Ultra-high voltage maintenance operator safety assessment method, device, equipment and medium

Abstract

The present invention relates to the technical field of ultra-high voltage maintenance, and discloses a safety assessment method, device, equipment and storage medium for ultra-high voltage maintenance operators. After removing the outlier noise from the point cloud data set of the ultra-high voltage transmission channel based on an improved k-nearest neighbor search algorithm, a target point cloud data set is obtained through preprocessing based on thinning and registration algorithms; point cloud classification is performed based on progressive morphology and Euclidean clustering to obtain power line point clouds and tower point clouds and select target points therefrom; the target distances between the cylindrical operator and the conductors, ground wires and towers are simulated based on the target points, and compared with the preset safety threshold to determine whether it is safe. The present invention quickly locates in the massive data through outlier noise removal, thinning and registration to reduce the calculation amount and improve the search efficiency. Classification is carried out based on progressive morphological filtering and Euclidean distance clustering to accurately and finely extract the power line and tower point clouds, and the distance judgment accuracy is improved through cylindrical simulation to improve the accuracy of safety judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of ultra-high voltage maintenance, and particularly to a safety assessment method, device, equipment and medium for ultra-high voltage maintenance operators. Background Technique

[0002] Overhead ultra-high voltage transmission lines are an important way for power transmission in the power industry. The safe and stable operation of the transmission lines directly affects the provision of highly reliable and stable electrical energy in the power system. Therefore, regular and effective inspection and maintenance of the transmission lines have become an important task in the power industry. At present, the inspection of overhead ultra-high voltage transmission lines mainly relies on manual operation and mechanical operation. The inspection of the transmission lines must ensure that various distances in the transmission conductors and the line corridors meet the requirements of the safety regulations for operators, especially the safety issues of maintenance personnel in manual operation.

[0003] Traditional safety assessment methods for operators need to obtain assessment results based on technologies such as image recognition, analysis, and operation, involving complex calculations such as ground filtering, DEM construction, ground object classification, and conductor fitting. There is a possibility of data deviation, which is not accurate enough and will affect the accuracy of the judgment results. Summary of the Invention

[0004] The main purpose of the present invention is to provide a safety assessment method, device, equipment and storage medium for ultra-high voltage maintenance operators, aiming to solve the technical problems of potential safety hazards in the existing maintenance process of overhead ultra-high voltage transmission lines and inaccurate existing safety assessment methods with misjudgments.

[0005] To achieve the above purpose, the present invention provides a safety assessment method for ultra-high voltage maintenance operators, including:

[0006] Obtain a point cloud data set of the ultra-high voltage transmission channel, and remove outliers and noise from the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set;

[0007] Preprocess the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set;

[0008] Classify the point cloud of the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain a power line point cloud and a tower point cloud;

[0009] Select target point cloud points according to the power line point cloud and the tower point cloud;

[0010] Simulate the target distances between a cylindrical operator and a target conductor, a target ground wire and a tower based on the target point cloud points;

[0011] Determine whether the operator is safe according to the comparison result between the target distance and the preset safety threshold.

[0012] In some embodiments, the obtaining the point cloud data set of the ultra-high voltage transmission channel and removing the outlier noise from the point cloud data set based on the improved k-nearest neighbor search algorithm to obtain the initial point cloud data set includes:

[0013] Collect the point cloud data set of the ultra-high voltage transmission channel based on LiDAR technology;

[0014] Project the scattered point cloud data of the point cloud data set onto a two-dimensional plane for grid division, and determine whether the scattered point cloud data is a noise point based on the improved k-nearest neighbor search algorithm;

[0015] Obtain the processed point cloud data after removing the noise points;

[0016] Perform secondary processing on the processed point cloud data by the image median filtering method to completely remove the outlier noise points in the scattered point cloud and obtain the initial point cloud data set.

[0017] In some embodiments, the projecting the scattered point cloud data of the point cloud data set onto a two-dimensional plane for grid division and determining whether the scattered point cloud data is a noise point based on the improved k-nearest neighbor search algorithm includes:

[0018] Import the scattered point cloud data of the point cloud data set;

[0019] Project the scattered point cloud data onto a two-dimensional plane to obtain a two-dimensional plane graph;

[0020] Perform grid division on the scattered point cloud data in the two-dimensional plane graph to obtain a plurality of squares;

[0021] Traverse the point cloud points in the scattered point cloud data, and use the point cloud point as the current measurement point;

[0022] Find k nearest neighbor points of the current measurement point based on the improved k-nearest neighbor search algorithm and the square;

[0023] Calculate the average value of the Euclidean distances between the current measurement point and the k nearest neighbor points;

[0024] Compare the average value with a preset threshold, and determine whether the current measurement point is a noise point according to the comparison result.

