Point cloud data simplification method for power line towers, and system

Abstract

The present invention relates to the technical field of data simplification, in particular to a point cloud data simplification method for power line towers, and a system. The simplification method comprises: collecting point cloud data of power line towers, and carrying out data processing; performing data feature extraction on the processed data; classifying the point cloud data on the basis of the extracted features; and simplifying the classified point cloud data by means of using a point cloud simplification algorithm. The present invention introduces a deep learning technology to automatically learn and extract rich semantic features in point cloud data, thereby providing efficient feature representations for subsequent classification and simplification; by using extracted advanced semantic features and by means of a deep learning algorithm, the present invention can perform accurate classification and recognition on power line towers, thereby achieving automatic target detection and recognition; the present invention achieves intelligent simplification of point cloud data of power line towers by means of an intelligent merging strategy, so as to greatly reduce the data volume while reserving key features, thereby improving storage and transmission efficiency.

Description

A method and system for simplifying point cloud data of power line towers Technical Field

[0001] The present invention relates to the technical field of data simplification, and in particular to a method and system for simplification of point cloud data of power line towers. Background Art

[0002] The point cloud data obtained by using a 3D laser scanner is often very dense, and the number of points in the point cloud data is often as high as tens of millions or even hundreds of millions. Even if the point cloud data is denoised, the number of points in the point cloud data is still large, so these original point cloud data are often not used directly for surface reconstruction and other tasks, because this will make the subsequent processing time-consuming and consume too many computer resources, and the accuracy of the reconstructed surface may not be high, and may even result in greater errors. Therefore, before performing spatial point cloud surface reconstruction, it is often necessary to perform point cloud simplification operations on high-density point cloud data. The purpose of point cloud simplification is to delete as many redundant data points as possible while maintaining the shape characteristics and geometric feature information of the original point cloud;

[0003] Existing point cloud reduction methods aren't specifically designed for power line tower point cloud data, but rather are generally applicable algorithms. However, these methods suffer from certain deficiencies in practice. While excessive point cloud reduction can meet the real-time computing needs of mobile platforms, the loss of point cloud information makes tower identification difficult, hindering subsequent operations. Insufficient point cloud reduction, retaining a significant portion of the data, can improve tower identification, but the excess data prevents mobile devices from responding to commands promptly, hindering subsequent power operations.

[0004] Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the present invention provides a point cloud data simplification method for power line towers. By adopting laser scanning and image fusion technology, it can obtain three-dimensional information of power line towers with high precision, ensuring the accuracy and reliability of the data; by introducing deep learning technology, it can automatically learn and extract rich semantic features in point cloud data, providing efficient feature representation for subsequent classification and simplification.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a method for simplifying point cloud data of power line towers, comprising the following steps:

[0008] Collect point cloud data of power line towers and perform data processing;

[0009] Extract data features from the processed data;

[0010] Classify point cloud data based on the extracted features;

[0011] The classified point cloud data is simplified using the point cloud simplification algorithm.

[0012] As a preferred embodiment of the method for simplifying point cloud data of a power line tower described in the present invention, the point cloud data of the power line tower is collected by using a laser radar device. The laser radar device transmits a laser beam to the power line tower. When the laser beam hits the target surface, the laser beam is reflected back to the laser radar device. By measuring the time difference and angle information of the reflected laser beam, the point cloud data of the power line tower can be completed.

[0013] The time difference of the reflected laser beam is measured by using the propagation speed of light c and the time difference t of the laser beam to calculate the distance from the laser radar device to the target point. The specific calculation formula is as follows:

[0014] Where c represents the propagation speed of the laser beam in the air, t represents the time difference between the round trip of the laser beam, and d represents the distance measured from the lidar device to the target point;

[0015] The angle information is the horizontal angle θ of the laser radar device. X and the vertical angle θ Y , and use the distance d measured from the lidar device to the target to convert it into the three-dimensional coordinates of the point cloud data. The specific conversion formula is as follows:

[0016] x=d×cos(θ X )×cos(θ Y )

[0017] y=d×sin(θ X )×cos(θ Y )

[0018] z=d×sin(θ Y )

[0019] Among them, θ X Represents the horizontal angle of the lidar device, θ Y It represents the vertical angle of the LiDAR device, d represents the distance from the LiDAR device to the target point, and x, y, and z represent the horizontal, vertical, and height coordinates of the point cloud data in the LiDAR device coordinate system, respectively.

