A high-precision 3D modeling method and platform for substation equipment

Abstract

The present invention discloses a high-precision three-dimensional modeling method and platform for substation equipment, which relates to the technical field of data processing. The method comprises: obtaining the equipment image acquisition results and equipment point cloud acquisition results of the substation equipment; obtaining the equipment image data stream and the equipment point cloud data stream; calculating the first Gaussian weight distribution according to the Gaussian weight analytical model; obtaining the first model of the substation equipment; obtaining the equipment feature guidance vector set; obtaining the second model of the substation equipment; performing UV expansion correction on the second model of the substation equipment according to the equipment image data stream and the equipment point cloud data stream to generate the third model of the substation equipment. The method solves the technical problems existing in the prior art of low precision in the fusion of point cloud data and image data and complex modeling process, which lead to insufficient precision and poor modeling efficiency of the three-dimensional modeling of the equipment, and achieves the technical effect of improving the precision and efficiency of the three-dimensional modeling of the substation equipment.

Description

Technical Field

[0001] The present application relates to the technical field related to data processing, and specifically to a high-precision three-dimensional modeling method and platform for substation equipment. Background Art

[0002] 3D modeling of substation equipment involves obtaining spatial geometric information of substation equipment through high-precision measurement methods and generating digital 3D models of the equipment in a computer. These 3D models not only accurately represent the equipment's geometry but also facilitate its full lifecycle management, including installation, operation, maintenance, and overhaul. Common 3D modeling methods include LiDAR (Light Detection and Ranging) and photogrammetry. LiDAR technology generates high-precision 3D point cloud data by emitting laser light and measuring the time difference between its reflections. Camera imaging technology provides rich visual information by capturing high-definition images of the equipment. However, in practical applications, efficiently and accurately combining LiDAR point cloud data with camera image data to generate high-precision 3D models of substation equipment still faces many challenges, including efficient filtering and smoothing of the point cloud data to reduce noise and errors, and effective registration of the image data with the point cloud data to enhance the detailed texture of the equipment model.

[0003] Therefore, in the current related technologies, there are technical problems such as low accuracy in the fusion of point cloud data and image data and complex modeling process, which in turn lead to insufficient accuracy and poor modeling efficiency in the three-dimensional modeling of the equipment. Summary of the Invention

[0004] This application provides a high-precision three-dimensional modeling method and platform for substation equipment, which solves the technical problems in the existing technology such as low accuracy of point cloud data and image data fusion and complex modeling process, which lead to insufficient accuracy and poor modeling efficiency of equipment three-dimensional modeling, and achieves the technical effect of improving the three-dimensional modeling accuracy and modeling efficiency of substation equipment.

[0005] The present application provides a high-precision three-dimensional modeling method for substation equipment, the method comprising: obtaining device image acquisition results and device point cloud acquisition results of the substation equipment based on a camera and a lidar device; inputting the device image acquisition results into an image filter to obtain a device image data stream, and inputting the device point cloud acquisition results into a point cloud filter to obtain a device point cloud data stream; based on the device point cloud data stream, calculating a first Gaussian weight distribution according to a Gaussian weight parsing model; performing smooth modeling on the device point cloud data stream according to the first Gaussian weight distribution to obtain a first model of the substation equipment; performing SIFT feature registration on the device image data stream and the device point cloud data stream to obtain a device feature guide vector set; based on the device feature guide vector set, guiding the device image data stream to perform detail texture enhancement on the first model of the substation equipment to obtain a second model of the substation equipment; performing UV expansion correction on the second model of the substation equipment according to the device image data stream and the device point cloud data stream to generate a third model of the substation equipment.

[0006] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment also performs the following processing: dividing the equipment point cloud data stream into neighborhoods of each point according to a predetermined neighborhood radius to obtain the neighborhoods of each point of the equipment; calculating the neighborhood centroid of each point according to the neighborhoods of each point of the equipment to obtain the neighborhood centroid of each point; calculating the covariance matrix of each point according to the neighborhoods of each point of the equipment to obtain the covariance matrix of each point; inputting the neighborhood centroid of each point and the covariance matrix of each point into the Gaussian weight analytical model to output multi-point Gaussian weights; and outputting the multi-point Gaussian weights as the first Gaussian weight distribution.

[0007] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment further performs the following processing: the Gaussian weight analytical model includes a Gaussian weight analytical function, and the Gaussian weight analytical function is:

[0008] ;

[0009] in, Represents the Gaussian weight of point i, i represents point i in the device point cloud data stream, point i is any point in the device point cloud data stream, Represents the i-point covariance matrix, Characterize the determinant of the covariance matrix of point i, Characterizes the inverse matrix of the i-point covariance matrix, Represents an exponential function with the natural constant e as the base, Represents the centroid of the neighborhood of point i, Characterizes the deviation vector between point i and the centroid of its neighborhood, The transposed vector representing the deviation vector between point i and the centroid of its neighborhood.

[0010] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment also performs the following processing: performing normalization calculation based on the first Gaussian weight distribution to obtain a second Gaussian weight distribution; performing Gaussian weighted smoothing optimization on the equipment point cloud data stream based on the second Gaussian weight distribution to obtain a smooth point cloud set of equipment; performing three-dimensional modeling based on the equipment smooth point cloud set to obtain the first model of the substation equipment.

[0011] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment also performs the following processing: performing SIFT feature detection based on the equipment image data stream to obtain a device image feature vector set; performing SIFT feature detection based on the equipment point cloud data stream to obtain a device point cloud feature vector set; performing matching cost identification on the equipment point cloud feature vector set based on each equipment image feature vector in the equipment image feature vector set to obtain multiple matching cost coefficient sets; performing matching cost minimization screening based on the multiple matching cost coefficient sets to determine the optimal matching cost distribution; based on the optimal matching cost distribution, aligning and associating the equipment image feature vector set and the equipment point cloud feature vector set to generate the equipment feature guidance vector set.

