Electromagnetic environment measurement method and device for complex terrain crossing area

Abstract

The invention relates to an electromagnetic environment measurement method and device for a complex terrain crossing area. The method comprises the following steps: acquiring a source point cloud and a target point cloud of a to-be-detected area, and acquiring a filtering point cloud set corresponding to the source point cloud; performing region growth segmentation on the filtering point cloud set, and determining region geometric information corresponding to each filtering point cloud in the filtering point cloud set; obtaining a matching pair between the first feature descriptor and the second feature descriptor to construct a feature matching relation set; determining a target rigid transformation matrix according to the feature matching relation set and the regional geometric information corresponding to the filtering point cloud, and performing point cloud registration on the filtering point cloud through the target rigid transformation matrix; and combining the filtered point clouds after point cloud registration to generate a three-dimensional point cloud database of the to-be-detected area, and modeling an electromagnetic environment of the to-be-detected area according to the three-dimensional point cloud database. By adopting the method, the measurement accuracy of the electromagnetic environment of the crossing area of the power transmission line can be improved.

Description

Technical Field

[0001] The present application relates to the field of electromagnetic measurement technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for measuring the electromagnetic environment in an area where complex terrain crosses. Background Art

[0002] Transmission lines are the core components of the power system, and their design is directly related to the stability of the system operation and the impact on the surrounding environment. In the context of global attention to energy conservation, emission reduction and sustainable development, the environmental performance and electromagnetic compatibility of transmission lines have become one of the key areas of industry research. With the diversification of social needs and the rapid development of transmission technology, how to minimize the impact on the environment while ensuring efficient transmission of electric energy and achieve coordinated development of the power system and the natural environment has become a key issue that needs to be solved urgently, especially in mountainous areas with complex terrain and special areas with cross-overs, where the electromagnetic environment of transmission lines is even more complex.

[0003] In traditional technology, three-dimensional lidar equipment and visible light cameras can be installed at the crossing section of the transmission line to obtain point cloud data and image data of the crossing section in real time, and the obtained two-dimensional image data and three-dimensional point cloud data are fused, and the texture information of the image is assigned to the point cloud under a unified coordinate system to increase the dimension of the point cloud data, and the electromagnetic environment of the crossing section of the transmission line is monitored in real time based on the processed point cloud data. However, this method is difficult to capture accurate point cloud data under complex working conditions due to the undulating terrain and the particularity of the line layout, and point cloud registration distortion is prone to occur, resulting in low measurement accuracy of the electromagnetic environment of the crossing area of ​​the transmission line. Summary of the invention

[0004] Based on this, it is necessary to provide an electromagnetic environment measurement method, device, computer equipment, computer-readable storage medium and computer program product for complex terrain crossing areas that can improve the measurement accuracy of the electromagnetic environment in the crossing areas of transmission lines in order to solve the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for measuring the electromagnetic environment of an area crossing a complex terrain, comprising:

[0006] Acquire a source point cloud and a target point cloud of the area to be detected, and acquire a filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes a filtered point cloud obtained by statistically filtering the source point cloud;

[0007] Performing region growing segmentation on the filtered point cloud set to determine regional geometric information corresponding to each of the filtered point clouds in the filtered point cloud set;

[0008] Extract the first feature descriptor of the filtered point clouds in the filtered point cloud set and the second feature descriptor of the target point cloud, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set;

[0009] Determine a target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix;

[0010] Merge the filtered point clouds after point cloud registration into a filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database for the region to be detected, and model the electromagnetic environment of the region to be detected according to the three-dimensional point cloud database.

[0011] In one embodiment, the obtaining the filtered point cloud set corresponding to the source point cloud includes:

[0012] Obtain a set of neighboring points of the source point cloud; the set of neighboring points includes k neighboring points closest to the source point cloud, and k is a positive integer;

[0013] Calculate the average distance between the source point cloud and each of the neighboring points;

[0014] Calculate the standard deviation and the average value of each of the average distances according to the average distances respectively corresponding to each of the source point clouds;

[0015] Determine the source point clouds that meet the filtering judgment conditions from each of the source point clouds as the filtered point clouds according to the average distance, the standard deviation, and the average value, so as to obtain the filtered point cloud set.

[0016] In one embodiment, the filtering judgment condition includes that the average distance between the source point cloud and each of the neighboring points is less than or equal to a distance threshold;

[0017] Wherein, the distance threshold is equal to the sum of the average value and the standard deviation parameter, and the standard deviation parameter is equal to the product of an empirical threshold factor and the standard deviation.

[0018] In one embodiment, the performing region growing segmentation on the filtered point cloud set to determine the regional geometric information corresponding to each of the filtered point clouds in the filtered point cloud set includes:

[0019] Determine the local normal vector and local curvature of each of the filtered point clouds in the filtered point cloud set;

[0020] Determine seed points from each of the filtered point clouds according to the local curvature;

[0021] For any one of the said seed points, an initialization region corresponding to the seed point is determined, and a region label of the initialization region is marked;

[0022] For any one of the filtered point clouds included in the initialization region, neighborhood points that satisfy the region growth discrimination condition are searched in the neighborhood of any one of the filtered point clouds;

[0023] The neighborhood points are added to the initialization region, the region label of the neighborhood points is determined as the region label of the initialization region, and region geometric information corresponding to the region label is obtained;

[0024] Wherein, the region growth discrimination condition includes: the angle between the normal vector of the neighborhood point and the normal vector of any one of the filtered point clouds is less than an angle threshold, and the distance between the neighborhood point and any one of the filtered point clouds is less than a distance threshold.

[0025] In one embodiment, the obtaining of the matching pairs between the first feature descriptor and the second feature descriptor includes:

[0026] A first feature set is constructed based on the first feature descriptor, and a second feature set is constructed based on the second feature descriptor;

[0027] A distance matching value between any one of the first feature descriptors in the first feature set and any one of the second feature descriptors in the second feature set is calculated;

[0028] According to the distance matching value and a matching threshold, multiple pairs of matching pairs that satisfy the confidence condition are screened out between the first feature descriptor and the second feature descriptor;

[0029] Wherein, the confidence condition includes that the distance matching value is less than the matching threshold.

[0030] In one embodiment, the determining of the target rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud includes:

[0031] An initial rigid transformation matrix is determined according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud;

[0032] The initial rigid transformation matrix is iteratively optimized through an ICP registration function to obtain the target rigid transformation matrix;

[0033] Wherein, the ICP registration function is expressed as:

[0034] ;

[0035] Wherein, is the ICP registration function, where R represents the rotation matrix, t represents the translation vector, N represents the total number of source point clouds, and M represents the total number of target point clouds. represents the feature descriptor associated with the i-th source point cloud. represents the feature descriptor associated with the j-th target point cloud. represents the normalization parameter. represents the i-th source point cloud. represents the j-th target point cloud. represents the norm, and D represents the distance metric function.

[0036] In a second aspect, the present application also provides an electromagnetic environment measurement device for a complex terrain cross-span area, including:

[0037] A data acquisition module, configured to obtain the source point cloud and the target point cloud of the area to be detected, and obtain the filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes the filtered point cloud obtained by performing statistical filtering on the source point cloud.