[0020] As a preferred solution of the point cloud data simplification method for power line towers described in the present invention, the data processing is to process the collected point cloud data, including noise suppression and data cleaning, thereby improving the efficiency of point cloud data simplification and reducing interference from irrelevant data. The noise suppression is to remove points that are inconsistent with the statistical characteristics of surrounding points by calculating the domain statistical characteristics of the point cloud data center. The specific implementation is as follows:

[0021] For each three-dimensional coordinate (x i ,y i ,z i ), calculate the mean value of the points in the point cloud data domain and standard deviation The specific calculation formula is as follows:

[0022] Where N represents the number of domain points, z j represents the height coordinate of the jth point in the domain, represents the mean height of the points in the area of ​​point i, represents the standard deviation of the heights of points within the area of ​​point i;

[0023] When the height coordinate, height mean and height difference of point i satisfy the formula When , it means that point i is noise and needs to be removed to improve the quality of point cloud data;

[0024] When the height coordinate, height mean and height standard deviation of point i satisfy the formula When , it means that point i is not noise, and the data of point i is retained.

[0025] As a preferred embodiment of the point cloud data simplification method for power line towers of the present invention, the data feature extraction of the processed data is performed by converting the point cloud data into a three-dimensional voxel network, extracting global feature information using a convolutional neural network, and selecting features representing the structure and spatial information of the power line tower from the extracted global feature information;

[0026] The point cloud data is converted into a three-dimensional voxel network by converting the collected three-dimensional coordinates (x i ,y i ,z i ) is divided into a three-dimensional network, and each network unit represents a voxel, and each voxel represents the information in the point cloud data. The specific implementation is as follows:

[0027] The three-dimensional coordinates of the point cloud data (x i ,y i ,z i) is divided into a voxel network of size L×W×H, then the three-dimensional coordinates of the point cloud data (x i ,y i ,z i ) is mapped to the coordinates of (x′ i ,y′ i ,z′ i ), the specific mapping formula is as follows:

[0028] Among them, x i 、y i 、z i They represent the horizontal coordinate, vertical coordinate and height coordinate of point i respectively, and L, W and H represent the pixel size of the voxel network respectively.

[0029] As a preferred solution of the point cloud data simplification method for power line towers described in the present invention, the extraction of global feature information using a convolutional neural network is to process point cloud data using a deep learning network, including performing feature mapping on the point cloud data using a multilayer perceptron and aggregating the mapped feature vectors; the feature mapping of the point cloud data using a multilayer perceptron is to convert the coordinates of the point cloud data into feature vectors in a high-dimensional feature space. The specific conversion formula is as follows:

[0030] T(x,y,z)=f(W1·x,W2·y,W3·z)

[0031] f(x)=max(0,x)

[0032] Where W represents the weight value corresponding to the neuron in the multilayer perceptron, T(x,y,z) represents the transformed feature vector, and f(x) represents the correction function to ensure that the transformed coordinates are positive and enhance the network's expressive power.

[0033] The mapped feature vectors are aggregated by using maximum pooling to extract the global feature vector representation from the local feature vectors. The maximum pooling selects the maximum value as the representative in the local feature vector to capture the most important information in the local feature vector and then aggregates it into a global feature vector. The specific implementation is as follows:

[0034] Among them, T i represents the local eigenvector, represents the average value of all local eigenvectors, T out The global feature vector after aggregation is represented by selecting the maximum value in the local vector as the global feature.

