Gaussian points based on the density of the reconstructed Gaussian points in the reconstructed region to generate a three-dimensional model of the photovoltaic power station includes: Based on the density of reconstructed Gaussian points in the reconstructed region, over-reconstructed sub-regions and under-reconstructed sub-regions are identified from the reconstructed region; A first target Gaussian point is selected from the under-reconstructed sub-region for replication, and the properties of the second target Gaussian point in the over-reconstructed sub-region are optimized to adaptively adjust the number and properties of the reconstructed Gaussian points. The first target Gaussian point is determined based on the attributes of each reconstructed Gaussian point in the underreconstructed sub-region, and the second target Gaussian point is determined based on the characteristics of each reconstructed Gaussian point in the overreconstructed sub-region.

[0009] In one possible implementation, the method further includes: Determine the rendering contribution of the reconstructed Gaussian points in the reconstructed region; The reconstructed Gaussian points whose rendering contribution does not meet the preset rendering conditions are deleted.

[0010] In one possible implementation, generating a sparse point cloud based on the remote sensing images from the multiple viewpoints includes: Extract image features from the remote sensing images from the multiple perspectives; Image matching is performed based on the image features to obtain multiple sets of image pairs; Relative camera pose estimation is performed on multiple sets of image pairs to determine the image camera pose corresponding to each remote sensing image in the image pair, and the corresponding three-dimensional point cloud is calculated based on the image camera pose to reconstruct the sparse point cloud corresponding to the photovoltaic power station.

[0011] In one possible implementation, the step of using Gaussian splashing to perform 3D reconstruction of the sparse point cloud and generate a reconstructed region includes: The primitives of Gaussian sputtering are initialized based on the sparse point cloud to obtain the corresponding initial Gaussian points; Determine the image frustum range of the remote sensing image corresponding to the sparse point cloud; The initial Gaussian point of the image view frustum is rendered using preset Gaussian point parameters to generate the initial rendered image. Calculate the loss value between the initial rendered image and the preset real image, and optimize the initial rendered image based on the loss value to generate the reconstructed region.

[0012] In one possible implementation, optimizing the initial rendered image based on the loss value to generate the reconstructed region includes: The initial rendered image is optimized at least once based on the loss value, wherein each iteration includes adjusting the Gaussian point parameters; The initial Gaussian point is rendered using the adjusted Gaussian point parameters to generate the reconstructed region.

[0013] Secondly, the present invention also provides a modeling apparatus for a photovoltaic power plant, comprising: The acquisition unit is used to acquire remote sensing images of the photovoltaic power station from multiple perspectives; The first generation unit is used to generate sparse point clouds based on the remote sensing images from the multiple perspectives. The second generation unit is used to perform three-dimensional reconstruction of the sparse point cloud using an improved Gaussian splashing processing method to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

[0014] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the photovoltaic power plant modeling method described in any of the above implementations.

[0015] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the photovoltaic power station modeling method described in any of the above implementations.

[0016] The beneficial effects of this invention are: The photovoltaic power plant modeling method provided by this invention acquires remote sensing images of the photovoltaic power plant from multiple perspectives. Because it processes high-precision remote sensing images, the accuracy of the sparse point cloud is improved. Furthermore, since the sparse point cloud contains key feature points of the scene, it is beneficial to improve subsequent modeling efficiency. An improved Gaussian splashing processing method is used to perform 3D reconstruction of the sparse point cloud, generating a 3D model of the photovoltaic power plant. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy. By adaptively adjusting the Gaussian points, the geometric and visual features of the photovoltaic power plant can be captured more accurately, improving the accuracy of the 3D model of the photovoltaic power plant. At the same time, it effectively controls the complexity and computational cost of the model, reduces unnecessary consumption of computational resources, and improves the modeling efficiency of the photovoltaic power plant, thereby improving the modeling quality and efficiency of the photovoltaic power plant. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of an embodiment of the photovoltaic power plant modeling method provided by the present invention; Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S103; Figure 3 For the present invention Figure 2 A schematic diagram of an embodiment of S202; Figure 4 A schematic flowchart illustrating an embodiment of the present invention for processing reconstructed Gaussian points in a reconstruction region; Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 6 For the present invention Figure 5 A schematic diagram of an embodiment of S301; Figure 7 For the present invention Figure 2 A schematic flowchart of an embodiment of S201 Figure 8 A schematic diagram of an embodiment of the photovoltaic power plant modeling device provided by the present invention; Figure 9 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] This invention provides a modeling method, apparatus, electronic device, and storage medium for photovoltaic power plants, which will be described below.