[0038] A data processing module, configured to perform region growing segmentation on the filtered point cloud set to determine the regional geometric information corresponding to each filtered point cloud in the filtered point cloud set; extract the first feature descriptor of the filtered point cloud in the filtered point cloud set and the second feature descriptor of the target point cloud, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set.

[0039] An environment measurement module, configured to determine the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix; merge the filtered point clouds after point cloud registration into the filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database of the area to be detected, and model the electromagnetic environment of the area to be detected according to the three-dimensional point cloud database.

[0040] In a third aspect, the present application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0041] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0042] In a fifth aspect, the present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0043] The electromagnetic environment measurement method, device, computer device, computer-readable storage medium, and computer program product for the above-mentioned complex terrain cross-span area obtain the source point cloud and the target point cloud of the area to be detected, and obtain the filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes the filtered point cloud obtained by performing statistical filtering on the source point cloud; perform region growing segmentation on the filtered point cloud set to determine the regional geometric information corresponding to each filtered point cloud in the filtered point cloud set; extract the first feature descriptor of the filtered point cloud in the filtered point cloud set and the second feature descriptor of the target point cloud, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set; determine the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix; merge the filtered point clouds after point cloud registration into the filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database of the area to be detected, and model the electromagnetic environment of the area to be detected according to the three-dimensional point cloud database. By performing statistical filtering on the source point cloud, the filtered point cloud set is obtained, which greatly reduces the noise and outliers in the data, making the subsequent feature extraction and registration processes more stable and accurate. Using the regional geometric information obtained by region growing segmentation can accurately characterize the characteristics of different local planes or curved surface regions in the point cloud, providing effective prior information for key point selection and feature descriptor calculation. Moreover, a high-confidence matching pair is constructed between the first feature descriptor in the source point cloud and the second feature descriptor in the target point cloud, effectively reducing the probability of incorrect matching. By performing point cloud registration on the filtered point cloud through the target rigid transformation matrix determined according to the feature matching relationship set and the target rigid transformation matrix, it is convenient to construct a high-quality three-dimensional point cloud database to achieve high-precision point cloud registration. Finally, according to the three-dimensional point cloud database, an accurate electromagnetic environment of the area to be detected can be modeled, improving the measurement accuracy of the electromagnetic environment of the cross-span area of the transmission line. Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0045] Figure 1 It is an application environment diagram of an electromagnetic environment measurement method for a complex terrain cross-span area in an embodiment;

[0046] Figure 2 It is a flowchart of an electromagnetic environment measurement method for a complex terrain cross-span area in an embodiment;

[0047] Figure 3 It is a schematic flow chart of a method for measuring the electromagnetic environment in a complex terrain cross - spanning area in another embodiment;

[0048] Figure 4 It is a structural block diagram of a device for measuring the electromagnetic environment in a complex terrain cross - spanning area in an embodiment;

[0049] Figure 5 It is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0050] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] The method for measuring the electromagnetic environment in a complex terrain cross - spanning area provided by the embodiments of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed on the cloud or other network servers. The terminal 102 acquires the source point cloud and the target point cloud of the area to be detected, and acquires the filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes the filtered point cloud obtained by performing statistical filtering on the source point cloud; the terminal 102 performs region growing segmentation on the filtered point cloud set to determine the region geometric information corresponding to each filtered point cloud in the filtered point cloud set; the terminal 102 extracts the first feature descriptor of the filtered point cloud in the filtered point cloud set and the second feature descriptor of the target point cloud, and obtains the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set; the terminal 102 determines the target rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud, and performs point cloud registration on the filtered point cloud through the target rigid transformation matrix; the terminal 102 merges the filtered point clouds after point cloud registration into the filtered point cloud in the same coordinate system, generates a three-dimensional point cloud database of the area to be detected, and models the electromagnetic environment of the area to be detected according to the three-dimensional point cloud database. Among them, the terminal 102 can be but is not limited to various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0052] In an exemplary embodiment, as Figure 2 shown, a method for measuring the electromagnetic environment of a complex terrain cross-span area is provided. Taking the terminal 102 in Figure 1 as an example for illustration, it includes:

[0053] Step S202, acquire the source point cloud and the target point cloud of the area to be detected, and acquire the filtered point cloud set corresponding to the source point cloud.

[0054] In specific implementation, multiple lidars can be used to obtain the source point cloud and the target point cloud of the area to be detected, and the filtered point cloud set corresponding to the source point cloud can be obtained. Among them, multiple lidars respectively correspond to different acquisition perspectives. Multiple lidars can be deployed at different positions to form multiple acquisition perspectives covering the cross-span area of the transmission line, avoiding the blind area of a single perspective. The multiple acquisition perspectives respectively corresponding to the multiple lidars can include horizontal, oblique, top-down, bottom-up, etc.

[0055] In specific implementation, the area to be detected may include the cross-span area of the transmission line. Among them, the cross-span area of the transmission line may refer to the area where the transmission line intersects or runs in parallel at a short distance with other lines, buildings, natural obstacles (such as rivers, valleys) or artificial facilities (such as roads, railways, bridges) in the air or on the ground. Due to the complex electromagnetic field distribution, safety distance requirements and potential interference risks in such areas, they are the key focus objects in the design, inspection and electromagnetic environment analysis of the power transmission network.

[0056] Among them, the source point cloud includes the point cloud data to be registered, and the target point cloud includes the point cloud data used as a reference benchmark. In practical applications, the target point cloud can be a reference benchmark for point cloud registration. The target point cloud can provide the reference coordinate system for registration, and the source point cloud can be aligned to this reference coordinate system through transformation. The source point cloud may include the observation data from different perspectives and may contain a lot of noise due to the limitation of the acquisition perspective. Therefore, it needs to be optimized through statistical filtering and region growing segmentation; while the target point cloud can be the observation data from the main perspective, and the observation data from this main perspective has high accuracy and good perspective. Exemplarily, the source point cloud can be the point cloud of the transmission line collected from the bottom of the valley looking up (including terrain occlusion noise), and the target point cloud can be the high-precision reference point cloud collected from the top of the mountain looking down (covering the complete line trend). The source point cloud from multiple perspectives can be aligned to the global coordinate system of the target point cloud to construct a complete three-dimensional model.

[0057] Among them, the filtered point cloud set includes the filtered point cloud obtained by performing statistical filtering on the source point cloud. Among them, statistical filtering can be to establish a statistical model by analyzing the geometric distribution characteristics (such as distance, density) of the local neighborhood of the point cloud and removing the outlier points (noise) that do not conform to the model expectation.