[0025] In some embodiments, the preprocessing the initial point cloud data set based on the thinning algorithm and the point cloud registration algorithm to obtain the target point cloud data set includes:

[0026] Take the initial point cloud data set as a sample set and set the sampling interval;

[0027] Select a starting data point among the data points of the sample set;

[0028] Based on the thinning algorithm, traverse the data points in the initial point cloud dataset according to the sampling interval and the starting data point to obtain updated data points;

[0029] Construct an updated point cloud dataset according to the updated data points;

[0030] Preprocess the updated point cloud dataset according to the point cloud registration algorithm to obtain a target point cloud dataset.

[0031] In some embodiments, the preprocessing the updated point cloud dataset according to the point cloud registration algorithm to obtain a target point cloud dataset includes:

[0032] Determine two pieces of point cloud from the updated point cloud dataset, and calculate the depth images of the two pieces of point cloud respectively;

[0033] Extract the features of the input depth image through a convolutional neural network, and use the feature difference vector as the input of the regression model to output registration parameters;

[0034] Perform point cloud registration on the updated point cloud dataset through the registration parameters to obtain a target point cloud dataset.

[0035] In some embodiments, the point cloud classification of the target point cloud dataset based on progressive morphological filtering and Euclidean distance clustering to obtain a power line point cloud and a tower point cloud includes:

[0036] Extract single-span point cloud data based on the target point cloud dataset;

[0037] Process the single-span point cloud data according to progressive morphological ground filtering to separate the ground point cloud data and the non-ground point cloud data;

[0038] Extract the power facility point cloud from the non-ground point cloud data according to Euclidean distance clustering and process it to obtain a power line point cloud;

[0039] Extract the tower point cloud according to the ground point cloud data.

[0040] In some embodiments, the simulating the target distances between the cylindrical operator and the target conductor, the target ground wire and the tower based on the target point cloud points includes:

[0041] Simulate the position of the operator according to the target point cloud points;

[0042] Simulate the operator with a cylinder;

[0043] Based on the position, simulate the distances between the operator and the target conductor, the target ground wire, and the tower respectively through the cylinder;

[0044] Obtain the target distance according to the sum of the distance between the operator and the target conductor, the distance between the operator and the target ground wire, and the distance between the operator and the tower.

[0045] In addition, to achieve the above object, the present invention also proposes a safety assessment device for ultra-high voltage maintenance operators, including:

[0046] A point cloud denoising module, configured to obtain a point cloud data set of an ultra-high voltage transmission channel, and perform outlier noise removal on the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set;

[0047] A point cloud processing module, configured to preprocess the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set;

[0048] A point cloud classification module, configured to perform point cloud classification on the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain a power line point cloud and a tower point cloud;

[0049] A point cloud selection module, configured to select target point cloud points according to the power line point cloud and the tower point cloud;

[0050] A distance calculation module, configured to simulate the target distances between the cylindrical operator and the target conductor, the target ground wire, and the tower based on the target point cloud points;

[0051] A safety judgment module, configured to judge whether the operator is safe according to the comparison result between the target distance and a preset safety threshold.

[0052] In addition, to achieve the above object, the present invention also proposes an ultra-high voltage maintenance operator safety assessment device, where the ultra-high voltage maintenance operator safety assessment device includes: a memory, a processor, and an ultra-high voltage maintenance operator safety assessment program stored on the memory and executable on the processor, and the ultra-high voltage maintenance operator safety assessment program is configured to implement the ultra-high voltage maintenance operator safety assessment method as described above.

[0053] In addition, to achieve the above object, the present invention also proposes a storage medium, where the storage medium stores an ultra-high voltage maintenance operator safety assessment program, and the ultra-high voltage maintenance operator safety assessment program is used to cause the processor to implement the ultra-high voltage maintenance operator safety assessment method as described above when executed.

[0054] The present invention obtains a point cloud data set of an extra-high voltage transmission channel, removes outlier noise from the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set; preprocesses the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set; classifies the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain a power line point cloud and a tower point cloud; selects target point cloud points according to the power line point cloud and the tower point cloud; simulates the target distances between a cylindrical operator and a target conductor, a target ground wire, and a tower based on the target point cloud points; and determines whether the operator is safe according to the comparison result between the target distance and a preset safety threshold. In the present invention, through outlier noise removal, thinning, and registration, a certain or certain points can be quickly located in a large amount of point cloud data, the calculation amount is significantly reduced, and the search efficiency is also greatly improved, facilitating subsequent processing. Based on progressive morphological filtering and Euclidean distance clustering, classification is performed using the spatial characteristics of towers and transmission lines to accurately and finely extract the power line and tower point clouds. Furthermore, operation simulation is performed through a cylinder to make the distance calculation more accurate, thereby further improving the distance judgment accuracy, and improving the accuracy of safety judgment, solving the technical problems of potential safety hazards existing in the maintenance process of existing overhead extra-high voltage transmission lines and inaccurate safety assessment methods with misjudgments, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic structural diagram of an extra-high voltage maintenance operator safety assessment device for the hardware operating environment involved in the embodiment scheme of the present invention;