[0024] Figure 1 This is a schematic flowchart of an embodiment of the photovoltaic power plant modeling method provided by the present invention, as shown below. Figure 1 As shown, the modeling methods for photovoltaic power plants include: S101. Acquire remote sensing images of the photovoltaic power station from multiple perspectives.

[0025] Remote sensing images refer to digital images of a real-world photovoltaic power station from different angles. For example, these images can be RGB images. Remote sensing images can be obtained by acquiring remote sensing video from different perspectives using remote sensing image acquisition equipment, or they can be obtained from pre-stored remote sensing video from multiple perspectives of the photovoltaic power station. To improve modeling efficiency, the acquired remote sensing images can undergo data cleaning processing to obtain remote sensing images suitable for modeling the photovoltaic power station.

[0026] Specifically, drones or other remote sensing equipment can be used to capture remote sensing images and videos of a photovoltaic power station from multiple perspectives. Then, a preset number of video frames are extracted from the remote sensing images and videos. The extracted video frames are cleaned and processed to remove video frames that do not meet the quality requirements, thereby obtaining remote sensing images. This improves the quality of the remote sensing images and facilitates subsequent improvements in the modeling quality of the photovoltaic power station.

[0027] In one specific implementation, a preset number of video frames can be extracted from remote sensing images and videos. This can be achieved by first determining the flight path and speed of the UAV, and then calculating the frame extraction interval based on the UAV's flight path and speed. For example, video frames can be extracted based on the UAV's moving speed and shooting range, ensuring that adjacent frames have more than 60% scene overlap to facilitate subsequent processing.

[0028] In one specific implementation, the data cleaning process for video frames can be to use an image quality assessment algorithm to evaluate the image quality and delete images with quality below a preset threshold, or to delete images with significant motion blur, or a combination of the two processing methods.

[0029] S102. Based on the remote sensing images from the multiple perspectives, generate a sparse point cloud.

[0030] In this context, a point cloud is a relatively dispersed set of points in three-dimensional space. Each point contains location information (usually X, Y, and Z coordinates) and may also contain other attributes such as color, intensity, or normal direction. In this embodiment, the sparse point cloud represents a photovoltaic power station scene and is relatively sparsely distributed. Sparse point clouds contain key feature points in the scene, such as corners and edges, which helps in quickly identifying and understanding the scene's structure. Furthermore, compared to dense point clouds, sparse point clouds contain fewer points, thus requiring less computational resources and time for processing and analysis, which improves subsequent modeling efficiency.

[0031] The inventors discovered that while LiDAR can collect point cloud data from photovoltaic power plants, the scenes of these plants are vast, resulting in a large volume of point cloud data. Furthermore, objects with low or strong reflectivity may cause data loss or distortion, reducing the accuracy of the point cloud data. Therefore, in this embodiment, sparse point clouds are generated based on remote sensing images from multiple perspectives. This improves the accuracy of sparse point clouds. Since sparse point clouds contain key feature points in the scene, they help to quickly identify and understand the scene's structure. In addition, sparse point clouds contain fewer points, thus requiring less computational resources and time for processing and analysis, which is beneficial for improving the efficiency of subsequent modeling.

