Sparse processing method and device for dense point cloud data of power transmission tower, terminal equipment and storage medium

Abstract

The invention discloses a sparse processing method and device for dense point cloud data of a power transmission tower, terminal equipment and a storage medium. The method comprises the following steps: acquiring the dense point cloud data of the power transmission tower; wherein the dense point cloud data comprises a plurality of nodes, coordinate information of each node and a connection relationship of each node; according to the coordinate information of each node of the dense point cloud data and the connection relationship of each node, integrating each node of the dense point cloud data and neighbor nodes of each node to obtain each piece of integrated data; generating sparse point cloud data of the power transmission tower according to the integrated data; and constructing a three-dimensional model of the power transmission tower according to the sparse point cloud data of the power transmission tower. By implementing the method, the construction efficiency of the three-dimensional model of the power transmission tower can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method, device, terminal device and storage medium for sparsifying dense point cloud data of transmission towers. Background Art

[0002] With the development of lidar (LiDAR) and three-dimensional scanning technology, it has become increasingly convenient to obtain high-density point cloud data of transmission towers. These dense point cloud data contain rich spatial information and are widely used in fields such as monitoring, detection and modeling of transmission towers. After obtaining the dense point cloud data, the traditional method for processing point cloud data is to directly register the point cloud data and perform three-dimensional reconstruction based on the dense point cloud data to generate a three-dimensional model of the transmission tower. However, this method of directly processing based on dense point cloud data has a low efficiency in constructing the three-dimensional model due to the large amount of data in the dense point cloud data. Summary of the Invention

[0003] Embodiments of the present invention provide a method, device, terminal device and storage medium for sparsifying dense point cloud data of transmission towers, which can improve the efficiency of constructing a three-dimensional model of a transmission tower.

[0004] An embodiment of the present invention provides a method for sparsifying dense point cloud data of transmission towers, including:

[0005] Obtaining dense point cloud data of a transmission tower; wherein, the dense point cloud data includes a plurality of nodes, coordinate information of each node, and connection relationships of each node;

[0006] Integrating each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node and the connection relationships of each node of the dense point cloud data to obtain respective integrated data;

[0007] Generating sparse point cloud data of the transmission tower according to the respective integrated data;

[0008] Constructing a three-dimensional model of the transmission tower according to the sparse point cloud data of the transmission tower.

[0009] Further, the integrating each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node and the connection relationships of each node of the dense point cloud data to obtain respective integrated data includes:

[0010] Constructing an undirected graph according to the coordinate information of each node and the connection relationships of each node of the dense point cloud data; wherein, the nodes of the undirected graph correspond to each node of the dense point cloud data, and the edges of the undirected graph are the connection relationships of each node;

[0011] For each node in the undirected graph, calculate the relative position between the current node and its neighbor nodes according to the position information of the current node and the position information of the neighbor nodes of the current node;

[0012] Integrate the coordinate information of each neighbor node with a relative position less than a preset threshold and the coordinate information of the current node to obtain each integrated data.

[0013] Further, generating the sparse point cloud data of the transmission tower according to each integrated data includes:

[0014] Generate a number of node feature vectors according to each integrated data;

[0015] Aggregate each node feature vector to obtain a global feature vector;

[0016] Input the global feature vector into the sparse processing model, so that the sparse processing model generates the sparse point cloud data of the transmission tower according to the global feature vector.

[0017] Further, aggregating each node feature vector to obtain a global feature vector includes:

[0018] Input each node feature vector into the graph neural network, so that the graph neural network fuses each node feature and the features of the neighbor nodes of each node in each node feature vector to obtain the fused feature vector of each node;

[0019] Aggregate the fused feature vectors of each node to obtain a global feature vector.