[0035] As a preferred embodiment of the point cloud data simplification method for power line towers described in the present invention, the classification of point cloud data based on the extracted features is performed by classifying the point cloud data using a deep learning algorithm, wherein the deep learning algorithm inputs the aggregated feature vector into a fully connected layer, maps the aggregated features to scores of each category through the fully connected layer, and converts the scores into probability distributions of different categories through a softmax activation function, thereby achieving classification of the point cloud data;

[0036] The conversion of scores into probability distributions of different categories through the softmax activation function is to use the softmax function to convert different elements in the output vector of the fully connected layer into probabilities of corresponding categories. The specific implementation is as follows:

[0037] Among them, (T′ out ) i Represents the score of the category corresponding to the i-th element in the output vector of the fully connected layer, N represents the total number of categories, Softmax(T′ out ) i Represents the probability of the category corresponding to the i-th element in the output vector.

[0038] As a preferred embodiment of the point cloud data simplification method for power line towers described in the present invention, the method comprises: simplifying the classified point cloud data using a point cloud simplification algorithm by using a triangulated network algorithm to simplify the point cloud data; the triangulated network algorithm constructs a triangulated network model using the coordinates of the point cloud data, calculates a triangle cost metric in the triangulated network model, and formulates simplification rules based on the triangle cost metric to simplify the point cloud data;

[0039] The method of constructing a triangulated network model using point cloud data coordinates is to convert the three-dimensional coordinates (x i ,y i ,z i ) is constructed into a triangle t(x t ,y t ,z t ) to complete the construction of the triangular network model. The triangle cost metric in the triangular network model is calculated based on the homogeneous coordinates corresponding to the constructed triangles. The specific calculation formula is as follows:

[0040] E t =t′Qt′

[0041] t′=(x t ,y t ,z t ,1)

[0042] Among them, E t represents the cost metric of the triangle, t′ is the homogeneous coordinate of triangle t, represents the plane equation of triangle t, Q represents a 4×4 symmetric matrix used to measure the error of the triangle, N i Represents vertex x i The corresponding normal vector is also the normal vector of triangle t.

[0043] Another object of the present invention is to provide a point cloud data simplification system for power line towers, which can accurately classify and identify power line towers by utilizing extracted high-level semantic features and relying on deep learning algorithms, thereby realizing automated target detection and recognition. Through intelligent merging strategies, intelligent simplification of power line tower point cloud data is achieved, which significantly reduces the amount of data while retaining key features, thereby improving storage and transmission efficiency.

[0044] As a preferred solution of the point cloud data simplification system for power line towers described in the present invention, it includes: a data acquisition and processing module, a feature extraction module, a data classification module and a point cloud data simplification module; the data acquisition and processing module is used to collect and process the point cloud data of the power line tower; the feature extraction module is used to extract features from the processed data; the data classification module is used to classify the point cloud data using the extracted features; and the point cloud data simplification module is used to simplify the classified point cloud data.

[0045] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, steps of a point cloud data simplification method for power line towers are implemented.

[0046] A computer-readable storage medium stores a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of a point cloud data simplification method for power line towers.

[0047] The beneficial effects of the present invention are as follows: by adopting laser scanning and image fusion technology, the present invention can obtain three-dimensional information of power line towers with high precision, ensuring the accuracy and reliability of the data; by introducing deep learning technology, it can automatically learn and extract rich semantic features in point cloud data, providing efficient feature representation for subsequent classification and simplification; by utilizing the extracted high-level semantic features and with the help of deep learning algorithms, it can accurately classify and identify power line towers, realizing automatic target detection and recognition; through intelligent merging strategies, it realizes intelligent simplification of power line tower point cloud data, while retaining key features, greatly reducing the amount of data and improving storage and transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0049] FIG1 is a schematic diagram showing the overall method steps of a method for simplifying point cloud data of a power line tower according to the present invention.

[0050] FIG2 is a schematic diagram of logic judgment of simplified rules of a point cloud data simplification method for power line towers according to the present invention.

[0051] FIG3 is a schematic diagram of the overall structure of a point cloud data simplification system for power line towers according to the present invention. DETAILED DESCRIPTION

[0052] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0054] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0055] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0056] In the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0057] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0058] Example 1

[0059] 1 and 2 , which are the first embodiment of the present invention, provide a method for simplifying point cloud data for a power line tower, comprising the following steps:

[0060] S1: Collect point cloud data of power line towers and perform data processing.