[0012] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment also performs the following processing: based on the equipment feature guidance vector set, the equipment image data stream is projected onto the first model of the substation equipment to obtain an image projection point set; the detail texture of the first model of the substation equipment is optimized according to the image projection point set to obtain an optimized model of the substation equipment; a detail texture transition smoothing evaluation is performed according to the optimized model of the substation equipment to obtain a model transition smoothing coefficient; it is determined whether the model transition smoothing coefficient is less than a predetermined transition smoothing coefficient; if the model transition smoothing coefficient is less than the predetermined transition smoothing coefficient, the substation equipment optimization model is transition smoothed according to the predetermined transition smoothing coefficient to generate the second model of the substation equipment.

[0013] In a possible implementation, the high-precision three-dimensional modeling method for substation equipment also performs the following processing: UV unfolding is performed based on the second model of the substation equipment to obtain a device model unfolding diagram; a device detail texture map is constructed based on the device image data stream; a device geometric feature map is drawn based on the device point cloud data stream; deviation detection is performed on the device model unfolding diagram based on the device detail texture map and the device geometric feature map to obtain a device model deviation detection result; based on the device model deviation detection result, the second model of the substation equipment is corrected to obtain the third model of the substation equipment.

[0014] The present application also provides a high-precision three-dimensional modeling platform for substation equipment, including: an acquisition result acquisition module for obtaining device image acquisition results and device point cloud acquisition results of substation equipment based on cameras and lidar equipment; a data stream acquisition module for inputting the device image acquisition results into an image filter to obtain a device image data stream, and inputting the device point cloud acquisition results into a point cloud filter to obtain a device point cloud data stream; a first Gaussian weight distribution calculation module for calculating a first Gaussian weight distribution based on the device point cloud data stream according to a Gaussian weight analytical model; a substation equipment first model acquisition module for analysing the device point cloud data stream according to the first Gaussian weight distribution. A smooth modeling is performed on the prepared point cloud data stream to obtain a first model of the substation equipment; a device feature guide vector set acquisition module is used to perform SIFT feature alignment according to the device image data stream and the device point cloud data stream to obtain a device feature guide vector set; a substation equipment second model acquisition module is used to guide the device image data stream to perform detail texture enhancement on the first model of the substation equipment based on the device feature guide vector set to obtain a second model of the substation equipment; a substation equipment third model generation module is used to perform UV expansion correction on the second model of the substation equipment according to the device image data stream and the device point cloud data stream to generate a third model of the substation equipment.

[0015] This application proposes a high-precision 3D modeling method and platform for substation equipment. The method and platform aim to obtain device image and point cloud acquisition results for the substation equipment; obtain device image data streams and device point cloud data streams; calculate a first Gaussian weight distribution based on a Gaussian weight analytic model; obtain a first model of the substation equipment; obtain a set of device feature guidance vectors; obtain a second model of the substation equipment; and generate a third model of the substation equipment. This method addresses the existing technical issues of low point cloud data and image data fusion accuracy and a complex modeling process, which in turn lead to insufficient accuracy and poor modeling efficiency in 3D equipment modeling. The method achieves the technical effect of improving the accuracy and efficiency of 3D modeling of substation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the platform according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. On the contrary, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0017] Figure 1 A flowchart of a high-precision three-dimensional modeling method for substation equipment provided in an embodiment of the present application.

[0018] Figure 2 A schematic structural diagram of a high-precision three-dimensional modeling platform for substation equipment provided in an embodiment of the present application.

[0019] Explanation of the accompanying drawings: collection result acquisition module 10, data stream acquisition module 20, first Gaussian weight distribution calculation module 30, substation equipment first model acquisition module 40, equipment feature guidance vector set acquisition module 50, substation equipment second model acquisition module 60, substation equipment third model generation module 70. DETAILED DESCRIPTION

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0021] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0022] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms “first\second” involved are merely to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions, for example, a process, method, platform, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0023] The present application embodiment provides a high-precision three-dimensional modeling method for substation equipment, such as Figure 1 As shown, the method includes:

[0024] Step S100: Obtain device image acquisition results and device point cloud acquisition results of the substation equipment based on the camera and lidar equipment.

[0025] Preferably, image data of the substation equipment is acquired through a camera device (such as a high-definition camera, a binocular high-definition camera, etc.). This is used as the device image acquisition result of the substation equipment, and typically includes information such as the appearance, surface features, color, and texture of the equipment. The camera can capture two-dimensional image data of the equipment based on different shooting angles, focal lengths, and lighting conditions, which is used for subsequent processing such as detail texture enhancement, image filtering, and feature matching. Three-dimensional point cloud data of the substation equipment is acquired through a laser radar (LiDAR) device to serve as the device point cloud acquisition result of the substation equipment. Specifically, the LiDAR emits a laser beam and accurately measures the distance to each point based on the time it takes for the laser to reflect back from the surface of an object, thereby constructing a three-dimensional dataset consisting of a large number of spatial points. The point cloud data represents the three-dimensional geometric form of the equipment, such as its position, shape, and surface contour. The point cloud data acquired through the LiDAR is typically highly accurate and can accurately represent the three-dimensional structure of the equipment. The device image provides rich visual information, which is helpful for subsequent texture enhancement and detail processing, while the point cloud data acquired by the LiDAR provides accurate spatial geometric structure information for the model. By fusing these two types of data, a more accurate and detailed three-dimensional model of the substation equipment can be generated.

[0026] Step S200: input the device image acquisition result into an image filter to obtain a device image data stream, and input the device point cloud acquisition result into a point cloud filter to obtain a device point cloud data stream.