[0020] Further, the construction of the sparse processing model includes:

[0021] Obtain a training sample set; wherein, the training sample set includes dense point cloud data samples of several transmission towers, global feature vector samples corresponding to the dense point cloud data samples of each transmission tower, and true sparse point cloud data corresponding to the dense point cloud data samples of each transmission tower;

[0022] Construct an initial sparse processing model; wherein, the initial sparse processing model includes: a generator and a discriminator;

[0023] The generator and the discriminator are trained adversarially with a training sample set until the combined loss is minimized, and then the generator is used as the sparse processing model; in each adversarial training, a global feature vector sample and random noise are input into the generator, so that the generator outputs predicted sparse point cloud data according to the global feature vector sample and the random noise, and the predicted sparse point cloud data is transmitted to the discriminator; real sparse point cloud data and dense point cloud data samples are used as the input of the discriminator, so that the discriminator determines a prediction error according to the predicted sparse point cloud data and the real sparse point cloud data, determines a content error according to the predicted sparse point cloud data and the dense point cloud data samples, and feeds back the prediction error and the content error to the generator, so that the generator adjusts the generator network parameters according to the prediction error and the content error; the combined loss of each adversarial training is determined according to the prediction error and the content error.

[0024] Further, determining the content error according to the predicted sparse point cloud data and the dense point cloud data samples includes:

[0025] Determining the content error according to the Hausdorff distance and the chamfer distance between the predicted sparse point cloud data and the dense point cloud data samples.

[0026] Further, after obtaining the dense point cloud data of the transmission tower, it further includes:

[0027] Denosing and filtering the dense point cloud data to obtain first preprocessed dense point cloud data;

[0028] Calculating the coordinate information of the point cloud centroid according to the first preprocessed dense point cloud data, and normalizing the coordinate information of each node of the first preprocessed dense point cloud data according to the coordinate information of the point cloud centroid to obtain second preprocessed dense point cloud data.

[0029] Based on the above method item embodiments, the present invention correspondingly provides device item embodiments;

[0030] An embodiment of the present invention correspondingly provides a sparse processing device for dense point cloud data of a transmission tower, including: a dense point cloud data acquisition module, an integration module, a sparse point cloud data generation module, and a three-dimensional model construction module;

[0031] The dense point cloud data acquisition module is used to acquire the dense point cloud data of the transmission tower; wherein, the dense point cloud data includes a plurality of nodes, the coordinate information of each node, and the connection relationship of each node;

[0032] The integration module is used to integrate each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node of the dense point cloud data and the connection relationship of each node to obtain each integrated data;

[0033] The sparse point cloud data generation module is used to generate sparse point cloud data of a transmission tower according to each integrated data;

[0034] The three-dimensional model construction module is used to construct a three-dimensional model of the transmission tower according to the sparse point cloud data of the transmission tower.

[0035] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for sparse processing of dense point cloud data of a transmission tower described in the above-mentioned embodiment of the present invention.

[0036] Another embodiment of the present invention provides a storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute the method for sparse processing of dense point cloud data of a transmission tower described in the above-mentioned embodiment of the present invention.

[0037] By implementing the present invention, the following beneficial effects are achieved:

[0038] The present invention provides a method, device, terminal device, and storage medium for sparse processing of dense point cloud data of a transmission tower. After obtaining the dense point cloud data of the transmission tower, the method integrates each node and its neighbor nodes based on the coordinate information of each node and the connection relationship of each node in the dense point cloud data, so that each node and its neighbor nodes can be centrally represented, reducing the data volume of the point cloud data. And the features of the neighbor nodes are not deleted in the integrated data. Then, sparse point cloud data of the transmission tower is generated according to the integrated data. When constructing a three-dimensional model of the transmission tower based on the sparse point cloud data of the transmission tower, the construction efficiency of the three-dimensional model can be accelerated, and the features of the dense point cloud data of the transmission tower are retained. On the basis of improving the generation efficiency of the three-dimensional model, the accuracy of the construction of the three-dimensional model of the transmission tower is ensured. Description of the Drawings

[0039] Figure 1 It is a schematic flowchart of a method for sparse processing of dense point cloud data of a transmission tower provided by an embodiment of the present invention.

[0040] Figure 2 It is a schematic structural diagram of a device for sparse processing of dense point cloud data of a transmission tower provided by an embodiment of the present invention. Detailed Embodiments

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0042] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present application.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above drawings are intended to cover non-exclusive inclusion.