[0061] Specifically, the point cloud data of the power line tower is collected by using a lidar device. The lidar device sends a laser beam to the power line tower. When the laser beam collides with the target surface, the laser beam is reflected back to the lidar device. By measuring the time difference and angle information of the reflected laser beam, the point cloud data of the power line tower can be completed.

[0062] Furthermore, the time difference of the reflected laser beam is measured by using the propagation speed of light c and the time difference t of the laser beam to calculate the distance from the laser radar device to the target point. The specific calculation formula is as follows:

[0063] Where c represents the propagation speed of the laser beam in the air, t represents the time difference between the round trip of the laser beam, and d represents the distance measured from the lidar device to the target point;

[0064] The angle information is the horizontal angle θ of the laser radar device. X and the vertical angle θ Y , and use the distance d measured from the lidar device to the target to convert it into the three-dimensional coordinates of the point cloud data. The specific conversion formula is as follows:

[0065] x=d×cos(θX )×cos(θ Y )

[0066] y=d×sin(θ X )×cos(θ Y )

[0067] z=d×sin(θ Y )

[0068] Among them, θ X Represents the horizontal angle of the lidar device, θ Y It represents the vertical angle of the LiDAR device, d represents the distance from the LiDAR device to the target point, and x, y, and z represent the horizontal, vertical, and height coordinates of the point cloud data in the LiDAR device coordinate system, respectively.

[0069] Furthermore, the data processing involves processing the collected point cloud data, including noise suppression and data cleaning, to improve the efficiency of point cloud data streamlining and reduce interference from irrelevant data. Noise suppression is achieved by calculating the statistical characteristics of the area of ​​the point cloud data center and removing points that are inconsistent with the statistical characteristics of surrounding points. The specific implementation is as follows:

[0070] For each point cloud data's three-dimensional coordinate (x i ,y i ,z i ), calculate the mean value of the points in the point cloud data domain and standard deviation The specific calculation formula is as follows:

[0071] Where N represents the number of domain points, z j represents the height coordinate of the jth point in the domain, represents the mean height of the points in the area of ​​point i, represents the standard deviation of the heights of points within the area of ​​point i;

[0072] When the height coordinate, height mean and height difference of point i satisfy the formula When , it means that point i is noise and needs to be removed to improve the quality of point cloud data;

[0073] When the height coordinate, height mean and height standard deviation of point i satisfy the formula When , it means that point i is not noise, and the data of point i is retained.

[0074] Specifically, the data cleaning is to calculate the distance between the midpoint of the point cloud data and the adjacent points, and perform a cleaning and removal operation on the distant adjacent points. The specific implementation of calculating the distance between the midpoint of the point cloud data and the adjacent points is as follows:

[0075] For each point i’s three-dimensional coordinates (x i ,y i ,z i ), calculate the value of the adjacent point (x j ,y j ,z j ) between the Euclidean distance d ij , the specific calculation formula is as follows:

[0076] Among them, x i 、y i 、z i Represents the horizontal coordinate, vertical coordinate and height coordinate of point i, x j 、y j 、z j Respectively represent the horizontal coordinate, vertical coordinate and height coordinate of point j adjacent to point i;

[0077] When the calculated Euclidean distance satisfies the formula When , it represents the neighboring point (x j ,y j ,z j ) is not abnormal data, so keep (x j ,y j ,z j ) data;

[0078] When the calculated Euclidean distance satisfies the formula When , it represents the neighboring point (x j ,y j ,z j ) is abnormal data, remove (x j ,y j ,z j ) data.

[0079] S2: Extract data features from the processed data.

[0080] Specifically, the data feature extraction of the processed data is performed by converting the point cloud data into a three-dimensional voxel network, using a convolutional neural network to extract global feature information, and selecting features representing the structure and spatial information of the power line tower from the extracted global feature information.