[0056] Figure 2 It is a schematic flowchart of an embodiment of the extra-high voltage maintenance operator safety assessment method of the present invention;

[0057] Figure 3 It is a schematic diagram of the k-nearest neighbor point search area of the extra-high voltage maintenance operator safety assessment method involved in the embodiment of the present invention;

[0058] Figure 4 It is a schematic diagram of the number of points within a grid of the extra-high voltage maintenance operator safety assessment method involved in the embodiment of the present invention;

[0059] Figure 5 It is a schematic flowchart of outlier noise removal of the extra-high voltage maintenance operator safety assessment method involved in the embodiment of the present invention;

[0060] Figure 6 It is a schematic diagram of a CNN-based point cloud registration model of the extra-high voltage maintenance operator safety assessment method involved in the embodiment of the present invention;

[0061] Figure 7Schematic diagram of the CNN structure adopted by the point cloud registration model for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0062] Figure 8 Schematic diagram of the iterative point cloud registration algorithm based on CNN for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0063] Figure 9 Schematic diagram of the classification process of transmission line point cloud data for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0064] Figure 10 Schematic diagram of the suspension analysis for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0065] Figure 11 Schematic diagram of the calculation process of the tower center position for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0066] Figure 12 Schematic diagram of the cross-section of the power line for the safety assessment method of ultra-high voltage maintenance operators in the embodiments of the present invention;

[0067] Figure 13 Structural block diagram of an embodiment of the safety assessment device for ultra-high voltage maintenance operators of the present invention.

[0068] The realization, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. Detailed implementation manners

[0069] 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0070] It should be noted that all the directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0071] In addition, the descriptions involving "first", "second", etc. in the present invention are for descriptive purposes only, and should not be construed as indicating or implying their relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, the technical solutions between various embodiments may be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] Refer to Figure 1 , Figure 1 which is a schematic structural diagram of an ultra-high voltage maintenance operator safety assessment device for the hardware operating environment involved in the embodiment solution of the present invention.

[0073] As Figure 1 shown, the ultra-high voltage maintenance operator safety assessment device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless-fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM memory), or may be a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0074] Those skilled in the art can understand that Figure 1 the structure shown in

[0075] does not constitute a limitation on the ultra-high voltage maintenance operator safety assessment device, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements. Figure 1 As

[0076] In Figure 1 the ultra-high voltage maintenance operator safety assessment device shown, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the ultra-high voltage maintenance operator safety assessment device of the present invention can be arranged in the ultra-high voltage maintenance operator safety assessment device, and the ultra-high voltage maintenance operator safety assessment device calls the ultra-high voltage maintenance operator safety assessment program stored in the memory 1005 through the processor 1001 and executes the ultra-high voltage maintenance operator safety assessment method provided by the embodiments of the present invention.

[0077] The present invention provides a method, device, equipment and storage medium for ultra-high voltage maintenance operator safety assessment.

[0078] An embodiment of the present invention provides a method for ultra-high voltage maintenance operator safety assessment. Referring to Figure 2 , Figure 2 is a schematic flowchart of an embodiment of a method for ultra-high voltage maintenance operator safety assessment of the present invention.

[0079] As Figure 2 shown, the ultra-high voltage maintenance operator safety assessment method includes:

[0080] Step S100: Obtain a point cloud data set of an ultra-high voltage transmission channel, and remove outliers from the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set;

[0081] Step S200: Preprocess the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set;

[0082] Step S300: Classify the point cloud of the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain a power line point cloud and a tower point cloud;

[0083] Step S400: Select target point cloud points according to the power line point cloud and the tower point cloud;

[0084] Step S500: Simulate the target distances between a cylindrical operator and a target conductor, a target ground wire and a tower based on the target point cloud points;

[0085] Step S600: Judge whether the operator is safe according to the comparison result between the target distance and a preset safety threshold.

[0086] It should be noted that the execution subject in this embodiment can be the safety assessment equipment for ultra-high voltage maintenance workers. The safety assessment equipment for ultra-high voltage maintenance workers can be a computer device with data processing functions, or other devices that can achieve the same or similar functions. This embodiment does not limit this. In this embodiment, a computer device is taken as an example for illustration.