[0027] Preferably, the device image acquisition result is input into an image filter for processing, that is, the image filter is used to perform denoising, contrast enhancement, brightness adjustment, smoothing, edge sharpening, etc. on the image data to improve image quality and remove noise. After the device image acquisition result is processed by the image filter, an optimized, denoised, and smoothed image data stream is output, that is, a device image data stream is obtained, which can be smoothly transmitted to subsequent processing modules (such as feature extraction, texture enhancement, etc.); the device point cloud acquisition result is input into a point cloud filter for processing. The point cloud filter is mainly used to remove noise points, fill data gaps, reduce outliers, etc., so that the point cloud data is more accurate and complete. The point cloud filter can also perform other operations, such as sparseness (reducing point cloud density) or resampling. Specifically, after the device point cloud acquisition result is processed by the point cloud filter, an optimized and denoised point cloud data stream is output, that is, a device point cloud data stream is obtained. The device point cloud data stream refers to the processed point cloud data. After filtering, these data can continue to be transmitted to subsequent processing stages (such as modeling, feature extraction, etc.) and ensure the accuracy of the generated three-dimensional model of the substation equipment.

[0028] Step S300: Calculate a first Gaussian weight distribution based on the device point cloud data stream and a Gaussian weight parsing model.

[0029] Preferably, the equipment point cloud data stream contains a three-dimensional point cloud data stream of the spatial information of the substation equipment, and the position of each point in space is represented by its coordinates (such as X, Y, Z), and may also contain other information, such as intensity value, color information, etc., which are used to represent the geometric shape and spatial structure of the equipment. The first Gaussian weight distribution is calculated using the equipment point cloud data and the Gaussian weight parsing model, wherein the Gaussian weight parsing model is a mathematical model used to assign a weight to each point in the point cloud data based on the Gaussian distribution, representing the importance of the point in space or the similarity with the surrounding points. The Gaussian distribution is used to describe the distribution of data points under certain conditions. The Gaussian weight usually depends on the distance between the points in space and other related factors, such as the spatial distance between points, the density of points, and and the local shape of the points. By applying the Gaussian weight parsing model to the point cloud data, a weighted value can be calculated for each point in the point cloud. Specifically, each point cloud data point is calculated with a weight value according to its spatial position (such as the distance from neighboring points) and other features (such as the density of the points) (usually the center point has a large weight, and the points far from the center point have a small weight), and is optimized according to the geometric shape and structure of the equipment. For example, if some points are close to the surface of an object or a key feature area, they are given a larger weight. Then, a mathematical formula (Gaussian function) is used to generate the first Gaussian weight distribution to help eliminate noise and better reflect the true shape of the object surface, thereby making the point cloud data more accurate. The model generated after smoothing modeling can better reflect the actual shape and details of the substation equipment.

[0030] Furthermore, step S300 also includes step S310, dividing the neighborhood of each point of the device point cloud data stream according to a predetermined neighborhood radius to obtain the neighborhood of each point of the device; step S320, calculating the neighborhood centroid according to the neighborhood of each point of the device to obtain the neighborhood centroid of each point; step S330, calculating the covariance matrix according to the neighborhood of each point of the device to obtain the covariance matrix of each point; step S340, inputting the neighborhood centroid of each point and the covariance matrix of each point into the Gaussian weight analytical model to output multi-point Gaussian weights; step S350, outputting the multi-point Gaussian weights as the first Gaussian weight distribution.

[0031] Preferably, the neighborhood radius is a predefined parameter, which represents a set of points within a range around each point. In point cloud processing, a radius range is usually selected to determine the neighborhood of each point. The neighborhood radius can be adjusted based on factors such as the scale of the device and the resolution of the point cloud. Specifically, the point cloud data is divided into neighborhoods according to the predetermined neighborhood radius, that is, all other points within the radius around each point in the point cloud data are identified to form the neighborhood of the point (that is, the spatial neighbors of the point constitute the local area of ​​the point); then the neighborhood centroid of each point in the device is calculated, that is, the coordinates of all points in the neighborhood are averaged, and the geometric center (center of mass) of the neighborhood is calculated, thereby obtaining the neighborhood centroid of each point, which represents the spatial center of the neighborhood.

[0032] Preferably, a covariance matrix is ​​calculated for the neighborhood of each point in the device, wherein the covariance matrix is ​​a mathematical tool that describes the spatial relationship between data points, and is used to represent the spatial distribution and directionality of all points in a neighborhood in the point cloud. Specifically, for any point in the point cloud data and all points in its neighborhood, the difference between each point and the center of mass is calculated based on the neighborhood center of mass of the point, that is, the deviation of each point from the center of mass on each coordinate axis is calculated, and these differences are used to calculate the elements of the covariance matrix, and finally a covariance matrix is ​​obtained, including variance (diagonal elements) and covariance (non-diagonal elements), wherein the variance represents the extension degree or distribution width of the neighborhood points in the x, y, and z directions. If the variance in a certain direction is large, it means that the points are more dispersed in this direction; if the variance is small, it means that the points are more concentrated in this direction; the covariance represents the correlation between different directions. If the covariance is large, it means that the points tend to be consistent in these two directions; if the covariance is negative, it means that they show opposite trends in these two directions.

[0033] Preferably, the centroid and covariance matrix of each point are used as inputs to a Gaussian weighted analytical model. The Gaussian weighted analytical model calculates the Gaussian weight of the point based on the neighborhood centroid and covariance matrix using a Gaussian distribution function. This weight is calculated by considering the geometric center and spatial distribution of the neighborhood points. These weights represent the importance or reliability of the point. Gaussian weights typically assign higher weights to points close to the centroid and lower weights to points farther away. The Gaussian weights of all points are then integrated to form a Gaussian weight distribution, or the first Gaussian weight distribution. The Gaussian weight distribution is a weighted representation of the entire point cloud data, where the weight of each point reflects the importance of the point within the local area. In this way, more important points in the point cloud data (such as device edges and key information points) receive greater weights, while irrelevant or noisy points receive smaller weights. By applying Gaussian weights, the model can better reflect the geometry and structure of the device, reduce noise and errors, and thus improve the accuracy of 3D modeling.