[0081] Furthermore, the point cloud data is converted into a three-dimensional voxel network by converting the collected three-dimensional coordinates (xi ,y i ,z i ) is divided into a three-dimensional network, and each network unit represents a voxel, and each voxel represents the information in the point cloud data. The specific implementation is as follows:

[0082] The three-dimensional coordinates of the point cloud data (x i ,y i ,z i ) is divided into a voxel network of size L×W×H, then the three-dimensional coordinates of the point cloud data (x i ,y i ,z i ) is mapped to the coordinates of (x′ i ,y′ i ,z′ i ), the specific mapping formula is as follows:

[0083] Among them, x i 、y i 、z i They represent the horizontal coordinate, vertical coordinate and height coordinate of point i respectively, and L, W and H represent the pixel size of the voxel network respectively.

[0084] Specifically, the extraction of global feature information using a convolutional neural network is to process point cloud data using a deep learning network, including performing feature mapping on the point cloud data using a multi-layer perceptron and aggregating the mapped feature vectors;

[0085] The feature mapping of point cloud data using a multilayer perceptron is to convert the coordinates of the point cloud data into feature vectors in a high-dimensional feature space. The specific conversion formula is as follows:

[0086] T(x,y,z)=f(W1·x,W2·y,W3·z)

[0087] f(x)=max(0,x)

[0088] Where W represents the weight value corresponding to the neuron in the multilayer perceptron, T(x,y,z) represents the transformed feature vector, and f(x) represents the correction function to ensure that the transformed coordinates are positive and enhance the network's expressive power.

[0089] The mapped feature vectors are aggregated by using maximum pooling to extract the global feature vector representation from the local feature vectors. The maximum pooling selects the maximum value as the representative in the local feature vector to capture the most important information in the local feature vector and then aggregates it into a global feature vector. The specific implementation is as follows:

[0090] Among them, T i represents the local eigenvector, represents the average value of all local eigenvectors, T out The global feature vector after aggregation is represented by selecting the maximum value in the local vector as the global feature.

[0091] S3: Classify point cloud data based on the extracted features.

[0092] Specifically, the classification of point cloud data based on the extracted features is to classify the point cloud data through a deep learning algorithm. The deep learning algorithm inputs the aggregated feature vector into a fully connected layer, and maps the aggregated features to the scores of each category through the fully connected layer, and converts the scores into probability distributions of different categories through a softmax activation function, thereby realizing the classification of point cloud data.

[0093] Furthermore, the mapping of the aggregated features to the scores of each category through the fully connected layer is to use the weight matrix and bias vector in the neurons of the fully connected layer to establish a mapping relationship between the input feature vector and the output score during the data training process. The specific implementation formula is as follows:

[0094] T′ out =W′·T out +b

[0095] Among them, W' represents the weight matrix in the neuron. When the dimension of the input feature vector is D and the number of output neurons is C, the weight matrix satisfies the formula W'=D×C. b represents the bias of the output neuron. When the number of output neurons is C', the dimension of the bias vector is also C'. T out Represents the feature vector of the input fully connected layer, which is the global feature vector after aggregation, T′ out Represents the output vector of the fully connected layer, which represents the score of each category.

[0096] Furthermore, the conversion of scores into probability distributions of different categories through the softmax activation function is to use the softmax function to convert different elements in the output vector of the fully connected layer into probabilities of corresponding categories. The specific implementation is as follows:

[0097] Among them, (T′ out ) i Represents the score of the category corresponding to the i-th element in the output vector of the fully connected layer, N represents the total number of categories, Softmax(T′ out ) i Represents the probability of the category corresponding to the i-th element in the output vector.

[0098] S4: Use the point cloud simplification algorithm to simplify the classified point cloud data.

[0099] Specifically, the classified point cloud data is simplified by using a point cloud simplification algorithm and the simplification of the point cloud data is achieved through a triangulated network algorithm. The triangulated network algorithm constructs a triangulated network model using the coordinates of the point cloud data, calculates the triangle cost metric in the triangulated network model, and formulates simplification rules based on the triangle cost metric to achieve the simplification of the point cloud data.