[0032] Specifically, the pose estimation of the camera corresponding to the remote sensing image can be obtained using the Structure-from-Motion (SfM) method, and a sparse point cloud can be established by combining the pose estimation. Understandably, in this embodiment, a sparse point cloud is generated from remote sensing images from multiple perspectives. Since the remote sensing images are processed with high precision, the accuracy of the sparse point cloud can be improved. Furthermore, because the sparse point cloud contains key feature points in the scene, it helps to quickly identify and understand the structure of the scene. At the same time, the sparse point cloud contains fewer points, thus requiring less computational resources and time for processing and analysis, which is beneficial to improving the efficiency of subsequent modeling. S103. The sparse point cloud is reconstructed in three dimensions using an improved Gaussian splashing processing method to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

[0033] Gaussian splashing is a volumetric rendering technique that uses Gaussian functions to represent points in three-dimensional space. This involves extracting data from point clouds or mesh models, converting it into Gaussian points, and then rendering it using Gaussian splashing technology to form the final image.

[0034] The improved Gaussian splashing processing method is based on Gaussian splashing, and adds an adaptive Gaussian point control strategy to the processing method.

[0035] The adaptive Gaussian point control strategy is a method that dynamically adjusts the distribution and properties of Gaussian points, which can optimize the 3D reconstruction process.

[0036] Specifically, an improved Gaussian splashing processing method is used for 3D reconstruction of sparse point clouds. This involves initial reconstruction using Gaussian splashing, followed by further reconstruction based on an adaptive Gaussian point control strategy to generate a 3D model of the photovoltaic power station, thus improving the accuracy of the 3D model. Understandably, in this embodiment, by adaptively adjusting the Gaussian points, the geometric and visual features of the photovoltaic power station can be captured more accurately, improving the accuracy of the 3D model while effectively controlling model complexity and computational cost, reducing unnecessary computational resource consumption, and improving the modeling efficiency of the photovoltaic power station.

[0037] In summary, the photovoltaic power plant modeling method provided in this embodiment of the invention acquires remote sensing images of the photovoltaic power plant from multiple perspectives. Since the processing of these high-precision remote sensing images improves the accuracy of the sparse point cloud, and because the sparse point cloud contains key feature points of the scene, it is beneficial to improve subsequent modeling efficiency. An improved Gaussian splashing processing method is used to perform 3D reconstruction of the sparse point cloud, generating a 3D model of the photovoltaic power plant. This improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy. By adaptively adjusting the Gaussian points, the geometric and visual features of the photovoltaic power plant can be captured more accurately, improving the accuracy of the 3D model. Simultaneously, it effectively controls the complexity and computational cost of the model, reduces unnecessary computational resource consumption, and improves the modeling efficiency of the photovoltaic power plant, thereby improving the modeling quality and efficiency of the photovoltaic power plant.

[0038] In some embodiments of the present invention, such as Figure 2 As shown, step S103 includes: S201. The sparse point cloud is reconstructed in three dimensions using Gaussian splashing to generate a reconstructed region. S202. Based on the density of the reconstructed Gaussian points in the reconstructed region, adaptively adjust the number and attributes of the reconstructed Gaussian points to generate a three-dimensional model of the photovoltaic power station.

[0039] The reconstructed region refers to the rendered image obtained by directly rendering the sparse point cloud using the Gaussian splashing method. The reconstructed region includes multiple reconstructed Gaussian points.

[0040] Specifically, a 3D reconstruction of sparse point clouds can be performed using a Gaussian splashing method. After generating the reconstructed region, the reconstruction effect of each sub-region within the reconstructed region is determined based on the density of the reconstructed Gaussian points. Based on the reconstruction effect, the number and attributes of the reconstructed Gaussian points are dynamically adjusted to generate a 3D model of the photovoltaic power station. Understandably, in this embodiment, the number and attributes of the reconstructed Gaussian points are dynamically adjusted based on the density of the reconstructed Gaussian points in the reconstructed region, thereby optimizing the rendering effect and improving the accuracy of the 3D model of the photovoltaic power station.

[0041] In some embodiments of the present invention, such as Figure 3 As shown, step S202 includes: S301. Based on the density of reconstructed Gaussian points in the reconstructed region, identify over-reconstructed sub-regions and under-reconstructed sub-regions in the reconstructed region; S302. Select a first target Gaussian point from the under-reconstructed sub-region for replication, and optimize the attributes of a second target Gaussian point in the over-reconstructed sub-region to adaptively adjust the number and attributes of the reconstructed Gaussian points; wherein, the first target Gaussian point is determined based on the attributes of each reconstructed Gaussian point in the under-reconstructed sub-region, and the second target Gaussian point is determined based on the characteristics of each reconstructed Gaussian point in the over-reconstructed sub-region.