[0034] Furthermore, step S300 further includes: the Gaussian weight analytical model includes a Gaussian weight analytical function, and the Gaussian weight analytical function is:

[0035] ;

[0036] in, Represents the Gaussian weight of point i, i represents point i in the device point cloud data stream, point i is any point in the device point cloud data stream, Represents the i-point covariance matrix, Characterize the determinant of the covariance matrix of point i, Characterizes the inverse matrix of the i-point covariance matrix, Represents an exponential function with the natural constant e as the base, Represents the centroid of the neighborhood of point i, Characterizes the deviation vector between point i and the centroid of its neighborhood, The transposed vector representing the deviation vector between point i and the centroid of its neighborhood.

[0037] Preferably, the Gaussian weighted analytical function , used in point cloud data processing, a mathematical function that calculates the weight of each point through Gaussian distribution, that is, assigning corresponding weights to points according to the relationship between the point and its neighborhood, where the covariance matrix Describes the spatial distribution of the point cloud in the neighborhood, which not only represents the variance of the points on each coordinate axis, but also includes the correlation between the points (i.e., the covariance of the points); the determinant of the covariance matrix is It provides the "volume" information of the spatial distribution of neighborhood points. A larger determinant indicates that the points are more widely distributed, while a smaller determinant indicates that the points are more concentrated. The inverse matrix of the covariance matrix is It is used to adjust the relative position of each point in the point cloud data in space, so that points far away from the center of mass in the point cloud are given lower weights; Used to define the attenuation characteristics of the Gaussian distribution. The shape of the Gaussian distribution is bell-shaped, with points close to the center of mass having higher weights and points far from the center of mass having lower weights; the transpose vector of the deviation vector , the transpose operation allows the bias vector to be matrix multiplied with the inverse of the covariance matrix.

[0038] Step S400: Smooth modeling is performed on the equipment point cloud data stream according to the first Gaussian weight distribution to obtain a first model of the substation equipment.

[0039] Step S400 further includes step S410, performing normalization calculation according to the first Gaussian weight distribution to obtain a second Gaussian weight distribution; step S420, performing Gaussian weighted smoothing optimization on the device point cloud data stream according to the second Gaussian weight distribution to obtain a device smooth point cloud set; step S430, performing three-dimensional modeling according to the device smooth point cloud set to obtain the first model of the substation equipment.

[0040] Preferably, the device point cloud data stream is smoothed and modeled according to the first Gaussian weight distribution (the weight distribution of each point calculated by the Gaussian weight analytical model), that is, the device point cloud data is processed by applying a smoothing algorithm to reduce the noise and error in the point cloud, so that the device model is smoother and more continuous. It usually involves weighted smoothing of each point in the point cloud so that the coordinates of each point are affected by the coordinates of other points in the neighborhood during calculation, thereby reducing the abnormal influence of a single point. Specifically, a normalization calculation is performed according to the first Gaussian weight distribution, that is, the first Gaussian weight distribution is adjusted to a standard range (usually between 0 and 1), and a second Gaussian weight distribution is obtained to ensure that the weights of all points add up to 1, avoiding certain Some points may cause deviations in the modeling results due to excessive or insufficient weights, so that the weights of all points in the point cloud become balanced; the second Gaussian weight distribution is then used to perform weighted smoothing on the device point cloud data, that is, the weight of each point is adjusted according to the points and covariance matrix in its neighborhood, further optimizing the smoothing effect of the point cloud data, thereby reducing the noise and error of the point cloud data and generating a smoothed point cloud set of the device; finally, three-dimensional modeling is performed based on the smoothed point cloud set of the device, that is, by processing the smoothed point cloud set of the device, an accurate three-dimensional model is generated as the first model of the substation equipment, which can truly reproduce the shape and details of the substation equipment, represent the geometric shape of the substation equipment, ensure the accuracy of the model, and at the same time improve the authenticity and visual effect of the model.

[0041] Step S500 : performing SIFT feature registration based on the device image data stream and the device point cloud data stream to obtain a device feature guidance vector set.

[0042] Preferably, SIFT feature registration (scale-invariant feature transform) is a commonly used feature matching method in computer vision, which is mainly used to extract and match key feature points in images. The purpose is to identify and match local features in images. Even when the scale, rotation, perspective and illumination change, the features can still remain consistent. SIFT feature registration is performed based on the device image data stream and the device point cloud data stream. Specifically, SIFT is used to extract features from the device image data stream, including detecting key points with high contrast from the image, usually at the edge of the image or in areas with more complex textures, and generating a multidimensional vector for each detected key point, representing the local image structure around the key point; feature extraction is performed on the device point cloud data stream, including determining key points by detecting sparse or significantly changed areas, such as corners, edges or other areas with relatively large geometric changes in the device. Large areas, as well as local geometric shapes based on point clouds, use some methods (such as normal direction, curvature, etc.) to generate descriptors for each key point to represent the geometric features of each key point in the point cloud; SIFT feature registration is to align the feature points extracted from the device image data stream and the device point cloud data stream, that is, to find the matching relationship between the image features and the point cloud features, that is, to use the nearest neighbor matching method to calculate the similarity or distance between the feature points in the two types of data streams, match the key points in the device image and the key points in the device point cloud, find the correspondence between the two, and then perform geometric transformations on the matching points (such as rotation, translation, etc.), align the key points in the image and the key points in the point cloud, and obtain matching point pairs between the image and the point cloud, and then generate a device feature guidance vector set, which contains the correspondence between the image features and the point cloud features, usually expressed as a vector or feature pair.