[0100] Furthermore, the use of point cloud data coordinates to construct a triangulated network model is to convert the three-dimensional coordinates of the point cloud data (x i ,y i ,z i ) is constructed into a triangle t(x t ,y t ,z t ) to complete the construction of the triangular network model. The triangle cost metric in the triangular network model is calculated based on the homogeneous coordinates corresponding to the constructed triangles. The specific calculation formula is as follows:

[0101] E t =t′Qt′

[0102] t′=(x t ,y t ,z t ,1)

[0103] Among them, E t represents the cost metric of the triangle, t′ is the homogeneous coordinate of triangle t, represents the plane equation of triangle t, Q represents a 4×4 symmetric matrix used to measure the error of the triangle, N i Represents vertex x i The corresponding normal vector is also the normal vector of triangle t.

[0104] Furthermore, the simplification rule formulated according to the triangle cost metric is to select triangles to be merged according to the cost metric, thereby simplifying the point cloud data. The specific simplification rules are as follows:

[0105] For each triangle t, the cost metric E is calculated t , select the triangle t with the minimum cost metric from all triangles min , and the triangle t with the minimum cost metric min Merge adjacent triangles that share two vertices. For the merged new triangle, calculate the cost metric of the new triangle. If the cost metric of the new triangle is greater than the smallest triangle t before the merger, minThe cost metric of the new triangle is t, and the merging operation is continued until the cost metric of the new triangle is less than the smallest triangle t before the merger. min The simplification of point cloud data is completed until the cost metric of is reached.

[0106] Example 2

[0107] 3 , which is a second embodiment of the present invention, provides a point cloud data simplification system for power line towers, including a data acquisition and processing module, a feature extraction module, a data classification module, and a point cloud data simplification module;

[0108] Specifically, the data acquisition and processing module is used to collect and process the point cloud data of the power line tower; the feature extraction module is used to extract features from the processed data; the data classification module is used to classify the point cloud data using the extracted features; and the point cloud data simplification module is used to simplify the classified point cloud data.

[0109] Furthermore, the data acquisition and processing module uses laser scanning (LiDAR) technology to collect point cloud data to obtain high-precision three-dimensional information of power line towers, and introduces image processing technology, combined with camera image data, to achieve the fusion of point cloud data and image data, improve the multimodality of the data, and by introducing sensor data (such as GPS, inertial measurement unit IMU), to achieve spatial positioning of point cloud data, improve the accuracy of the data, and thus provide high-quality, multimodal point cloud data, providing a sufficient information basis for the feature extraction module, and through spatial positioning information, providing more accurate scene information for the data classification module; the feature extraction module uses deep learning technology to design the network structure of the convolutional neural network (CNN), extract high-level semantic features from the point cloud data, and combine the shape, color, texture and other information of the point cloud data to construct a multimodal feature extraction network to improve the richness and robustness of the features, while introducing time series information, considering the changing characteristics of the power line towers in different time periods, and increasing the feature extraction of the time dimension, and extracting rich semantic features from the multimodal point cloud data provided by the data acquisition and processing module, and passing the extracted features to the data classification module as the classification task. Input: The data classification module uses a convolutional neural network (CNN) model in deep learning to classify point cloud data, and introduces transfer learning technology to transfer models pre-trained in other fields to the power line tower data classification task to improve the generalization ability of the model. At the same time, considering the classification requirements of different power line tower shapes and states, a multi-task learning model is designed to simultaneously process multiple related tasks. By receiving the features extracted by the feature extraction module, the point cloud data classification task is performed, and the classification results are passed to the point cloud data simplification module as a basis for simplification processing. The point cloud data simplification module, based on the output of the classification module, adopts the aforementioned merging rules to merge triangles of similar or similar categories to achieve point cloud data simplification. At the same time, considering the particularity of point cloud data, a simplification strategy that maintains feature points is introduced to ensure that the simplified point cloud still retains key feature points. An adaptive simplification strategy is introduced to adjust the degree of simplification according to the point cloud density and importance of different areas to ensure that the simplified point cloud data meets application requirements. By receiving the output of the classification module, the area and degree of simplification required are determined, and the point cloud data of the corresponding area is simplified to generate a simplified point cloud model.