[0058] In specific implementation, the terrain of the area to be detected can be first surveyed in detail to clarify key parts such as steep slopes, canyons, and crossing structures (such as bridges and tunnel entrances). Analyze the occlusion problems in the area to be detected, such as mountain corners, building occlusions, and densely vegetated areas, and identify them in advance. Determine the key areas and complementary perspectives to be collected according to the terrain characteristics of the area to be detected, and use maps and terrain models to assist in designing the collection plan to ensure that all complex areas have sufficient perspective coverage. The data collection of multiple lidars uses a combination of fixed and mobile collection methods. Fixed lidars can be arranged at high places or key nodes (such as the top of a steep slope and beside a crossing structure), while the lidars for mobile collection can be installed on drones or vehicles to supplement areas where it is difficult to install fixed equipment. For complex terrains, high-density and high-precision LiDAR devices are selected to reduce the influence of spot scattering and uneven local reflectivity. Due to the complexity of the terrain, strict calibration is required when different lidars adopt different installation methods to ensure external calibration and real-time time synchronization, and to ensure the consistency of multi-perspective data during subsequent registration.

[0059] Furthermore, taking advantage of the multi-perspective, multiple lidars collect data from different angles (horizontal, oblique, top-down) to make up for the data loss caused by terrain occlusion. For intersection areas (such as road intersections and valley junctions), it is ensured that there is sufficient overlap of the common area in the collected data of each lidar. For complex slopes and areas with large height changes, the collection parameters of the lidar, such as scanning frequency and horizontal resolution, can be adjusted in real time. A combination of preset collection modes and manual adjustment is adopted, and the operator flexibly adjusts the operating state according to the actual situation on site to reduce data loss or noise interference caused by changes in the reflection angle. To reduce the "dead corners" in strongly occluded areas, auxiliary collection devices (such as using auxiliary drones to collect local data) can also be deployed, or additional sensors can be added in key areas.

[0060] In one example, the LiDAR is carried on a drone to collect large-scale terrain data. Set the horizontal and vertical angle ranges of the device and adjust the sampling frequency according to the transmission line corridor. For short distances (<100 meters), the sampling frequency can be set to 100 kHz to 200 kHz; for medium distances (100 meters to 300 meters), the sampling frequency is increased to 300 kHz to 600 kHz; for long distances (>300 meters), the sampling frequency can be increased to a frequency of 1 MHz, so as to flexibly obtain the source point cloud and target point cloud of the area to be detected.

[0061] Step S204: Perform region growing segmentation on the filtered point cloud set to determine the regional geometric information corresponding to each filtered point cloud in the filtered point cloud set.

[0062] Among them, region growing segmentation is a technique for dividing point clouds into continuous regions based on local similarity criteria. By analyzing the geometric characteristics of point clouds (such as normal vectors and curvatures), points with consistent characteristics are clustered into the same region, providing structured data for subsequent feature matching and registration. After region growing segmentation is performed on the filtered point cloud set, each filtered point cloud carries a region label, and each region label corresponds to a region and the regional geometric information of that region.

[0063] Among them, the regional geometric information may include regional boundaries, regional areas, regional centers, regional principal directions, etc.

[0064] In specific implementation, the basic idea of the region growing algorithm is: starting from a seed point, neighboring points similar to it (in terms of distance, normal vector direction, etc.) are grouped into the same segmentation region.

[0065] Assume that the normal vector n(p filtered ) has been calculated for each filtered point cloud in the filtered point cloud set. Two main condition parameters can be set: the distance threshold d threshold and the angle threshold θ threshold . Among them, the distance threshold requires that the Euclidean distance between adjacent points is less than this threshold; the angle threshold requires that the angle between the normal vectors of adjacent points is less than this threshold to ensure surface smoothness and similarity.

[0066] In specific implementation, assume that S is the set of already segmented points. Determine the filtered point clouds that do not belong to the set of already segmented points from all filtered point clouds p filtered , and select an unsegmented seed point p seed . Input the seed point p seed into the initialized region R. For each point p filtered in the initialized region R, find its neighboring points u (k-neighborhood or radius query can be used). If the neighboring point u meets the conditions corresponding to the distance threshold and the angle threshold, then add the neighboring point u to the initialized region R until all points are segmented into a certain region. Among them, the initialized region R can be expressed as:

[0067] .

[0068] Therefore, for the neighboring point u to be judged, it can be judged whether the following conditions are met:

[0069] ;

[0070] ;

[0071] If all the above conditions are met, then this neighboring point can be classified into the same region as the point p filtered .

[0072] Step S206: Extract the first feature descriptor of the filtered point cloud and the second feature descriptor of the target point cloud in the filtered point cloud set, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set.

[0073] In specific implementation, the feature types to which the extracted first feature descriptor and second feature descriptor belong may include features such as FPFH (Fast Point Feature Histogram) feature and SHOT (Signature of Histograms of Orientations) feature. For each point in the source point cloud, search for the point with the most similar features in the target point cloud, and calculate the feature distance (such as Euclidean distance or cosine similarity, etc.) between the feature descriptors. Retain the matching pairs whose feature distances meet the matching threshold. For example, the matching pairs with feature distances less than the matching threshold can be retained to construct a feature matching relationship set. The feature matching relationship set may include multiple pairs of matching pairs between the first feature descriptor and the second feature descriptor.

[0074] Step S208: Determine the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix.

[0075] Among them, the regional geometric information corresponding to the filtered point cloud is the regional geometric information corresponding to the region to which the filtered point cloud belongs after region growing segmentation.

[0076] Among them, the target rigid transformation matrix can be composed of a rotation matrix and a translation vector, and is used for point cloud registration.

[0077] In specific implementation, according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, the geometric relationship and regional constraints of the matching pairs in the feature matching relationship set can be used to solve the optimal rigid transformation, obtain the target rigid transformation matrix, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix to minimize the alignment error.

[0078] In one embodiment, determining the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud includes:

[0079] Determine the initial rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud; perform iterative optimization on the initial rigid transformation matrix through the ICP registration function to obtain the target rigid transformation matrix.

[0080] Among them, the initial rigid transformation matrix is composed of a rotation matrix and a translation vector. Exemplarily, the initial rigid transformation matrix can be expressed as [R, t], including the rotation matrix R and the translation vector t.

[0081] Exemplarily, to calculate the initial rigid transformation matrix according to the matching pairs and regional geometric information, the centroid of the matching pairs can be calculated first. The expression of the centroid is:

[0082] ;

[0083] where, can represent the coordinates of the i-th point in the source point cloud, can represent the coordinates of the i-th point in the target point cloud; K can represent the number of matching pairs; can represent the centroid coordinates of the matching pairs.

[0084] The constructed de-centered points can be expressed as:

[0085] ;

[0086] where, can represent the de-centered coordinates.

[0087] The covariance matrix can be expressed as:

[0088] .

[0089] Performing singular value decomposition on the covariance matrix H, we can obtain:

[0090] ;

[0091] where, U represents the left singular vector matrix, and V T represents the right singular vector matrix.

[0092] The expressions for the rotation matrix R0 and the translation vector t0 are:

[0093] ;

[0094] ;

[0095] The initial rigid transformation matrix can be composed of the rotation matrix R0 and the translation vector t0.