[0044] In the description of the embodiments of the present application, technical terms such as "first" and "second" are only used to distinguish different objects and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity, specific order or primary-secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0045] Referring to "embodiments" herein means that specific features, structures or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase does not necessarily refer to the same embodiment at all positions in the specification, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0046] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0047] In the description of the embodiments of the present application, the term "a plurality of" means two or more (including two). Similarly, "a plurality of groups" means two or more groups (including two groups), and "a plurality of pieces" means two or more pieces (including two pieces).

[0048] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installation", "connection", "connection", "fixation" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can also be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to specific situations.

[0049] As Figure 1 shown, it is a method for sparsifying dense point cloud data of a transmission tower provided by an embodiment of the present invention, including:

[0050] Step S1: Obtain the dense point cloud data of the transmission tower; wherein, the dense point cloud data includes a number of nodes, the coordinate information of each node, and the connection relationship of each node;

[0051] Step S2: Integrate each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node and the connection relationship of each node of the dense point cloud data to obtain each integrated data;

[0052] Step S3: Generate sparse point cloud data of the transmission tower according to each integrated data;

[0053] Step S4: Construct a three-dimensional model of the transmission tower according to the sparse point cloud data of the transmission tower.

[0054] For step S1, use a high-precision lidar (LiDAR) or three-dimensional scanning device to obtain the dense point cloud data of the transmission tower. This dense point cloud data contains a number of nodes, the coordinate information of each node, and the connection relationship of each node. The coordinate information of each node is represented in the form of three-dimensional coordinates, and each node contains three coordinates x, y, and z.

[0055] In a preferred embodiment, after obtaining the dense point cloud data of the transmission tower, it further includes: denoising and filtering the dense point cloud data to obtain first preprocessed dense point cloud data; calculating the coordinate information of the point cloud centroid according to the first preprocessed dense point cloud data, and normalizing the coordinate information of each node of the first preprocessed dense point cloud data according to the coordinate information of the point cloud centroid to obtain second preprocessed dense point cloud data.

[0056] Specifically, denoising and filtering are performed on the above-collected dense point cloud data of the transmission tower to remove the noise and abnormal points in the dense point cloud data, so as to improve the quality of the obtained dense point cloud data. Denoising and filtering can be carried out by statistical filtering and outlier removal methods, and through denoising and filtering, the uniformity and accuracy of the dense point cloud data can be ensured. The data (i.e., the above-mentioned first preprocessed dense point cloud data) after the above denoising and filtering processing is subjected to coordinate normalization processing. The essence of coordinate normalization processing is to calculate the centroid of the point cloud of the dense point cloud data, and then normalize the coordinate information of each node based on the centroid of the point cloud, so that each node in the dense point cloud data can be normalized to a unified scale and coordinate system, eliminating the scale and position differences. Specifically, calculate the coordinate information of the centroid of the point cloud of the dense point cloud data. After subtracting the coordinate of the centroid of the point cloud from the coordinate of each node in the dense point cloud data, and then dividing by the scale factor, the normalized dense point cloud data (i.e., the above-mentioned second preprocessed dense point cloud data) can be obtained. The scale factor can be set according to the standard of normalization processing, and its value is not specifically limited in the present invention. Preferably, after normalization processing, data augmentation can also be performed on the dense point cloud data by randomly selecting, scaling, and adding noise to the dense point cloud data.

[0057] For step S2, in a preferred embodiment, integrating each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node and the connection relationship of each node of the dense point cloud data to obtain each integrated data includes: constructing an undirected graph according to the coordinate information of each node and the connection relationship of each node of the dense point cloud data; wherein, the nodes of the undirected graph correspond to each node of the dense point cloud data, and the edges of the undirected graph are the connection relationships of each node; for each node in the undirected graph, calculate the relative position between the current node and its neighbor nodes according to the position information of the current node and the position information of its neighbor nodes; integrate the coordinate information of each neighbor node with a relative position less than a preset threshold and the coordinate information of the current node to obtain each integrated data.