[0110] Furthermore, in order to demonstrate the technical effects of the present invention, a multi-faceted data comparison is now conducted between the prior art and the technical features of the present invention, thereby highlighting the beneficial effects of the present invention. The specific comparison results are shown in the following table:

[0111] The comparison table above clearly demonstrates that the present invention offers significant advantages in data acquisition accuracy, multimodal data fusion, automated feature extraction, classification and recognition accuracy, intelligent data simplification, processing efficiency, and automation. Compared to existing technologies, it boasts higher accuracy, diversity, intelligence, and efficiency. These advantages highlight the present invention's significant improvement over existing technologies, bringing a higher-level solution to the field of power line tower point cloud data processing. The present invention achieves efficient, accurate, and intelligent processing of power line tower point cloud data, providing strong support for power line tower operation and maintenance management while also improving the automation and efficiency of data processing.

[0112] Furthermore, if the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0113] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0114] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0115] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment (ie, those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention) may not be described.

[0116] It will be appreciated that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but will, for those of ordinary skill having the benefit of this disclosure, be a routine undertaking of design, fabrication, and production without undue experimentation.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for simplifying point cloud data of a power line tower, characterized in that: The following steps are included: Collect point cloud data of power line towers and perform data processing; Extract data features from the processed data; Classify point cloud data based on the extracted features; The classified point cloud data is simplified using the point cloud simplification algorithm.

2. A method for simplifying point cloud data for a power line tower according to claim 1, characterized in that: The point cloud data of the power line tower is collected by using a laser radar device. The laser radar device sends a laser beam to the power line tower. When the laser beam hits the target surface, the laser beam is reflected back to the laser radar device. By measuring the time difference and angle information of the reflected laser beam, the point cloud data of the power line tower can be completed. The time difference of the reflected laser beam is measured by using the propagation speed of light c and the time difference t of the laser beam to calculate the distance from the laser radar device to the target point. The specific calculation formula is as follows: Where c represents the propagation speed of the laser beam in the air, t represents the time difference of the laser beam's round trip, and d represents the distance from the laser radar device to the target point; The angle information is the horizontal angle θ of the laser radar device. X and the vertical angle θ Y , and use the distance d measured from the laser radar device to the target to convert it into the three-dimensional coordinates of the point cloud data. The specific conversion formula is as follows: x=d×cos(θ X )×cos(θ Y ) y=d×sin(θ X )×cos(θ Y ) z=d×sin(θ Y ) Among them, θ X Represents the horizontal angle of the LiDAR device, θ Y It represents the vertical angle of the LiDAR device, d represents the distance from the LiDAR device to the target point, and x, y, and z represent the horizontal, vertical, and height coordinates of the point cloud data in the LiDAR device coordinate system, respectively.

3. A method for simplifying point cloud data for a power line tower as claimed in claim 2, characterized in that: The data processing is to process the collected point cloud data. Including noise suppression and data cleaning, thereby improving the efficiency of point cloud data streamlining and reducing the interference of irrelevant data. The noise suppression is to remove points that are inconsistent with the statistical characteristics of surrounding points by calculating the field statistical characteristics of the point cloud data center. The specific implementation is as follows: For each point cloud data's three-dimensional coordinate (x i ,y i ,z i ), calculate the mean of the points in the point cloud data domain and standard deviation The specific calculation formula is as follows: Where N represents the number of domain points, z j represents the height coordinate of the jth point in the domain, represents the mean height of points in the area of ​​point i, represents the standard deviation of the heights of points within the domain of point i; When the height coordinate, height mean and height difference of point i satisfy the formula When , it means that point i is noise and needs to be removed to improve the quality of point cloud data; When the height coordinate, height mean and height standard deviation of point i satisfy the formula , it means that point i is not noise, and the data of point i is retained.

4. A method for simplifying point cloud data for a power line tower as claimed in claim 3, characterized in that: The data feature extraction of the processed data is performed by converting the point cloud data into a three-dimensional voxel network, extracting global feature information using a convolutional neural network, and selecting features representing the structure and spatial information of the power line tower from the extracted global feature information; The point cloud data is converted into a three-dimensional voxel network by converting the collected three-dimensional coordinates (x i ,y i ,z i ) is divided into a three-dimensional network, and each network unit represents a voxel, and each voxel represents the information in the point cloud data. The specific implementation is as follows: The three-dimensional coordinates (x i ,y i ,z i ) is divided into a voxel network of size L×W×H, then the three-dimensional coordinates of the point cloud data (x i ,y i ,z i ) is mapped to the coordinates of (x′ i ,y′ i ,z′ i ), the specific mapping formula is as follows: Among them, x i ,y i 、z i They represent the horizontal coordinate, vertical coordinate and height coordinate of point i respectively, and L, W and H represent the pixel size of the voxel network respectively.