[0096] Among them, the ICP (Iterative Closest Point) registration function is an algorithm that gradually minimizes the overall alignment error between two point clouds by iteratively searching for the closest neighbor point pairs and optimizing the rigid body transformation parameters (rotation and translation), ultimately achieving precise alignment of the point clouds. In the specific implementation, the ICP registration function starts with the initial rigid transformation matrix for optimization iteration. Starting with the initial rigid transformation matrix can significantly reduce the number of iterations required by ICP, accelerate convergence, and in scenarios with repetitive structures such as transmission lines, the initial matrix can guide ICP to avoid incorrect alignment and prevent local optima. Then, for each point in the source point cloud, the closest corresponding point is searched in the target point cloud to form a set of temporary matching pairs. According to the geometric relationship of the matching point pairs, the current optimal rotation matrix and translation vector are solved to minimize the overall alignment error of the matching point pairs. The transformation parameters calculated currently are applied to the source point cloud to make it move closer to the target point cloud. If the error change between two consecutive iterations is less than the set threshold or the maximum number of iterations is reached, the iteration is terminated, and the final transformation matrix, that is, the target rigid transformation matrix, is output. Among them, the target rigid transformation matrix is composed of the optimized rotation matrix and translation vector.

[0097] Point cloud registration is performed on the filtered point cloud through the target rigid transformation matrix, that is, the optimized transformation is applied to the filtered point cloud to achieve precise point cloud registration, and the filtered point clouds after point cloud registration are merged into a filtered point cloud in the same coordinate system, and the fused result of the multi-source point clouds after registration can be obtained, thereby constructing a complete three-dimensional point cloud database.

[0098] Step S210: Merge the filtered point clouds after point cloud registration into a filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database of the area to be detected, and model the electromagnetic environment of the area to be detected based on the three-dimensional point cloud database.

[0099] In the specific implementation, after all the filtered point clouds are registered through the rigid transformation matrix, the coordinates of all the registered point cloud sets are merged into a large point cloud. The voxel filter is used to merge the redundant points in the dense area, which can not only reduce the amount of point cloud data but also smooth the noise. Statistical filtering or radius filtering is used to exclude the outliers caused by the small errors existing in the registration. The merged point cloud is exported in a common 3D file format (such as PLY, PCD, STL, etc.) for use on other platforms, and a three-dimensional point cloud database is constructed to record all the data points that have been filtered, registered, and merged in the area to be detected.

[0100] In specific implementation, according to the three-dimensional point cloud database, the electromagnetic environment of the area to be detected is modeled, which may include: according to the three-dimensional point cloud database and the electromagnetic field simulation parameters of the area to be detected, the spatial distribution of the electromagnetic field can be calculated through numerical simulation, and the electromagnetic environment model of the area to be detected is modeled. For example, a visual or numerical electromagnetic field distribution map can be used as the modeling result of the electromagnetic environment. Among them, the electromagnetic field simulation parameters may include, but are not limited to, parameters required for simulation such as wire voltage level, current load, material properties, excitation source, dielectric constant, etc.

[0101] Exemplarily, the three-dimensional point cloud database can be input into the COMSOL platform for modeling the electromagnetic environment of the area to be detected. The COMSOL platform can be a multi-physics coupling simulation software platform based on the finite element method.

[0102] In practical applications, according to the data formats supported by COMSOL (such as STL, IGES, STEP, etc.), the point cloud data in the three-dimensional point cloud database can be converted into a compatible format. If directly importing the point cloud does not meet the requirements, surface reconstruction (such as using the Poisson reconstruction algorithm or other mesh generation methods) can be performed first to generate a geometric model. Import the converted geometric model into COMSOL, check whether the model is complete. If it is incomplete, geometric repair or mesh reconstruction can be performed to ensure that there are no geometric defects (such as self-intersection, holes, etc.). According to the actual situation of the area to be detected, select a suitable physical field module (such as a radio frequency module or an electromagnetic field module). Set the adaptive mesh according to the geometric complexity of the model to ensure the simulation accuracy. Then run the electromagnetic simulation solver to calculate the electromagnetic field distribution in the area to be detected. The analysis results generally include field distribution maps, electromagnetic wave propagation paths, coupling effects, etc. At the same time, define material properties, boundary conditions, excitation sources and other relevant parameters. Use the built-in post-processing function of COMSOL to export the simulation data and field distribution maps. Quantitative analysis such as local field strength, electromagnetic energy density, scattering parameters, etc. can be carried out in combination with actual requirements to evaluate the characteristics of the electromagnetic environment.

[0103] Furthermore, a three-dimensional modeling method of KPConv (Kernel Point Convolution) can also be adopted to process the three-dimensional point cloud database. KPConv captures local geometric features by defining convolution kernel points, can process point clouds with delicate local structures, can effectively capture the local characteristics of point clouds, and is suitable for processing fine models of complex terrains. The specific implementation steps are as follows: input the three-dimensional point cloud database established by lidar scanning, then detect the overlapping parts of adjacent regions in the radar scanning data, extract spatial features (such as planes, edges or corners), perform preliminary alignment of the point cloud data, then initialize two sets of point cloud data, and set the initial transformation matrix (R, t) to ensure the starting conditions of point cloud registration, and determine its neighborhood through the k-NN algorithm. For each test sample, the k-NN algorithm calculates its distances from all points in the training set, and its calculation formula is:

[0104] ;

[0105] where, and are two sample points respectively, and m is the feature dimension of the sample.

[0106] According to the calculated distances, select the k points with the closest distances to the test point from the training set, select the category that appears most frequently in the k neighborhoods as the category of the test point, then determine the shape of the convolution kernel according to the local geometric features of the point cloud, and then design the number of convolution kernel points according to the complexity of the point cloud data. During the training process, adjust the positions of the convolution kernel points through backpropagation and gradient descent so that they can adapt to different local structures of the point cloud. The deformable kernel points improve the ability of local feature extraction through learning and optimization. In the deep network, through max pooling, in each local neighborhood of the feature map, select the maximum value in this area as its local feature.

[0107] The feature generated by KPConv through convolution operations represents the local geometric information of each point and its neighborhood. Each point cloud point obtains a local feature vector through convolution operations, and this vector can describe the geometric shape of the point and its surrounding environment. In the subsequent layers of the network, the local features of all points are aggregated into global features.

[0108] Then start point cloud classification and segmentation. KPConv classifies through the global features learned by the previous network, uses a softmax classifier to predict the category of the point cloud, then outputs the probability that each point belongs to a certain category, and uses the cross-entropy loss function to predict the label of each point.

[0109] Its formula is:

[0110] ;

[0111] Among them, z i is the actually determined label, and [z] is the predicted value.

[0112] Then, KPConv is used to process point cloud denoising. It can identify and remove outliers and noise based on the relationship between points and neighborhoods. Finally, the point cloud is completed according to the captured local geometric features, so as to obtain a more complete three-dimensional reconstruction model.

[0113] After obtaining the transmission corridor model of complex terrain through the KPConv deep learning convolutional network, it is imported into COMSOL for modeling and simulation calculations. Set the electrostatic physical field, set the insulating boundary conditions for the uncharged domain, and simulate the shielding or isolation effect of the electric field. Build a three-phase voltage through the circuit physical field. The expression is:

[0114] ;

[0115] V peak The calculation formula of [V] is:

[0116] ;

[0117] Among them, V line is the line voltage of the transmission corridor.