[0042] Among them, the under-reconstructed sub-region and the over-reconstructed sub-region can be determined based on the density of the reconstructed Gaussian points in the reconstructed region and the expected density threshold. That is, the region where the density of the reconstructed Gaussian points is lower than the expected density threshold is regarded as the under-reconstructed sub-region.

[0043] The primary target Gaussian point refers to a point crucial for capturing and representing the geometric and visual features of the reconstructed region during 3D reconstruction. It can be determined based on the attributes of each reconstructed Gaussian point in the under-reconstructed sub-region, such as using reconstructed Gaussian points located on important features (e.g., edges, corners, texture changes) as the primary target Gaussian point.

[0044] The second target Gaussian point refers to the point among the reconstructed Gaussian points that best represents the overall characteristics of the over-reconstructed sub-region. These reconstructed Gaussian points are representative in terms of geometry, color, texture, and lighting, and can capture and reflect the main visual and geometric information of the over-reconstructed sub-region in which they are located. Located at the geometric center or centroid of the region, they can evenly reflect the geometric characteristics of the surrounding area. For example, they can be reconstructed Gaussian points whose color and texture match the average or dominant characteristics of the region they represent; reconstructed Gaussian points in areas with significant lighting changes or significant shadows, where representative Gaussian points should reflect these lighting effects; reconstructed Gaussian points located at edges, corners, or other significant visual features; or reconstructed Gaussian points located in areas with obvious structural features, etc.

[0045] In one specific implementation, clustering algorithms (such as K-means, DBSCAN, etc.) are used to group the reconstructed Gaussian points of the over-reconstructed sub-regions, and the center point or pattern point in each group can be used as the second target Gaussian point. Alternatively, features of the Gaussian points, such as color histograms, texture descriptors, and shape descriptors, are extracted using deep learning or traditional image processing techniques. Then, feature fusion techniques (such as feature weighting, feature selection, etc.) are used to select target features, and the second target Gaussian point is selected based on the target features.

[0046] Specifically, a first target Gaussian point is selected and copied from the under-reconstructed sub-region, thereby increasing the number of first target Gaussian points in the under-reconstructed sub-region. This improves the detail and realism of the rendering, achieving refined reconstruction processing of the under-reconstructed sub-region and optimizing its reconstruction quality. The attributes of the second target Gaussian point in the over-reconstructed sub-region are optimized, thus improving its reconstruction quality. In essence, this embodiment achieves refined reconstruction processing of both under-reconstructed and over-reconstructed sub-regions by adaptively adjusting the number and attributes of the reconstructed Gaussian points, thereby optimizing the reconstruction quality and improving the accuracy of the 3D model of the photovoltaic power station.

[0047] In some embodiments of the present invention, such as Figure 4 As shown, the method further includes: S401. Determine the rendering contribution of the reconstructed Gaussian points in the reconstructed region; S402. Delete the reconstructed Gaussian points whose rendering contribution does not meet the preset rendering conditions.

[0048] The rendering contribution refers to the visual influence of the reconstructed Gaussian points in the final rendered image. The contribution of each reconstructed Gaussian point can be evaluated using visual importance metrics (such as attention-based methods).

[0049] Specifically, Gaussian points whose rendering contribution does not meet the preset rendering conditions are deleted, that is, Gaussian points whose rendering contribution is lower than the preset contribution threshold are deleted, thereby reducing the complexity and computational cost of the reconstruction process and further improving modeling efficiency.

[0050] It is worth noting that steps S401-S402 can be performed before or after step 302, and this embodiment does not impose any restrictions.

[0051] In some embodiments of the present invention, such as Figure 5 As shown, step S102 includes: S501. Extract image features from the remote sensing images from the multiple perspectives; S502. Perform image matching based on the image features to obtain multiple sets of image pairs; S503. Perform relative camera pose estimation on multiple sets of image pairs to determine the image camera pose corresponding to each remote sensing image in the image pair, and calculate the corresponding three-dimensional point cloud based on the image camera pose to reconstruct the sparse point cloud corresponding to the photovoltaic power station.