[0058] Specifically, each node of the above-mentioned normalized dense point cloud data is used as the node of the undirected graph to be constructed. The connection relationship between each node and the remaining nodes in the undirected graph constructs the edge set of each node. Using the k-nearest neighbor algorithm, for each node in the undirected graph, find its k nearest neighbor nodes and establish undirected edges. After determining each node and each undirected edge, an undirected graph of the dense point cloud data is generated. Based on this undirected graph, the adjacency matrix of the undirected graph is calculated. Among them, the adjacency matrix is the matrix representation of the undirected graph, and its essence is to convert the undirected graph into a matrix representation. Each element in the adjacency matrix represents the connection relationship between each node. For example, the element A ij indicates whether there is an edge between node i and node j, A ij= 0 indicates that there is no edge between node i and node j, A ij = 1 indicates that there is an edge between node i and node j. Converting an undirected graph into the representation form of an adjacency matrix is mainly to save storage space. Calculate the neighborhood features of each node based on the undirected graph to capture local geometric information. For each node in the undirected graph, according to the position information of the current node and the position information of k neighbor nodes of this node, calculate the relative positions between the current node and the k neighbor nodes respectively, and then combine the coordinate information of each neighbor node with a relative position less than the preset threshold and the coordinate information of the current node, that is, the integrated data is obtained. The integrated data contains the coordinate information of the current node and the coordinate information of each neighbor node with a relative position less than the preset threshold.

[0059] For step S3, in a preferred embodiment, the generating the sparse point cloud data of the transmission tower according to the integrated data includes: generating a number of node feature vectors according to the integrated data; aggregating the node feature vectors to obtain a global feature vector; inputting the global feature vector into a sparse processing model, so that the sparse processing model generates the sparse point cloud data of the transmission tower according to the global feature vector.

[0060] In a preferred embodiment, the aggregating the node feature vectors to obtain a global feature vector includes: inputting the node feature vectors into a graph neural network, so that the graph neural network fuses the node features in the node feature vectors and the features of the neighbor nodes of each node according to the node feature vectors to obtain the fused feature vectors of each node; aggregating the fused feature vectors of each node to obtain a global feature vector.

[0061] Specifically, a corresponding number of node feature vectors are generated according to the integrated data. The node feature vectors are input into the graph neural network GNN. The graph neural network GNN includes a multi-layer perceptron (MLP), a graph convolutional layer, a multi-scale feature fusion layer, a global feature extraction layer, and an attention mechanism. When the node feature vectors are input into the graph neural network GNN, they first enter the multi-layer perceptron, and the multi-layer perceptron performs a non-linear mapping on the node feature vectors, mapping the node feature vectors to a high-dimensional feature space, and at the same time using an activation function to increase non-linearity during the mapping process. Then, a convolution operation is performed on the graph convolutional layer, and the node feature vectors are updated according to the neighbor nodes of the node feature vectors. The feature update formula is as follows:

[0062]

[0063] Among them, represents the feature vector of node i at the l-th layer; represents the set of neighbor nodes of node i; d iDenotes the degree of node i, i.e., the number of connected edges; d j Denotes the degree of neighbor node j, i.e., the number of connected edges; Denotes the feature vector of node j at the (l - 1)-th layer; W (l) Denotes the weight matrix of the l-th layer; σ is the activation function, which can be the ReLU activation function.

[0064] Furthermore, input the feature vectors of each node after feature update into the multi-scale feature fusion layer. By capturing and fusing the structural features of the feature vectors of each node at different scales through the multi-scale feature fusion layer, the fused feature vectors of each node can be obtained. In addition, an attention mechanism is introduced in the graph convolution to enhance the model's attention to important nodes and edges. Calculate the attention coefficients between nodes, perform weighted summation on the features of neighbor nodes, and update the node features. Finally, through the global feature extraction layer using average pooling or max pooling, the fused feature vectors of all nodes are aggregated into a global feature vector.

[0065] Input the global feature vector into the sparse processing model, so that the sparse processing model generates the sparse point cloud data of the transmission tower according to the global feature vector.