[0043] Furthermore, step S500 also includes step S510, performing SIFT feature detection based on the device image data stream to obtain a device image feature vector set; including step S520, performing SIFT feature detection based on the device point cloud data stream to obtain a device point cloud feature vector set; including step S530, performing matching cost identification on the device point cloud feature vector set based on each device image feature vector in the device image feature vector set to obtain multiple matching cost coefficient sets; including step S540, performing matching cost minimization screening based on the multiple matching cost coefficient sets to determine the optimal matching cost distribution; including step S550, performing registration and association on the device image feature vector set and the device point cloud feature vector set based on the optimal matching cost distribution to generate the device feature guidance vector set.

[0044] Preferably, SIFT feature detection is performed on the device image data stream to extract feature points in each device image (usually located at the edges, corners or areas with obvious texture in the image). Each feature point will have a corresponding descriptor (feature vector) that captures the geometric or texture information of the local area around the point. All feature vectors constitute the device image feature vector set; SIFT feature detection is performed on the device point cloud data stream, that is, by analyzing the key points and local geometric shapes in the point cloud data, key points are extracted from the device point cloud data stream, and a feature vector is generated for each point, reflecting the local geometric shape and spatial position of each point. All feature vectors constitute the device point cloud feature vector set.

[0045] Preferably, each image feature in the device image feature vector set is matched with the feature in the device point cloud feature vector set, that is, the distance between the two (such as Euclidean distance) is calculated to measure their similarity. The smaller the matching cost, the more likely the two feature points are corresponding points at the same position. Each device image feature vector is matched with multiple feature vectors in the device point cloud feature vector set, and the matching cost (i.e. similarity) is calculated, thereby obtaining multiple matching cost coefficient sets, which represent the degree of matching between the image features and the point cloud features. Then, a minimization strategy is used to analyze all matching costs of the multiple matching cost coefficient sets, and screen The matching pairs with the smallest matching cost are selected, that is, those feature pairs with the smallest distance or the strongest similarity are selected, and which image features and point cloud features are the best matches are determined, thereby forming the optimal matching cost distribution; finally, according to the optimal matching cost distribution, the features in the device image feature vector set are aligned with the features in the device point cloud feature vector set, that is, the two-dimensional image features and the three-dimensional point cloud features are aligned to ensure that they correspond to the same position or the same physical structure, and then the device feature guidance vector set is obtained, which can effectively associate local features in the image and point cloud data, and provide accurate input for subsequent processing such as three-dimensional modeling and texture mapping.

[0046] Step S600: Based on the equipment feature guidance vector set, guide the equipment image data stream to perform detail texture enhancement on the first model of the substation equipment to obtain a second model of the substation equipment.

[0047] Preferably, detail texture enhancement refers to refining and optimizing the texture of an existing three-dimensional point cloud model through image data, wherein the initially generated first model of the substation equipment may have some problems such as unclear texture, insufficient details or low surface quality. Specifically, by using the device feature guidance vector set, the texture information (such as color, texture, surface details, etc.) in the device image data stream is mapped to the first model of the substation equipment. This can not only improve the presentation of surface details, but also ensure accurate spatial alignment of the image and the three-dimensional model according to the feature vector set, so that the texture fits the geometric shape of the equipment more closely. That is, through texture mapping technology, the details in the image data (such as the color, brightness, reflection characteristics, etc. of the object surface) are transferred to the surface of the three-dimensional model, thereby improving the realism and fineness of the model, and then generating a second model of the substation equipment with richer surface texture and higher visual effect, which can more realistically reproduce the appearance and details of the substation equipment, thereby improving the quality and effect of modeling.

[0048] Furthermore, step S600 also includes step S610, projecting the device image data stream to the first model of the substation equipment based on the device feature guidance vector set to obtain an image projection point set; step S620, performing detail texture optimization on the first model of the substation equipment according to the image projection point set to obtain a substation equipment optimization model; step S630, performing detail texture transition smoothing evaluation based on the substation equipment optimization model to obtain a model transition smoothing coefficient; step S640, judging whether the model transition smoothing coefficient is less than a predetermined transition smoothing coefficient; step S650, if the model transition smoothing coefficient is less than the predetermined transition smoothing coefficient, performing transition smoothing optimization on the substation equipment optimization model according to the predetermined transition smoothing coefficient to generate the second model of the substation equipment.

[0049] Preferably, the texture and image information in the device image data stream is mapped to the three-dimensional model of the substation equipment. Specifically, the image data is "projected" through a feature-guided vector set, that is, the pixel position of the image is corresponded to the surface of the three-dimensional model, and each pixel in the image corresponds to a point on the surface of the three-dimensional model, forming an image projection point set, which contains all texture mapping positions; the image projection point set is then used to optimize the texture of the first model of the substation equipment, that is, to determine which parts of the device model need to be enhanced in detail, such as texture clarity, contrast, color, etc., so that the surface details of the model are more realistic, thereby making the model more visually accurate and delicate, and obtaining an optimized model of the substation equipment.