[0052] The image features can be at least one of SIFT (Scale Invariant Feature Transform), SURF (Speed ​​Robust Feature Transform), and ORB (Orientation Fast and Rotation Invariant BRIEF). For example, SIFT feature points of a remote sensing image can be extracted using the colmap tool as image features.

[0053] Specifically, the colmap tool can be used to match image features between different remote sensing images using matching algorithms (such as the nearest neighbor method). To accelerate the matching process, the colmap tool employs approximate nearest neighbor search algorithms such as the bag-of-words model (Vocabulary Tree) or FLANN to perform image feature matching, resulting in multiple image pairs. Then, relative camera pose estimation is performed on these multiple image pairs to determine the camera pose corresponding to each remote sensing image in the image pair. Based on the camera pose, the corresponding 3D point cloud is calculated, which can be obtained by triangulation to reconstruct the sparse point cloud corresponding to the photovoltaic power station. In this embodiment, the SfM method is used, combining the camera pose of the remote sensing images to construct the point cloud, thus improving the accuracy of the sparse point cloud.

[0054] It should be noted that, in order to ensure accurate matching of image features, after identifying multiple sets of image pairs, geometric verification can be performed using tools such as colmap, RANSAC, to eliminate incorrect image pairs.

[0055] It should be further explained that, in order to improve the generation efficiency of sparse point clouds, an incremental construction method can be adopted to generate sparse point clouds. The specific process is as follows: For at least one set of image pairs, relative camera pose estimation is performed to determine the camera pose corresponding to each remote sensing image in the image pair. Based on the camera pose, the corresponding 3D point cloud is calculated, thus initializing the 3D point cloud for a set of image pairs. More images are gradually introduced using the colmap tool. For each new image introduced, the camera pose of the new image is estimated using the previously calculated 3D point cloud and matching relationships. This process typically uses the PnP (Perspective-n-Point) algorithm, which estimates the camera pose based on the known 3D point cloud and corresponding points in the image. For each new image, the colmap tool calculates new 3D points using triangulation by comparing them with existing images. Through this incremental process, the sparse 3D point cloud of the entire scene is gradually reconstructed. In some embodiments of the present invention, such as... Figure 6 As shown, step S201 includes: S601. Initialize the primitives of Gaussian sputtering based on the sparse point cloud to obtain the corresponding initial Gaussian points; S602. Determine the image frustum range of the remote sensing image corresponding to the sparse point cloud; S603. Render the initial Gaussian points of the image view frustum range using preset Gaussian point parameters to generate the initial rendered image. S604. Calculate the loss value between the initial rendered image and the preset real image, and optimize the initial rendered image based on the loss value to generate the reconstructed region.

[0056] Specifically, the `colmap` tool can be used to initialize Gaussian sputtering primitives based on sparse point clouds to obtain corresponding initial Gaussian points, essentially converting the sparse point cloud into initial Gaussian points. Then, the image frustum extent of the remote sensing image corresponding to the sparse point cloud is determined, the loss value between the initial rendered image and the preset real image is calculated, and the initial rendered image is optimized based on the loss value to generate the reconstructed region. Understandably, by optimizing the initial rendered image based on the loss value, the consistency between the rendered image and the real image is ensured, improving the visual quality and geometric accuracy of the reconstructed region.

[0057] In one specific implementation, Gaussian sputtering is used to perform 3D reconstruction of the sparse point cloud to generate a reconstructed region. The specific process is as follows: First, the primitives for Gaussian sputtering are initialized. The sparse point cloud includes position coordinates (X, Y, Z) and color information (R, G, B). The parameters of the Gaussian primitives include position coordinates (X, Y, Z), covariance matrix Σ, opacity α, and spherical harmonics. The position coordinates of the Gaussian primitives are inherited from the position coordinates of the sparse point cloud, and the spherical harmonics are obtained by transforming (R, G, B). Based on the camera pose estimated by colmap, all Gaussian points within the view frustum are rendered, compared with the real image, and the loss is calculated and optimized. The image rendering process is consistent with the point cloud image rendering process.