[0066] In a preferred embodiment, the construction of the sparse processing model includes:

[0067] Obtain a training sample set; wherein, the training sample set includes dense point cloud data samples of several transmission towers, global feature vector samples corresponding to the dense point cloud data samples of each transmission tower, and real sparse point cloud data corresponding to the dense point cloud data samples of each transmission tower; construct an initial sparse processing model; wherein, the initial sparse processing model includes: a generator and a discriminator; perform adversarial training on the generator and the discriminator with the training sample set until the joint loss is minimized, and take the generator as the sparse processing model; in each adversarial training, input the global feature vector sample and random noise into the generator, so that the generator outputs predicted sparse point cloud data according to the global feature vector sample and random noise, and transmit the predicted sparse point cloud data to the discriminator; use the real sparse point cloud data and the dense point cloud data sample as the input of the discriminator, so that the discriminator determines the prediction error according to the predicted sparse point cloud data and the real sparse point cloud data, determines the content error according to the predicted sparse point cloud data and the dense point cloud data sample, and feedbacks the prediction error and the content error to the generator, so that the generator adjusts the network parameters of the generator according to the prediction error and the content error; the joint loss of each adversarial training is determined according to the prediction error and the content error.

[0068] In a preferred embodiment, determining the content error according to the predicted sparse point cloud data and the dense point cloud data samples includes: determining the content error according to the Hausdorff distance and the chamfer distance between the predicted sparse point cloud data and the dense point cloud data samples.

[0069] Specifically, obtain a training sample set including dense point cloud data samples of several transmission towers, global feature vector samples corresponding to the dense point cloud data samples of each transmission tower, and true sparse point cloud data corresponding to the dense point cloud data samples of each transmission tower. Construct an initial sparse processing model, which includes a generator G and a discriminator D.

[0070] The loss function of the discriminator is:

[0071]

[0072] where, L D is the loss function of the discriminator; Q real is the true sparse point cloud data; Q fake is the predicted sparse point cloud data generated by the generator; D(Q) is the discriminant output of the discriminator for the input point cloud data Q.

[0073] The loss function of the generator is:

[0074]

[0075] where, L G is the loss function of the generator; λ is a trade-off coefficient that controls the balance between the prediction error and the content error; L content is the content error, which measures the difference between the predicted sparse point cloud data generated by the generator and the dense point cloud data samples.

[0076] Train the generator G and the discriminator D by using the obtained training sample set in an alternating optimization manner. Specifically, for the loss function, in each alternating training, fix the generator G respectively, minimize the loss function L D , update the parameters of the discriminator D; fix the discriminator D, minimize the loss function L G , update the parameters of the generator G.

[0077] For the model training process, in each adversarial training, a global feature vector sample and random noise are input into the generator, so that the generator outputs predicted sparse point cloud data according to the global feature vector sample and random noise, and transmits the predicted sparse point cloud data to the discriminator; the real sparse point cloud data and the dense point cloud data sample are used as the input of the discriminator, so that the discriminator determines the prediction error according to the predicted sparse point cloud data and the real sparse point cloud data, determines the content error according to the predicted sparse point cloud data and the dense point cloud data sample, and feeds back the prediction error and the content error to the generator, so that the generator adjusts the generator network parameters, such as the trade-off coefficient λ, according to the prediction error and the content error. The content error between the predicted sparse point cloud data and the dense point cloud data sample, that is, the structural similarity, can be determined by calculating the Hausdorff distance and the Chamfer distance between the predicted sparse point cloud data and the dense point cloud data sample. When the joint loss, that is, the prediction error and the content error, is minimized, the generator at this time is used as the sparse processing model.

[0078] Preferably, during the model training process, a regularization method is used to prevent the model from overfitting and improve the generalization ability. Among them, the commonly used regularization methods include: weight decay, Dropout, and batch normalization. Weight decay controls the model complexity by adding a weight decay term to the loss function. Dropout randomly discards some neurons in the fully connected layer of the network to prevent overfitting. Batch normalization accelerates the training convergence by performing batch normalization after the output of each layer.