5. A method for simplifying point cloud data for a power line tower as claimed in claim 4, characterized in that: The extraction of global feature information using a convolutional neural network is to process point cloud data using a deep learning network, including using a multi-layer perceptron to perform feature mapping on point cloud data and aggregating the mapped feature vectors; the feature mapping of point cloud data using a multi-layer perceptron is to convert the coordinates of the point cloud data into feature vectors in a high-dimensional feature space. The specific conversion formula is as follows: T(x,y,z)=f(W1·x,W2·y,W3·z) f(x)=max(0,x) Where W represents the weight value corresponding to the neuron in the multilayer perceptron, T(x,y,z) represents the transformed feature vector, and f(x) represents the correction function to ensure that the transformed coordinates are positive values ​​and enhance the expression ability of the network. force; The mapping feature vectors are aggregated by extracting the global feature vector representation from the local feature vectors using maximum pooling. The maximum pooling selects the maximum value as a representative in the local feature vectors to capture the most important information in the local feature vectors and then aggregates them into the global feature vectors. The specific implementation is as follows: Among them, T i represents the local eigenvector, represents the average value of all local eigenvectors, T out The global feature vector after aggregation is represented by selecting the maximum value in the local vector as the global feature.

6. A method for simplifying point cloud data for a power line tower according to claim 5, characterized in that: The classification of point cloud data based on the extracted features is to classify the point cloud data through a deep learning algorithm, wherein the deep learning algorithm inputs the aggregated feature vector into a fully connected layer, maps the aggregated features to the scores of each category through the fully connected layer, and converts the scores into probability distributions of different categories through a softmax activation function, thereby realizing the classification of point cloud data; The conversion of scores into probability distributions of different categories through the softmax activation function is to use the softmax function to convert different elements in the output vector of the fully connected layer into probabilities of corresponding categories. The specific implementation is as follows: Among them, (T′ out ) i represents the score of the category corresponding to the i-th element in the output vector of the fully connected layer, N represents the total number of categories, and Soft max(T′ out ) i Represents the probability of the category corresponding to the i-th element in the output vector.

7. A method for simplifying point cloud data for a power line tower according to claim 6, characterized in that: The point cloud data after classification is simplified by using a point cloud simplification algorithm, and the simplification of the point cloud data is achieved by using a triangular network algorithm. The triangular network algorithm is to construct a triangular network model by using the coordinates of the point cloud data, calculate the triangle cost metric in the triangular network model, and formulate simplification rules according to the triangle cost metric to achieve the simplification of the point cloud data; The method of using point cloud data coordinates to construct a triangulated network model is to convert the three-dimensional coordinates (x i ,y i ,z i ) to form a triangle t(x t ,y t ,z t ), and then complete the construction of the triangular network model. The triangle cost metric in the triangular network model is calculated according to the homogeneous coordinates corresponding to the constructed triangles. The specific calculation formula is as follows: E t =t′Qt′ t′=(x t ,y t ,z t ,1) Among them, E t represents the cost metric of the triangle, t′ is the homogeneous coordinate of triangle t, represents the plane equation of triangle t, Q represents a 4×4 symmetric matrix used to measure the error of the triangle, N i Represents vertex x i The corresponding normal vector is also the normal vector of triangle t.

8. A system using the point cloud data simplification method for power line towers as claimed in any one of claims 1 to 7, characterized in that: Including data acquisition and processing module, feature extraction module, data classification module and point cloud data simplification module; The data acquisition and processing module is used to collect and process the point cloud data of the power line tower; The feature extraction module is used to extract features from the processed data; The data classification module is used to classify the point cloud data using the extracted features; The point cloud data simplification module is used to simplify the classified point cloud data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images