[0058] In some embodiments of the present invention, such as Figure 6 As shown, step S604 includes: S701. Based on the loss value, perform at least one iterative optimization on the initial rendered image, wherein each iterative optimization includes adjusting the Gaussian point parameters; S702. Render the initial Gaussian point using the adjusted Gaussian point parameters to generate the reconstructed region.

[0059] Among them, the loss can be pixel loss, such as mean squared error (MSE) or structural similarity (SSIM); geometric loss, such as depth loss or normal loss, used to ensure geometric consistency; regularization loss, such as smoothness or sparsity constraints of Gaussian points.

[0060] Specifically, the initial rendered image is optimized at least once based on the loss value. Each iteration includes adjusting the Gaussian point parameters, rendering the initial Gaussian points using the adjusted parameters, and generating the reconstructed region. This embodiment of the invention leverages the advantages of 3D Gaussian sputtering and SfM techniques to achieve the conversion from sparse point clouds to a 3D model. The optimization process ensures consistency between the rendered image and the real image, improving the visual quality and geometric accuracy of the reconstructed region.

[0061] To better implement the photovoltaic power plant modeling method in this embodiment of the invention, based on the photovoltaic power plant modeling method, correspondingly, as follows: Figure 8 As shown, this embodiment of the invention also provides a modeling device for a photovoltaic power station. The photovoltaic power station modeling device 800 includes: The acquisition unit 801 is used to acquire remote sensing images of the photovoltaic power station from multiple perspectives; The first generation unit 802 is used to generate sparse point clouds based on the remote sensing images from the multiple perspectives. The second generation unit 803 is used to perform three-dimensional reconstruction of the sparse point cloud using an improved Gaussian splashing processing method to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

[0062] The photovoltaic power plant modeling device 800 provided in the above embodiments can realize the technical solutions described in the above photovoltaic power plant modeling method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above photovoltaic power plant modeling method embodiments, and will not be repeated here.

[0063] like Figure 9 As shown, the present invention also provides an electronic device 900. The electronic device 900 includes a processor 901, a memory 902, and a display 903. Figure 9 Only some components of the electronic device 900 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0064] In some embodiments, processor 901 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 902 or process data, such as the photovoltaic power plant modeling method of the present invention.

[0065] In some embodiments, processor 901 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 901 may be local or remote. In some embodiments, processor 901 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0066] In some embodiments, memory 902 may be an internal storage unit of electronic device 900, such as a hard disk or memory of electronic device 900. In other embodiments, memory 902 may also be an external storage device of electronic device 900, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 900.

[0067] Furthermore, the memory 902 may include both internal storage units of the electronic device 900 and external storage devices. The memory 902 is used to store application software and various types of data installed on the electronic device 900.

[0068] In some embodiments, display 903 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 903 is used to display information from electronic device 900 and to display a visual user interface. Components 901-903 of electronic device 900 communicate with each other via a system bus.

[0069] In one embodiment, when processor 901 executes the photovoltaic power plant modeling program in memory 902, the following steps can be implemented: Acquire remote sensing images of photovoltaic power plants from multiple perspectives; Based on the remote sensing images from the multiple perspectives, a sparse point cloud is generated; An improved Gaussian splashing processing method is used to perform three-dimensional reconstruction of the sparse point cloud to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

[0070] It should be understood that when the processor 901 executes the photovoltaic power plant modeling program in the memory 902, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0071] Furthermore, this embodiment of the invention does not specifically limit the type of electronic device 900 mentioned. Electronic device 900 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the invention, electronic device 900 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0072] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions in the photovoltaic power plant modeling methods provided in the above-described method embodiments.