[0079] For step S4, a three-dimensional model of the transmission tower is constructed based on the sparse point cloud data of the transmission tower. Specifically, first, the generated sparse point cloud data is sorted to determine the connection order of each node in the sparse point cloud data. This sorting method can be sorted according to the spatial position of each node, such as the height z coordinate or the feature vector of each node. The sorted nodes are connected in sequence to form line segments to construct the basic topological structure of the transmission tower. For adjacent points, they can be connected according to certain distance and angle thresholds to ensure the rationality of the topological structure. The basic topological structure of the transmission tower is optimized, and some unreasonable connections are corrected, such as removing connections that exceed a certain distance threshold to prevent connections with too large spans; according to the structural characteristics of the transmission tower, the angles of the connections are constrained to maintain the rationality of the structure; horizontal or diagonal connections are added as needed to improve the structure. After the above processing is completed, the processed sparse point cloud data is rendered into a three-dimensional model of the transmission tower through a three-dimensional modeling tool for reference and analysis by transmission tower engineering personnel.

[0080] Based on the above method item embodiments, the present invention correspondingly provides device item embodiments.

[0081] Such as Figure 2As shown in the figure, an embodiment of the present invention provides a sparsification processing device for dense point cloud data of transmission towers, including: a dense point cloud data acquisition module, an integration module, a sparse point cloud data generation module, and a three-dimensional model construction module;

[0082] The dense point cloud data acquisition module is used to acquire dense point cloud data of transmission towers; wherein, the dense point cloud data includes a number of nodes, the coordinate information of each node, and the connection relationship of each node;

[0083] The integration module is used to integrate each node and its neighbor nodes of the dense point cloud data according to the coordinate information of each node and the connection relationship of each node of the dense point cloud data to obtain each integrated data;

[0084] The sparse point cloud data generation module is used to generate sparse point cloud data of transmission towers according to each integrated data;

[0085] The three-dimensional model construction module is used to construct a three-dimensional model of a transmission tower according to the sparse point cloud data of the transmission tower.

[0086] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement without creative work.

[0087] Those skilled in the art can clearly understand that for the convenience and conciseness, the specific working process of the device described above can refer to the corresponding process in the foregoing method embodiment, which will not be elaborated here.

[0088] Based on the foregoing method item embodiment, the present invention correspondingly provides a terminal device item embodiment.

[0089] An embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a sparsification processing method for dense point cloud data of transmission towers according to any one of the present invention.

[0090] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0091] The so-called processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device through various interfaces and lines.

[0092] The memory can be used to store the computer program. The processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0093] Based on the above method item embodiments, the present invention correspondingly provides storage medium item embodiments.

[0094] An embodiment of the present invention provides a storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute any one of the methods for sparse processing of transmission tower dense point cloud data in the present invention.

[0095] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of the foregoing method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0096] The foregoing is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art of the present technology, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A method for sparse processing of dense point cloud data of transmission towers, characterized in that: include: Acquire dense point cloud data of the transmission tower; wherein the dense point cloud data includes a number of nodes, coordinate information of each node and connection relationship of each node; Integrate each node and its neighboring nodes in the dense point cloud data according to the coordinate information of each node and the connection relationship of each node to obtain integrated data; Generate sparse point cloud data of the transmission tower based on each integrated data; Construct a 3D model of the transmission tower based on the sparse point cloud data of the transmission tower.

2. The method for sparse processing of dense point cloud data of a transmission tower according to claim 1, characterized in that: The step of integrating each node and its neighboring nodes according to the coordinate information of each node of the dense point cloud data and the connection relationship of each node to obtain integrated data includes: Constructing an undirected graph according to the coordinate information of each node of the dense point cloud data and the connection relationship of each node; wherein the nodes of the undirected graph correspond to each node of the dense point cloud data, and the edges of the undirected graph are the connection relationships of each node; For each node in the undirected graph, the relative position of the current node and the neighboring nodes of the current node is calculated according to the position information of the current node and the position information of the neighboring nodes of the current node; The coordinate information of each neighboring node whose relative position is less than a preset threshold is integrated with the coordinate information of the current node to obtain integrated data.