[0118] Then, the set three-phase voltage in the circuit physical field is output to the electrostatic field as the voltage input terminal. Set the ground voltage to 0V through the terminal. Then, due to the large size of the model and the presence of domains with a large aspect ratio, a sub-domain grid is established through the distributed grid step in the mesh to achieve a grid transition with a large aspect ratio difference, which is convenient for convergence. By setting the transient study settings, the change of the electric field over time can be obtained. Especially when the voltage input changes (such as the fluctuation of the three-phase voltage), this analysis can help identify phenomena such as sudden changes and fluctuations in the electric field intensity. Through the above steps, the modeling results of the electromagnetic environment in the area to be detected can be obtained.

[0119] In the above electromagnetic environment measurement method for complex terrain crossing areas, the source point cloud and the target point cloud of the area to be detected are obtained, and the filtered point cloud set corresponding to the source point cloud is obtained; the filtered point cloud set includes the filtered point cloud obtained by statistically filtering the source point cloud; the filtered point cloud set is segmented by region growing to determine the regional geometric information corresponding to each filtered point cloud in the filtered point cloud set; the first feature descriptor of the filtered point cloud in the filtered point cloud set and the second feature descriptor of the target point cloud are extracted, and the matching pairs between the first feature descriptor and the second feature descriptor are obtained to construct a feature matching relationship set; according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, the target rigid transformation matrix is determined, and the filtered point cloud is registered by the target rigid transformation matrix; the filtered point clouds after point cloud registration are merged into a filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database of the area to be detected, and the electromagnetic environment of the area to be detected is modeled according to the three-dimensional point cloud database. By statistically filtering the source point cloud, the filtered point cloud set is obtained, which greatly reduces the noise and outliers in the data, making the subsequent feature extraction and registration processes more stable and accurate. Using the regional geometric information obtained by region growing segmentation can accurately describe the characteristics of different local planes or curved surface regions in the point cloud, providing effective prior information for key point selection and feature descriptor calculation. Moreover, high-confidence matching pairs are constructed between the first feature descriptor in the source point cloud and the second feature descriptor in the target point cloud, effectively reducing the probability of incorrect matching. By determining the target rigid transformation matrix according to the feature matching relationship set and the target rigid transformation matrix, the filtered point cloud is registered by point cloud, which is convenient for constructing a high-quality three-dimensional point cloud database to achieve high-precision point cloud registration. Finally, according to the three-dimensional point cloud database, the accurate electromagnetic environment of the area to be detected can be modeled, improving the measurement accuracy of the electromagnetic environment of the crossing area of the transmission line.

[0120] In some other embodiments, determining the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud includes: determining an initial rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud; and iteratively optimizing the initial rigid transformation matrix through an ICP registration function to obtain the target rigid transformation matrix.

[0121] Among them, the ICP registration function is expressed as:

[0122] ;

[0123] Among them, is the ICP registration function, R represents the rotation matrix, t represents the translation vector, N represents the total number of source point clouds, M represents the total number of target point clouds, represents the feature descriptor associated with the i-th source point cloud, denotes the feature descriptor associated with the j-th target point cloud, denotes the normalization parameter, denotes the i-th source point cloud, denotes the j-th target point cloud, denotes the norm, and D denotes the distance metric function.

[0124] wherein, denotes the feature descriptor associated with the i-th source point cloud, that is, the first feature descriptor corresponding to the i-th source point cloud; denotes the feature descriptor associated with the j-th target point cloud, that is, the second feature descriptor corresponding to the j-th target point cloud.

[0125] wherein, the distance metric function is a function for measuring the difference degree between two feature descriptors and The function. Optionally, the distance metric function can be a function for calculating the Euclidean distance, cosine similarity, etc.

[0126] In some other embodiments, obtaining the filtered point cloud set corresponding to the source point cloud includes: obtaining the neighboring point set of the source point cloud; the neighboring point set includes k neighboring points closest to the source point cloud, and k is a positive integer; calculating the average distance between the source point cloud and each neighboring point; calculating the standard deviation and average value of each average distance according to the average distances respectively corresponding to each source point cloud; determining the source point cloud that meets the filtering judgment condition from each source point cloud as the filtered point cloud according to the average distance, standard deviation and average value, so as to obtain the filtered point cloud set.

[0127] wherein, the neighboring point set of the source point cloud can be the neighboring point set found through the neighborhood. For each source point cloud, the k-nearest neighbor (KNN) method can be used to screen the neighboring point set, and the k-nearest neighbor query is performed for each source point cloud. For any source point cloud, the k-d tree (a binary tree with each node being a k-dimensional point) can be used for query, the distances between the source point cloud and other points are sorted, and the k points with the smallest distances are extracted as the k nearest neighbor points, and the k nearest neighbor points found are used as the neighboring point set of the source point cloud. For each source point cloud, the corresponding neighboring point set is saved for subsequent calculations, such as calculating statistical metrics such as the average distance between the source point cloud and each neighboring point in the neighboring point set. Once the k neighboring points are determined, the average value of the distances between the k neighboring points and the source point cloud can be calculated to obtain the average distance, which reflects the distribution density of the source point cloud in the local area and is used to statistically analyze the distance distribution of the entire point cloud or for subsequent filtering judgment.

[0128] wherein, the average value of each average distance can be used to reflect the global average density, and the standard deviation of each average distance can be used to reflect the degree of density fluctuation.

[0129] Among them, the filtering judgment condition is used to judge whether the average distance, standard deviation, and average value corresponding to each source point cloud meet certain criteria, so as to retain the source point cloud that meets the filtering judgment condition as the filtered point cloud, obtain the filtered point cloud set, and eliminate the source point cloud that does not meet the filtering judgment condition.

[0130] In one embodiment, the filtering judgment condition includes that the average distance between the source point cloud and each neighboring point is less than or equal to the distance threshold; wherein, the distance threshold is equal to the sum of the average value and the standard deviation parameter, and the standard deviation parameter is equal to the product of the empirical threshold factor and the standard deviation.

[0131] Among them, the empirical threshold factor can be used to control the filtering strictness. For example, it can take values from 1 to 3.

[0132] Exemplarily, the filtering judgment condition can be expressed as:

[0133] ;

[0134] ;

[0135] ;

[0136] ;

[0137] Among them, d i represents the average distance between the i-th source point cloud p i and each neighboring point in the neighboring point set; k represents the neighborhood search parameter, that is, search for k nearest neighbor points; S(p i ) represents the result set after neighborhood search of the source point cloud p i with the neighborhood search parameter k, that is, the neighboring point set; p l represents the l-th point in the neighboring point set of the source point cloud p i ; represents the norm; μ d represents the average neighborhood distance of all source point clouds, that is, the average value of the average distances corresponding to all source point clouds; N represents the total number of source point clouds, σ d represents the standard deviation of the average distances corresponding to all source point clouds, and α represents the empirical threshold factor.

[0138] In specific implementation, the obtained source point cloud is , and the target point cloud is . For each point p i in the source point cloud, use the points within its k-neighborhood to form the neighboring point set S(p i ), and then calculate the average distance dᵢ of the point p i in this neighborhood. The average neighborhood distance μ d of all points and the standard deviation σ dAfter calculation, the empirical threshold factor α is used to retain the points that meet the condition to form the filtered point cloud p filtered , and obtain the filtered point cloud set.