[0050] Preferably, the substation equipment optimization model is evaluated for detail texture transition smoothness. Specifically, the texture of the model after texture optimization may have unnatural texture changes in some areas, especially the texture transition between different areas may be uneven. The model is subjected to transition smoothing processing to make the transition between different texture areas more natural and smooth, and a coefficient measuring the texture transition smoothness (model transition smoothing coefficient) is calculated. If the transition smoothing coefficient is high, it means that the texture transition is more natural. If the coefficient is low, there may be obvious transition uneven areas. Then, it is determined whether the model transition smoothing coefficient is less than a predetermined transition smoothing coefficient, wherein the predetermined transition smoothing coefficient is a set transition smoothing standard value. If the model transition smoothing coefficient is greater than or equal to the predetermined transition smoothing coefficient, the model does not need to be further optimized. On the contrary, if the transition smoothing coefficient is less than the predetermined transition smoothing coefficient, the substation equipment optimization model is subjected to transition smoothing optimization using the predetermined transition smoothing coefficient, that is, the model is adjusted by interpolation, gradient adjustment, etc., so that the texture transition between different areas is smoother, and the final second substation equipment model is generated, which can more realistically represent the appearance and details of the substation equipment and has higher quality and credibility in practical applications (such as simulation, analysis, display, etc.).

[0051] Step S700 : Perform UV unwrapping correction on the second model of the substation equipment according to the equipment image data stream and the equipment point cloud data stream to generate a third model of the substation equipment.

[0052] Preferably, UV unfolding correction unfolds the surface of a three-dimensional object into a two-dimensional plane and maps the texture onto it, thereby ensuring that the texture of the three-dimensional model surface can naturally fit the geometric shape of the model. Among them, UV unfolding is to map the surface of a three-dimensional object to a two-dimensional plane. Each vertex of the three-dimensional model has a corresponding two-dimensional coordinate (U, V), which represents the position of each point on the surface of the object in the two-dimensional image. Through UV unfolding, the surface of the three-dimensional model can be "flattened" so that it can be mapped (such as texture mapping) on ​​it, but it may cause the texture to be stretched, distorted or unnaturally deformed. The goal of UV unfolding correction is to correct this irregular texture mapping so that the texture More evenly and naturally covered on the model surface, specifically, the three-dimensional surface of the substation equipment is "unfolded" into a two-dimensional plane, and a UV coordinate system is assigned to each face or local area. By adjusting the UV coordinates, the texture can be correctly and evenly mapped to the three-dimensional model surface to avoid obvious stretching or distortion. If there is texture stretching or distortion in some areas of the model, it is corrected by adjusting the UV coordinates to ensure a smooth transition and consistency of the texture map, ensuring the accuracy and authenticity of the texture, and then obtaining the final third model of the substation equipment with high-quality surface texture mapping, which can better show the details and appearance of the substation equipment.

[0053] Furthermore, step S700 also includes step S710, performing UV unfolding according to the second model of the substation equipment to obtain a device model unfolding diagram; step S720, constructing a device detail texture map according to the device image data stream; step S730, drawing a device geometric feature map according to the device point cloud data stream; step S740, performing deviation detection on the device model unfolding diagram according to the device detail texture map and the device geometric feature map to obtain a device model deviation detection result; step S750, correcting the second model of the substation equipment according to the device model deviation detection result to obtain the third model of the substation equipment.

[0054] Preferably, the second model of the substation equipment is UV unfolded, that is, the surface of the three-dimensional model is flattened to a two-dimensional plane (expanded view) to provide an accurate coordinate system for texture mapping, ensuring that each surface area in the two-dimensional plane can correctly represent its position in three-dimensional space, and obtaining the equipment model expansion view, which contains the UV coordinate mapping information of all surface areas; based on the equipment image data stream, the texture is extracted, optimized and synthesized to obtain the equipment detail texture map, which contains the detailed information of the equipment surface, such as material, color, gloss, etc.; the geometric features of the equipment are extracted according to the equipment point cloud data stream, such as the contour, angle, curvature and other information of the equipment surface, to form the equipment geometric feature map; then the equipment model expansion map is subjected to deviation detection, that is, the texture mapping and geometric features on the equipment model expansion map are compared to see whether they match the equipment detail texture map and geometric feature map, so as to detect Check whether the surface texture of the equipment model is correctly fitted to the geometric form of the equipment, and then obtain the equipment model deviation detection result, which reflects the difference or mismatch between the surface texture and the geometric structure of the equipment during the texture mapping process. For example, the texture may be stretched, distorted or misplaced in certain areas. The purpose of deviation detection is to locate and quantify these differences to ensure the accuracy of the final model; finally, based on the deviation detection results, the second model of the substation equipment is corrected, including adjusting UV unfolding, realigning textures, repairing geometric inconsistencies, etc., to ensure that the texture and geometric form can be seamlessly connected, and the third model of the substation equipment is obtained, which has significant improvements in visual effects and geometric accuracy, can ensure that the texture and geometric structure are fully matched, there is no texture distortion or geometric inconsistency, and more realistically reflects the actual appearance and details of the substation equipment.

[0055] In the above, refer to Figure 1 A high-precision three-dimensional modeling method for substation equipment according to an embodiment of the present invention is described in detail. Figure 2 A high-precision three-dimensional modeling platform for substation equipment according to an embodiment of the present invention is described.

[0056] According to an embodiment of the present invention, a high-precision 3D modeling platform for substation equipment is used to solve the technical problems existing in the prior art, such as low precision in the fusion of point cloud data and image data, complex modeling process, and thus insufficient precision and poor modeling efficiency in equipment 3D modeling, thereby achieving the technical effect of improving the precision and efficiency of 3D modeling of substation equipment. Figure 2 As shown, a high-precision three-dimensional modeling platform for substation equipment includes: an acquisition result acquisition module 10, a data stream acquisition module 20, a first Gaussian weight distribution calculation module 30, a substation equipment first model acquisition module 40, an equipment feature guidance vector set acquisition module 50, a substation equipment second model acquisition module 60, and a substation equipment third model generation module 70.