3. The method for sparse processing of dense point cloud data of a transmission tower according to claim 2, characterized in that: The step of generating sparse point cloud data of the transmission tower according to each integrated data comprises: Generate several node feature vectors according to each integrated data; Aggregate the feature vectors of each node to obtain the global feature vector; The global feature vector is input into a sparse processing model so that the sparse processing model generates sparse point cloud data of the transmission tower according to the global feature vector.

4. The method for sparse processing of dense point cloud data of a transmission tower according to claim 3, characterized in that: The process of aggregating the feature vectors of each node to obtain a global feature vector includes: Inputting each node feature vector into the graph neural network, so that the graph neural network fuses the features of each node in each node feature vector and the features of neighboring nodes of each node according to the feature vector of each node, to obtain a fused feature vector of each node; The fused feature vectors of each node are aggregated to obtain the global feature vector.

5. The method for sparse processing of dense point cloud data of a transmission tower according to claim 3, characterized in that: The construction of the sparse processing model includes: Acquire a training sample set; wherein the training sample set includes dense point cloud data samples of several transmission towers, global feature vector samples corresponding to the dense point cloud data samples of each transmission tower, and real sparse point cloud data corresponding to the dense point cloud data samples of each transmission tower; Constructing an initial sparse processing model; wherein the initial sparse processing model includes: a generator and a discriminator; The generator and the discriminator are trained adversarially with the training sample set until the joint loss is minimized, and the generator is used as the sparse processing model; in each adversarial training, the global feature vector sample and random noise are input into the generator, so that the generator predicts sparse point cloud data according to the global feature vector sample and the random noise output, and transmits the predicted sparse point cloud data to the discriminator; the real sparse point cloud data and the dense point cloud data samples are used as the input of the discriminator, so that the discriminator determines the prediction error according to the predicted sparse point cloud data and the real sparse point cloud data, determines the content error according to the predicted sparse point cloud data and the dense point cloud data samples, and feeds the prediction error and the content error back to the generator, so that the generator adjusts the generator network parameters according to the prediction error and the content error; the joint loss of each adversarial training is determined according to the prediction error and the content error.

6. The method for sparse processing of dense point cloud data of a transmission tower according to claim 5, characterized in that: The determining of the content error according to the predicted sparse point cloud data and dense point cloud data samples includes: The content error is determined based on the Hausdorff distance and chamfer distance of the predicted sparse point cloud data and dense point cloud data samples.

7. The method for sparse processing of dense point cloud data of a transmission tower according to claim 1, characterized in that: After obtaining the dense point cloud data of the transmission tower, it also includes: De-noising and filtering the dense point cloud data to obtain first pre-processed dense point cloud data; The coordinate information of the centroid of the point cloud is calculated according to the first preprocessed dense point cloud data, and the coordinate information of each node of the first preprocessed dense point cloud data is normalized according to the coordinate information of the centroid of the point cloud to obtain the second preprocessed dense point cloud data.

8. A sparse processing device for dense point cloud data of a transmission tower, characterized in that: include: Dense point cloud data acquisition module, integration module, sparse point cloud data generation module and 3D model construction module; The dense point cloud data acquisition module is used to acquire dense point cloud data of the transmission tower; wherein the dense point cloud data includes a number of nodes, coordinate information of each node and connection relationship of each node; The integration module is used to integrate each node of the dense point cloud data and its neighboring nodes according to the coordinate information of each node of the dense point cloud data and the connection relationship of each node to obtain each integrated data; The sparse point cloud data generation module is used to generate sparse point cloud data of the transmission tower according to each integrated data; The three-dimensional model building module is used to build a three-dimensional model of the transmission tower according to the sparse point cloud data of the transmission tower.

9. A terminal device, characterized in that: It comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a sparse processing method for dense point cloud data of a transmission tower as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the sparse processing method for dense point cloud data of a transmission tower as described in any one of claims 1 to 7.

Images