[0139] In the technical solution of this embodiment, the average distance of each source point cloud in the neighborhood can reflect the true distribution of the local density of the point cloud data, has good sensitivity to outliers and local noise, and by calculating the average neighborhood distance and the standard deviation of the neighborhood distance of the entire source point cloud, reliable filtering judgment conditions can be established to effectively remove outliers and isolated points, improve the overall data quality. The filtering method based on local statistical information can adapt to point cloud data in different density regions, retain more effective information in dense regions, and perform more stringent filtering in sparse or noisy regions to achieve adaptive optimization of the data, while reducing the risk of introducing incorrect data caused by sensor errors, environmental changes, etc.

[0140] In another embodiment, region growing segmentation is performed on the filtered point cloud set to determine the regional geometric information corresponding to each filtered point cloud in the filtered point cloud set, including: determining the local normal vector and local curvature of each filtered point cloud in the filtered point cloud set; determining seed points from each filtered point cloud according to the local curvature; for any seed point, determining a corresponding initialization region based on the seed point and marking the region label of the initialization region; for any filtered point cloud included in the initialization region, searching for neighborhood points that meet the region growing discrimination condition in the neighborhood of the any filtered point cloud; adding the neighborhood points to the initialization region, determining the region label of the neighborhood points as the region label of the initialization region, and obtaining the regional geometric information corresponding to the region label;

[0141] Among them, the region growing discrimination condition includes: the angle between the normal vector of the neighborhood point and the normal vector of any filtered point cloud is less than the angle threshold, and the distance between the neighborhood point and any filtered point cloud is less than the distance threshold.

[0142] In specific implementation, for each point p in the filtered point cloud set filtered , the local neighbor point set of point p is obtained by K-Nearest Neighbor (KNN) or fixed radius search in the point cloud, the mean value μ of all neighbor points in the neighborhood is calculated, then the covariance matrix is constructed, the covariance matrix is subjected to eigenvalue decomposition, and three eigenvalues and corresponding eigenvectors are obtained. Since the main local change directions of the surface are depicted by the largest and middle eigenvalues, and the eigenvector corresponding to the smallest eigenvalue is approximately perpendicular to the local surface. Therefore, the eigenvector corresponding to the smallest eigenvalue is selected as the local normal vector of point p filtered . filtered

[0143] For each point pfiltered In the neighborhood, calculate the local curvature through the eigenvalues of the covariance matrix (e.g., the ratio of the minimum eigenvalue to other eigenvalues). If the local curvature is low, it indicates that the point is in a relatively flat area and can be used as a candidate seed point. At the same time, the neighborhood density of the point can be considered. If the number of points in the neighborhood is small, it may be in the edge or noise area and generally will not be selected as a seed point. Add all the points that meet the above conditions to the seed point list, and then start region growing from these seed points. Starting from the selected seed points, initialize a region for each seed point to obtain the initialized region corresponding to the seed point and mark the region label of the seed point. For the points in the current region, perform neighborhood search and judge the following region growing discrimination conditions:

[0144] Distance condition: The Euclidean distance between the new point and the current point must be within the distance threshold.

[0145] Normal vector consistency condition: The angle between the local normal vector of the new point and the main direction in the current region is less than the angle threshold (e.g., 10° - 20°).

[0146] Other auxiliary conditions: Such as curvature or color information can also be used as judgment bases.

[0147] Furthermore, use a queue or recursive method to add the new points that meet the above conditions to the current region and mark them as the same region. For each newly added point, search its neighborhood in the same way and gradually expand the entire region through diffusion. If no new qualified points can be found in the current region, the region growing ends. If a point can meet the conditions of multiple seed regions at the same time, priorities can be set (e.g., closer distance, more consistent normal vector, etc.) for the attribution decision. For boundary points, special labels may need to be set or fine-tuning may be performed in post-processing.

[0148] During the region growing process, record the same region label for each point added to the region. In this way, the entire point cloud is finally divided into several labeled regions. For each region, calculate the region center (i.e., the average coordinates of all points in the region), the region boundary (i.e., according to the dispersion of points in the region, the minimum bounding box, convex hull, etc. can be calculated to represent the spatial range of the region), and the main direction (use principal component analysis PCA to perform eigen decomposition on all points in the region to obtain the main direction of the region, which corresponds to the main extension direction of the local surface) as the region geometric information corresponding to the region.

[0149] In the technical solution of this embodiment, through region growing segmentation, an accurate local normal vector can be obtained for each filtered point cloud, which reflects the directional characteristics of the point cloud surface in each local area. The local normal vector provides an accurate geometric description for subsequent feature matching and surface reconstruction, improving the accuracy of the overall modeling. According to the region growing discrimination conditions, the point cloud is divided into regions, and each region is labeled to facilitate the identification of different local geometric structures. This labeled division can help identify planar or curved surface regions inside the point cloud, enhance the distinguishability of different regions, and provide prior information for global registration. Using the region labels, calculate the regional geometric information (such as regional boundaries, areas, local feature distributions, etc.) for each local partition, which is beneficial for further analysis and understanding of the entire scene. The regional geometric information can provide a reliable basis for subsequent feature matching, initial rigid transformation estimation, and local optimization, improving the overall registration effect.

[0150] In another embodiment, obtaining the matching pairs between the first feature descriptor and the second feature descriptor includes: constructing a first feature set based on the first feature descriptor, and constructing a second feature set based on the second feature descriptor; calculating the distance matching value between any first feature descriptor in the first feature set and any second feature descriptor in the second feature set; and screening out multiple pairs of matching pairs that meet the confidence condition between the first feature descriptor and the second feature descriptor according to the distance matching value and the matching threshold.

[0151] Among them, the confidence condition includes that the distance matching value is less than the matching threshold.

[0152] In a specific implementation, for the key point x selected from the filtered point cloud i extract the first feature descriptor f i , and form the first feature set F p . At the same time, for the corresponding key point y in the target point cloud i extract the second feature descriptor g j , and form the second feature set F q . The first feature set F p and the second feature set F q can be expressed as:

[0153] ;

[0154] ;

[0155] Among them, represents the i-th first feature descriptor, represents the j-th second feature descriptor.

[0156] Feature matching using a distance metric function can be expressed as:

[0157] ;

[0158] Among them, the result output by the distance metric function is the distance matching value .

[0159] Set the distance threshold T D , if , it is considered that the confidence condition is satisfied and the matching is successful. The random sample consensus algorithm (RANSAC) is used to further eliminate the outlier matches, and finally the set of matching pairs is obtained .

[0160] In another embodiment, before performing region growing segmentation on the filtered point cloud set, random noise can be injected into any filtered point cloud in the filtered point cloud set and the target point cloud, and density perturbation processing can be performed on any filtered point cloud in the filtered point cloud set and the target point cloud to adjust the sampling density.