[0057] The acquisition result acquisition module 10 is used to obtain the device image acquisition results and device point cloud acquisition results of the substation equipment based on the camera and the lidar equipment; the data stream acquisition module 20 is used to input the device image acquisition results into the image filter to obtain the device image data stream, and input the device point cloud acquisition results into the point cloud filter to obtain the device point cloud data stream; the first Gaussian weight distribution calculation module 30 is used to calculate the first Gaussian weight distribution based on the device point cloud data stream according to the Gaussian weight analytical model; the substation equipment first model acquisition module 40 is used to smooth the device point cloud data stream according to the first Gaussian weight distribution. Obtain a first model of substation equipment; a device feature guide vector set acquisition module 50, used to perform SIFT feature alignment based on the device image data stream and the device point cloud data stream to obtain a device feature guide vector set; a second model acquisition module 60 of substation equipment, used to guide the device image data stream to perform detail texture enhancement on the first model of substation equipment based on the device feature guide vector set to obtain a second model of substation equipment; a third model generation module 70 of substation equipment, used to perform UV expansion correction on the second model of substation equipment based on the device image data stream and the device point cloud data stream to generate a third model of substation equipment.

[0058] The specific configuration of the first Gaussian weight distribution calculation module 30 will be described in detail below. The first Gaussian weight distribution calculation module 30 further includes: dividing the device point cloud data stream into neighborhoods according to a predetermined neighborhood radius to obtain the neighborhoods of each device point; calculating the neighborhood centroid based on the neighborhoods of each device point to obtain the neighborhood centroid of each point; calculating the covariance matrix based on the neighborhoods of each device point to obtain the covariance matrix of each point; inputting the neighborhood centroid of each point and the covariance matrix of each point into the Gaussian weight analytical model to output multi-point Gaussian weights; and outputting the multi-point Gaussian weights as the first Gaussian weight distribution.

[0059] The specific configuration of the first Gaussian weight distribution calculation module 30 will be described in detail below. The first Gaussian weight distribution calculation module 30 further includes: the Gaussian weight analytical model includes a Gaussian weight analytical function, and the Gaussian weight analytical function is:

[0060] ;

[0061] in, Represents the Gaussian weight of point i, i represents point i in the device point cloud data stream, point i is any point in the device point cloud data stream, Represents the i-point covariance matrix, Characterize the determinant of the covariance matrix of point i, Characterizes the inverse matrix of the i-point covariance matrix, Represents an exponential function with the natural constant e as the base, Represents the centroid of the neighborhood of point i, Characterizes the deviation vector between point i and the centroid of its neighborhood, The transposed vector representing the deviation vector between point i and the centroid of its neighborhood.

[0062] The specific configuration of the substation equipment first model acquisition module 40 will be described in detail below. The substation equipment first model acquisition module 40 further includes: performing a normalization calculation based on the first Gaussian weight distribution to obtain a second Gaussian weight distribution; performing Gaussian weighted smoothing optimization on the equipment point cloud data stream based on the second Gaussian weight distribution to obtain a smoothed equipment point cloud set; and performing three-dimensional modeling based on the smoothed equipment point cloud set to obtain the first substation equipment model.

[0063] The specific configuration of the device feature guidance vector set acquisition module 50 will be described in detail below. The device feature guidance vector set acquisition module 50 further includes: performing SIFT feature detection based on the device image data stream to obtain a device image feature vector set; performing SIFT feature detection based on the device point cloud data stream to obtain a device point cloud feature vector set; performing matching cost identification on the device point cloud feature vector set based on each device image feature vector in the device image feature vector set to obtain multiple matching cost coefficient sets; performing matching cost minimization screening based on the multiple matching cost coefficient sets to determine the optimal matching cost distribution; and performing registration association on the device image feature vector set and the device point cloud feature vector set based on the optimal matching cost distribution to generate the device feature guidance vector set.

[0064] The specific configuration of the substation equipment second model acquisition module 60 will be described in detail below. The substation equipment second model acquisition module 60 further includes: projecting the equipment image data stream onto the substation equipment first model based on the equipment feature guidance vector set to obtain an image projection point set; performing detail texture optimization on the substation equipment first model according to the image projection point set to obtain a substation equipment optimization model; performing detail texture transition smoothing evaluation on the substation equipment optimization model to obtain a model transition smoothing coefficient; determining whether the model transition smoothing coefficient is less than a predetermined transition smoothing coefficient; if the model transition smoothing coefficient is less than the predetermined transition smoothing coefficient, performing transition smoothing optimization on the substation equipment optimization model according to the predetermined transition smoothing coefficient to generate the substation equipment second model.

[0065] The specific configuration of the substation equipment third model generation module 70 will be described in detail below. The substation equipment third model generation module 70 further includes: performing UV unfolding based on the second substation equipment model to obtain a device model unfolded diagram; constructing a device detail texture map based on the device image data stream; drawing a device geometric feature map based on the device point cloud data stream; performing deviation detection on the device model unfolded diagram based on the device detail texture map and the device geometric feature map to obtain a device model deviation detection result; and correcting the second substation equipment model based on the device model deviation detection result to obtain the third substation equipment model.