[0073] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0074] The above provides a detailed description of the photovoltaic power plant modeling method, apparatus, electronic equipment, and storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A modeling method for a photovoltaic power station, characterized in that, include: Acquire remote sensing images of photovoltaic power plants from multiple perspectives; Based on the remote sensing images from the multiple perspectives, a sparse point cloud is generated; An improved Gaussian splashing processing method is used to perform three-dimensional reconstruction of the sparse point cloud to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

2. The modeling method for photovoltaic power plants according to claim 1, characterized in that, The method of using an improved Gaussian splashing technique to perform 3D reconstruction of the sparse point cloud to generate a 3D model of the photovoltaic power station includes: The sparse point cloud is reconstructed in three dimensions using a Gaussian splashing method to generate a reconstructed region. Based on the density of the reconstructed Gaussian points in the reconstructed region, the number and properties of the reconstructed Gaussian points are adaptively adjusted to generate a three-dimensional model of the photovoltaic power station.

3. The modeling method for photovoltaic power plants according to claim 2, characterized in that, The step of adaptively adjusting the number and properties of the reconstructed Gaussian points based on the density of the reconstructed Gaussian points in the reconstructed region to generate a three-dimensional model of the photovoltaic power station includes: Based on the density of reconstructed Gaussian points in the reconstructed region, over-reconstructed sub-regions and under-reconstructed sub-regions are identified from the reconstructed region; A first target Gaussian point is selected from the under-reconstructed sub-region for replication, and the properties of the second target Gaussian point in the over-reconstructed sub-region are optimized to adaptively adjust the number and properties of the reconstructed Gaussian points. The first target Gaussian point is determined based on the attributes of each reconstructed Gaussian point in the underreconstructed sub-region, and the second target Gaussian point is determined based on the characteristics of each reconstructed Gaussian point in the overreconstructed sub-region.

4. The modeling method for photovoltaic power plants according to claim 3, characterized in that, The method further includes: Determine the rendering contribution of the reconstructed Gaussian points in the reconstructed region; The reconstructed Gaussian points whose rendering contribution does not meet the preset rendering conditions are deleted.

5. The modeling method for photovoltaic power plants according to claim 1, characterized in that, The generation of sparse point clouds based on the remote sensing images from the multiple perspectives includes: Extract image features from the remote sensing images from the multiple perspectives; Image matching is performed based on the image features to obtain multiple sets of image pairs; Relative camera pose estimation is performed on multiple sets of image pairs to determine the image camera pose corresponding to each remote sensing image in the image pair, and the corresponding three-dimensional point cloud is calculated based on the image camera pose to reconstruct the sparse point cloud corresponding to the photovoltaic power station.

6. The modeling method for photovoltaic power plants according to claim 2, characterized in that, The method of using Gaussian splashing to perform three-dimensional reconstruction of the sparse point cloud to generate the reconstructed region includes: The primitives of Gaussian sputtering are initialized based on the sparse point cloud to obtain the corresponding initial Gaussian points; Determine the image frustum range of the remote sensing image corresponding to the sparse point cloud; The initial Gaussian point of the image view frustum is rendered using preset Gaussian point parameters to generate the initial rendered image. Calculate the loss value between the initial rendered image and the preset real image, and optimize the initial rendered image based on the loss value to generate the reconstructed region.

7. The modeling method for photovoltaic power plants according to claim 6, characterized in that, The step of optimizing the initial rendered image based on the loss value to generate the reconstructed region includes: The initial rendered image is optimized at least once based on the loss value, wherein each iteration includes adjusting the Gaussian point parameters; The initial Gaussian point is rendered using the adjusted Gaussian point parameters to generate the reconstructed region.

8. A modeling device for a photovoltaic power station, characterized in that, include: The acquisition unit is used to acquire remote sensing images of the photovoltaic power station from multiple perspectives; The first generation unit is used to generate sparse point clouds based on the remote sensing images from the multiple perspectives. The second generation unit is used to perform three-dimensional reconstruction of the sparse point cloud using an improved Gaussian splashing processing method to generate a three-dimensional model of the photovoltaic power station. The improved Gaussian splashing processing method includes an adaptive Gaussian point control strategy.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the modeling method for a photovoltaic power plant according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the modeling method for a photovoltaic power plant as described in any one of claims 1 to 7.

Images