[0161] In specific implementation, due to the influence of the environment (such as rain, fog, vibration) or hardware errors on the lidar, the collected point cloud has coordinate offsets. Noise injection (such as Gaussian noise) can reproduce such errors to test the tolerance of the registration algorithm (such as ICP) to noise. The lidar is affected by the environment (such as rain, fog, vibration) or hardware errors, and the collected point cloud has coordinate offsets. Noise injection (such as Gaussian noise) can reproduce such errors. By simulating the uncertainties in the real scene, the robustness of the algorithm is verified and the actual application performance is optimized.

[0162] In another embodiment, as Figure 3 shown, a method for measuring the electromagnetic environment in a complex terrain crossing area is provided. Taking the application of this method to Figure 1 the terminal 102 in

[0163] Step S302, obtain the source point cloud and the target point cloud of the area to be detected, and obtain the filtered point cloud set corresponding to the source point cloud.

[0164] Among them, the filtered point cloud set includes the filtered point cloud obtained by performing statistical filtering on the source point cloud.

[0165] Step S304, obtain the set of neighboring points of the source point cloud, and calculate the average distance between the source point cloud and each neighboring point.

[0166] Among them, the set of neighboring points includes k neighboring points closest to the source point cloud, and k is a positive integer.

[0167] Step S306, calculate the standard deviation and the average value of each average distance according to the average distances respectively corresponding to each source point cloud.

[0168] Step S308: Determine the source point clouds that meet the filtering judgment conditions from each source point cloud according to the average distance, standard deviation, and mean value as the filtered point clouds to obtain a filtered point cloud set.

[0169] Among them, the filtering judgment conditions include that the average distance between the source point cloud and each neighboring point is less than or equal to the distance threshold; where the distance threshold is equal to the sum of the mean value and the standard deviation parameter, and the standard deviation parameter is equal to the product of the empirical threshold factor and the standard deviation.

[0170] Step S310: Determine the local normal vector and local curvature of each filtered point cloud in the filtered point cloud set, and determine the seed points from each filtered point cloud according to the local curvature.

[0171] Step S312: For any seed point, determine the corresponding initialization region based on the seed point and mark the region label of the initialization region.

[0172] Step S314: For any filtered point cloud included in the initialization region, search for neighboring points that meet the region growth discrimination conditions in the neighborhood of the any filtered point cloud.

[0173] Step S316: Add the neighboring points to the initialization region, determine the region label of the neighboring points as the region label of the initialization region, and obtain the region geometric information corresponding to the region label.

[0174] Among them, the region growth discrimination conditions include: the included angle between the normal vector of the neighboring point and the normal vector of any filtered point cloud is less than the angle threshold, and the distance between the neighboring point and any filtered point cloud is less than the distance threshold.

[0175] Step S318: Extract the first feature descriptor of the filtered point cloud in the filtered point cloud set and the second feature descriptor of the target point cloud, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set.

[0176] In one of the embodiments, obtaining the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set includes: constructing a first feature set based on the first feature descriptor and constructing a second feature set based on the second feature descriptor; calculating the distance matching value between any first feature descriptor in the first feature set and any second feature descriptor in the second feature set; screening out multiple pairs of matching pairs that meet the confidence condition between the first feature descriptor and the second feature descriptor according to the distance matching value and the matching threshold; where the confidence condition includes that the distance matching value is less than the matching threshold.

[0177] Step S320: Determine the initial rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud.

[0178] Step S322: Iteratively optimize the initial rigid transformation matrix through the ICP registration function to obtain the target rigid transformation matrix, and perform point cloud registration on the filtered point cloud using the target rigid transformation matrix.

[0179] Among them, the ICP registration function is expressed as:

[0180] ;

[0181] Among them, is the ICP registration function, R represents the rotation matrix, t represents the translation vector, N represents the total number of source point clouds, M represents the total number of target point clouds, represents the feature descriptor associated with the i-th source point cloud, represents the feature descriptor associated with the j-th target point cloud, represents the normalization parameter, represents the i-th source point cloud, represents the j-th target point cloud, represents the norm, and D represents the distance metric function.

[0182] Step S324: Merge the filtered point clouds after point cloud registration into a filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database corresponding to the area to be detected, and model the electromagnetic environment of the area to be detected based on the three-dimensional point cloud database.

[0183] It should be noted that the specific limitations of the above steps can refer to the specific limitations of a method for measuring the electromagnetic environment in a complex terrain cross-span area described above.

[0184] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown sequentially according to the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily have to be sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0185] Based on the same inventive concept, an embodiment of the present application further provides an electromagnetic environment measurement device for a complex terrain crossing area for implementing the electromagnetic environment measurement method for the complex terrain crossing area involved above. The implementation solutions for solving problems provided by this device are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the electromagnetic environment measurement device for the complex terrain crossing area provided below can refer to the limitations on the electromagnetic environment measurement method for the complex terrain crossing area in the above text, and will not be repeated here.

[0186] In an exemplary embodiment, as Figure 4 shown, an electromagnetic environment measurement device for a complex terrain crossing area is provided, including:

[0187] A data acquisition module 410, configured to obtain a source point cloud and a target point cloud of an area to be detected, and obtain a filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes filtered point clouds obtained by performing statistical filtering on the source point cloud;

[0188] A data processing module 420, configured to perform region growing segmentation on the filtered point cloud set to determine region geometric information corresponding to each filtered point cloud in the filtered point cloud set; extract a first feature descriptor of the filtered point clouds in the filtered point cloud set and a second feature descriptor of the target point cloud, and obtain matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set;

[0189] An environment measurement module 430, configured to determine a target rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix; merge the filtered point clouds after point cloud registration into a filtered point cloud in the same coordinate system to generate a three-dimensional point cloud database of the area to be detected, and model the electromagnetic environment of the area to be detected according to the three-dimensional point cloud database.

[0190] In one of the embodiments, the data acquisition module 410 is specifically configured to obtain a set of neighboring points of the source point cloud; the set of neighboring points includes k neighboring points closest to the source point cloud, where k is a positive integer; calculate the average distance between the source point cloud and each of the neighboring points; calculate the standard deviation and the average value of each of the average distances according to the average distances respectively corresponding to each source point cloud; determine, from each source point cloud, the source point cloud that meets the filtering judgment condition as the filtered point cloud according to the average distance, the standard deviation, and the average value, so as to obtain the filtered point cloud set.

[0191] In one embodiment, the filtering judgment condition includes that the average distance between the source point cloud and each adjacent point is less than or equal to a distance threshold; wherein, the distance threshold is equal to the sum of the average value and the standard deviation parameter, and the standard deviation parameter is equal to the product of the empirical threshold factor and the standard deviation.

[0192] In one embodiment, the data processing module 420 is specifically configured to determine the local normal vector and local curvature of each of the filtered point clouds in the filtered point cloud set; determine seed points from each of the filtered point clouds according to the local curvature; for any one of the seed points, determine a corresponding initialization region based on the seed point and mark the region label of the initialization region; for any one of the filtered point clouds included in the initialization region, search for neighborhood points that meet the region growth discrimination condition in the neighborhood of any one of the filtered point clouds; add the neighborhood points to the initialization region, determine the region label of the neighborhood points as the region label of the initialization region, and obtain region geometric information corresponding to the region label; wherein, the region growth discrimination condition includes: the angle between the normal vector of the neighborhood point and the normal vector of any one of the filtered point clouds is less than an angle threshold, and the distance between the neighborhood point and any one of the filtered point clouds is less than a distance threshold.