[0066] A high-precision three-dimensional modeling platform for substation equipment provided by an embodiment of the present invention can execute a high-precision three-dimensional modeling method for substation equipment provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0067] Although this application makes various references to certain modules in the platform according to the embodiments of this application, any number of different modules can be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0068] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A high-precision three-dimensional modeling method for substation equipment, characterized in that: The method comprises: Obtain device image acquisition results and device point cloud acquisition results for substation equipment using cameras and lidar equipment; Input the device image acquisition result into an image filter to obtain a device image data stream, and input the device point cloud acquisition result into a point cloud filter to obtain a device point cloud data stream; Calculating a first Gaussian weight distribution based on the device point cloud data stream and a Gaussian weight parsing model; Performing smooth modeling on the device point cloud data stream according to the first Gaussian weight distribution to obtain a first model of substation equipment; Perform SIFT feature registration based on the device image data stream and the device point cloud data stream to obtain a device feature guidance vector set; Based on the device feature guidance vector set, guiding the device image data stream to perform detail texture enhancement on the first model of the substation equipment to obtain a second model of the substation equipment; Perform UV unwrapping correction on the second model of the substation equipment according to the device image data stream and the device point cloud data stream to generate a third model of the substation equipment; The step of calculating a first Gaussian weight distribution based on the device point cloud data stream and a Gaussian weight parsing model includes: Dividing the device point cloud data stream into neighborhoods of each point according to a predetermined neighborhood radius to obtain neighborhoods of each point of the device; Calculating the neighborhood centroid of each point in the neighborhood of the device to obtain the neighborhood centroid of each point; Calculating the covariance matrix of each point in the neighborhood of the device to obtain the covariance matrix of each point; Inputting the neighborhood centroid of each point and the covariance matrix of each point into the Gaussian weight analytical model, and outputting multi-point Gaussian weights; Outputting the multi-point Gaussian weights as the first Gaussian weight distribution; The Gaussian weighted analytical model includes a Gaussian weighted analytical function, which is: ； in, Represents the Gaussian weight of point i, i represents point i in the device point cloud data stream, point i is any point in the device point cloud data stream, Represents the i-point covariance matrix, Characterize the determinant of the covariance matrix of point i, Characterizes the inverse matrix of the i-point covariance matrix, Represents an exponential function with the natural constant e as the base, Represents the centroid of the neighborhood of point i, Characterizes the deviation vector between point i and the centroid of its neighborhood, The transposed vector representing the deviation vector between point i and the centroid of its neighborhood.

2. A high-precision three-dimensional modeling method for substation equipment according to claim 1, characterized in that: Smoothing the device point cloud data stream according to the first Gaussian weight distribution to obtain a first model of substation equipment includes: Performing normalization calculation based on the first Gaussian weight distribution to obtain a second Gaussian weight distribution; Performing Gaussian weighted smoothing optimization on the device point cloud data stream according to the second Gaussian weight distribution to obtain a device smoothed point cloud set; Three-dimensional modeling is performed based on the smooth point cloud set of the equipment to obtain a first model of the substation equipment.

3. A high-precision three-dimensional modeling method for substation equipment according to claim 1, characterized in that: Performing SIFT feature registration according to the device image data stream and the device point cloud data stream to obtain a device feature guide vector set includes: Perform SIFT feature detection on the device image data stream to obtain a device image feature vector set; Perform SIFT feature detection based on the device point cloud data stream to obtain a device point cloud feature vector set; According to each device image feature vector in the device image feature vector set, matching cost identification is performed on the device point cloud feature vector set to obtain multiple matching cost coefficient sets; Perform matching cost minimization screening based on the multiple matching cost coefficient sets to determine the optimal matching cost distribution; According to the optimal matching cost distribution, the device image feature vector set and the device point cloud feature vector set are registered and associated to generate the device feature guidance vector set.

4. A high-precision three-dimensional modeling method for substation equipment according to claim 1, characterized in that: Based on the device feature guidance vector set, guiding the device image data stream to perform detail texture enhancement on the first model of the substation equipment to obtain a second model of the substation equipment, including: Based on the equipment feature guidance vector set, projecting the equipment image data stream onto the first model of the substation equipment to obtain an image projection point set; Performing detail texture optimization on the first model of the substation equipment according to the image projection point set to obtain an optimized model of the substation equipment; Performing detail texture transition smoothness evaluation based on the substation equipment optimization model to obtain a model transition smoothness coefficient; Determining whether the model transition smoothing coefficient is less than a predetermined transition smoothing coefficient; If the model transition smoothing coefficient is less than the predetermined transition smoothing coefficient, the substation equipment optimization model is subjected to transition smoothing optimization according to the predetermined transition smoothing coefficient to generate the second substation equipment model.

5. A high-precision three-dimensional modeling method for substation equipment according to claim 1, characterized in that: Performing UV unwrapping correction on the second model of the substation equipment according to the device image data stream and the device point cloud data stream to generate a third model of the substation equipment, including: Perform UV expansion on the second model of the substation equipment to obtain an equipment model expansion diagram; constructing a device detail texture map according to the device image data stream; Drawing a device geometric feature map based on the device point cloud data stream; performing deviation detection on the device model expansion image according to the device detail texture image and the device geometric feature image to obtain a device model deviation detection result; According to the equipment model deviation detection result, the second substation equipment model is corrected to obtain the third substation equipment model.

6. A high-precision 3D modeling platform for substation equipment, characterized by: The platform is used to implement the high-precision three-dimensional modeling method for substation equipment according to any one of claims 1 to 5, and the platform includes: The acquisition result acquisition module is used to obtain the device image acquisition results and device point cloud acquisition results of the substation equipment based on the camera and lidar equipment; a data stream acquisition module, configured to input the device image acquisition result into an image filter to obtain a device image data stream, and input the device point cloud acquisition result into a point cloud filter to obtain a device point cloud data stream; A first Gaussian weight distribution calculation module, configured to calculate a first Gaussian weight distribution based on the device point cloud data stream and a Gaussian weight parsing model; a substation equipment first model obtaining module, configured to perform smooth modeling on the equipment point cloud data stream according to the first Gaussian weight distribution to obtain a first substation equipment model; a device feature guide vector set acquisition module, configured to perform SIFT feature registration based on the device image data stream and the device point cloud data stream to obtain a device feature guide vector set; a substation equipment second model obtaining module, configured to guide the equipment image data stream to perform detail texture enhancement on the first substation equipment model based on the equipment feature guidance vector set, so as to obtain a second substation equipment model; The substation equipment third model generation module is used to perform UV expansion correction on the substation equipment second model according to the equipment image data stream and the equipment point cloud data stream to generate a substation equipment third model.

Images