[0193] In one embodiment, the data processing module 420 is specifically configured to: construct a first feature set based on a first feature descriptor and construct a second feature set based on a second feature descriptor; calculate a distance matching value between any first feature descriptor in the first feature set and any second feature descriptor in the second feature set; screen out multiple pairs of matching pairs that meet the confidence condition between the first feature descriptor and the second feature descriptor according to the distance matching value and the matching threshold; wherein, the confidence condition includes that the distance matching value is less than the matching threshold.

[0194] In one embodiment, the environment measurement module 430 is specifically configured to determine an initial rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud; perform iterative optimization on the initial rigid transformation matrix through an ICP registration function to obtain the target rigid transformation matrix;

[0195] wherein, the ICP registration function is expressed as:

[0196] ;

[0197] wherein, is the ICP registration function, R represents the rotation matrix, t represents the translation vector, N represents the total number of source point clouds, M represents the total number of target point clouds, represents the feature descriptor associated with the i-th source point cloud, represents the feature descriptor associated with the j-th target point cloud, represents a normalization parameter represents the i-th source point cloud represents the j-th target point cloud represents a norm, and D represents a distance metric function.

[0198] Each module in the electromagnetic environment measurement device in the above-mentioned complex terrain crossover area can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0199] In an exemplary embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 5 shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used for the processor to exchange information with external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals, and the wireless method can be implemented through WIFI, a mobile cellular network, near field communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for measuring the electromagnetic environment in a complex terrain crossover area. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad set on the shell of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0200] Those skilled in the art can understand that Figure 5 the structure shown in

[0201] In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0202] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0203] In an embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0204] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0205] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0206] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope recorded in the present application.

[0207] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. An electromagnetic environment measurement method for complex terrain cross - spanning areas, characterized in that, The method includes: Obtaining a source point cloud and a target point cloud of a region to be detected, and obtaining a filtered point cloud set corresponding to the source point cloud; the filtered point cloud set includes filtered point clouds obtained by performing statistical filtering on the source point cloud; Performing region growing segmentation on the filtered point cloud set to determine region geometric information corresponding to each of the filtered point clouds in the filtered point cloud set; Extracting a first feature descriptor of the filtered point clouds in the filtered point cloud set and a second feature descriptor of the target point cloud, and obtaining matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set; Determining a target rigid transformation matrix according to the feature matching relationship set and the region geometric information corresponding to the filtered point cloud, and performing point cloud registration on the filtered point cloud by using the target rigid transformation matrix; Merging the filtered point clouds after point cloud registration into filtered point clouds in the same coordinate system to generate a three-dimensional point cloud database of the region to be detected, and modeling the electromagnetic environment of the region to be detected according to the three-dimensional point cloud database.

2. The method according to claim 1, wherein The obtaining the filtered point cloud set corresponding to the source point cloud includes: Obtaining a set of neighboring points of the source point cloud; the set of neighboring points includes k neighboring points closest to the source point cloud, and k is a positive integer; Calculating the average distance between the source point cloud and each of the neighboring points; Calculating the standard deviation and the average value of the average distances according to the average distances respectively corresponding to the respective source point clouds; Determining, from the respective source point clouds, source point clouds that meet the filtering judgment condition as filtered point clouds according to the average distance, the standard deviation, and the average value, so as to obtain the filtered point cloud set.

3. The method according to claim 2, wherein The filtering judgment condition includes that the average distance between the source point cloud and each of the neighboring points is less than or equal to a distance threshold; Wherein, the distance threshold is equal to the sum of the average value and a standard deviation parameter, and the standard deviation parameter is equal to the product of an empirical threshold factor and the standard deviation.

4. The method according to claim 1, characterized in that The performing region growing segmentation on the filtered point cloud set to determine region geometric information corresponding to each of the filtered point clouds in the filtered point cloud set includes: Determining the local normal vector and local curvature of each of the filtered point clouds in the filtered point cloud set; Determining seed points from the respective filtered point clouds according to the local curvature; For any one of the seed points, determining a corresponding initialization region based on the seed point and marking the region label of the initialization region; For any one of the filtered point clouds included in the initialization region, searching for neighborhood points that meet the region growing discrimination condition in the neighborhood of the filtered point cloud; Adding the neighborhood points to the initialization region, determining the region label of the neighborhood points as the region label of the initialization region, and obtaining region geometric information corresponding to the region label; Wherein, the region growing discrimination condition includes: the included angle between the normal vector of the neighborhood point and the normal vector of any one of the filtered point clouds is less than an angle threshold, and the distance between the neighborhood point and any one of the filtered point clouds is less than a distance threshold.

5. The method according to claim 1, characterized in that, The obtaining the matching pairs between the first feature descriptor and the second feature descriptor includes: Construct a first feature set based on the first feature descriptor, and construct a second feature set based on the second feature descriptor; Calculate the distance matching value between any one of the first feature descriptors in the first feature set and any one of the second feature descriptors in the second feature set; According to the distance matching value and the matching threshold, filter out multiple pairs of matching pairs that meet the confidence condition between the first feature descriptor and the second feature descriptor; Wherein, the confidence condition includes that the distance matching value is less than the matching threshold.

6. The method according to claim 1, wherein The determining the target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud includes: Determine an initial rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud; Iteratively optimize the initial rigid transformation matrix through an ICP registration function to obtain the target rigid transformation matrix; Wherein, the ICP registration function is expressed as: ； Among them, is the ICP registration function, R represents the rotation matrix, t represents the translation vector, N represents the total number of source point clouds, M represents the total number of target point clouds, represents the feature descriptor associated with the i-th source point cloud, represents the feature descriptor associated with the j-th target point cloud, represents the normalization parameter, represents the i-th source point cloud, represents the j-th target point cloud, represents the norm, and D represents the distance metric function.

7. An electromagnetic environment measurement device for a complex terrain crossover area, characterized in that The device includes: A data acquisition module, configured to acquire a source point cloud and a target point cloud of a region to be detected, and acquire a set of filtered point clouds corresponding to the source point cloud; the set of filtered point clouds includes filtered point clouds obtained by statistically filtering the source point cloud; A data processing module, configured to perform region growing segmentation on the set of filtered point clouds to determine the regional geometric information corresponding to each of the filtered point clouds in the set of filtered point clouds; extract the first feature descriptor of the filtered point cloud in the set of filtered point clouds and the second feature descriptor of the target point cloud, and obtain the matching pairs between the first feature descriptor and the second feature descriptor to construct a feature matching relationship set; An environment measurement module, configured to determine a target rigid transformation matrix according to the feature matching relationship set and the regional geometric information corresponding to the filtered point cloud, and perform point cloud registration on the filtered point cloud through the target rigid transformation matrix; merge the filtered point clouds after point cloud registration into a filtered point cloud in the same coordinate system, generate a three-dimensional point cloud database of the region to be detected, and model the electromagnetic environment of the region to be detected according to the three-dimensional point cloud database.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images