[0087] It should be noted that this embodiment takes the method of ultra-high voltage maintenance based on the single-point method as an example for illustration. Specifically, a position is selected to simulate the worker and the distances between the worker and the live body and the grounding body are calculated to judge the safety of the worker.

[0088] It can be understood that the core steps of the ultra-high voltage maintenance worker safety assessment method proposed in this embodiment include: Step 1: Outlier removal; Step 2: Point cloud thinning: Resample the original point cloud data to improve the point cloud processing efficiency without losing key features; Step 3: Point cloud registration: Register multiple segments of point cloud data according to feature points to ensure that the data has no offset; Step 4: Point cloud classification (tower body, conductor, ground wire); Step 5: Operation simulation.

[0089] In one embodiment, a point cloud data set of an ultra-high voltage transmission channel is obtained, and outlier noise removal is performed on the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set, including: collecting the point cloud data set of the ultra-high voltage transmission channel based on LiDAR technology; projecting the scattered point cloud data of the point cloud data set onto a two-dimensional plane for grid division, and judging whether the scattered point cloud data is a noise point based on the improved k-nearest neighbor search algorithm; removing the noise points to obtain processed point cloud data; performing secondary processing on the processed point cloud data by the image median filtering method to completely remove the outlier noise points in the scattered point cloud to obtain an initial point cloud data set.

[0090] Specifically, outlier noise removal is performed based on the improved k-nearest neighbor search algorithm. According to the characteristic that the data attribute of the obtained point is the three-dimensional coordinates of the point, the scattered point cloud is projected onto a two-dimensional plane for grid division, and k nearest neighbor points of the measuring points in the point cloud data are found. In this way, within the rectangle where the point is located or its adjacent rectangles, a total of 9 rectangles can be searched, as Figure 3 and Figure 4 shown. Compared with searching in 27 grids in space, the calculation amount of this embodiment is significantly reduced, and the search efficiency is also greatly improved; the processed point cloud data is processed again by the image median filtering method to completely remove the outlier noise points in the scattered point cloud.

[0091] In one example, the scattered point cloud data of the point cloud data set is projected onto a two-dimensional plane for mesh generation, and whether the scattered point cloud data is noise points is determined based on an improved k-nearest neighbor search algorithm, including: importing the scattered point cloud data of the point cloud data set; projecting the scattered point cloud data onto the two-dimensional plane to obtain a two-dimensional plan view; performing mesh generation on the scattered point cloud data in the two-dimensional plan view to obtain a plurality of squares; traversing the point cloud points in the scattered point cloud data, and using the point cloud point as the current measurement point; finding k nearest neighbor points of the current measurement point based on the improved k-nearest neighbor search algorithm and the square; calculating the average value of the Euclidean distances between the current measurement point and the k nearest neighbor points; comparing the average value with a preset threshold, and determining whether the current measurement point is a noise point according to the comparison result.

[0092] Specifically, referring to Figure 5 the outlier noise removal process shown, the main implementation steps are as follows:

[0093] Step 1: Observe the distribution of outlier noise points, select the top view (x, y) or side views (x, z), (y, z) projections to obtain a two-dimensional plan view;

[0094] Step 2: Perform mesh generation on the scattered point cloud data set in the plane (for example, taking the top view as an example);

[0095] Step 3: Search for k nearest neighbor points of any point p i ∈C, and calculate the average value of the Euclidean distances between the measurement point and the k nearest neighbor points

[0096]

[0097] where K(p i ) is the set of k nearest neighbor points of the measurement point. If there are still not enough k nearest neighbor points in the rectangle where the current measurement point is located and its adjacent rectangles, it can be determined that the current measurement point is an isolated noise point and directly deleted.

[0098] Step 4: Compare the average distance d between the current measurement point p i and its nearest neighbor points with the set threshold D. If then it is considered that the point p i is a non-noise point and is retained. Otherwise, it is determined that p i is a noise point and is deleted.

[0099] In one embodiment, the initial point cloud dataset is preprocessed based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud dataset, including: using the initial point cloud dataset as a sample set and setting a sampling interval; selecting a starting data point from the data points of the sample set; traversing the data points in the initial point cloud dataset based on the thinning algorithm according to the sampling interval and the starting data point to obtain updated data points; constructing an updated point cloud dataset according to the updated data points; and preprocessing the updated point cloud dataset according to the point cloud registration algorithm to obtain a target point cloud dataset.

[0100] Specifically, system thinning is used to compress the point cloud. The input original data points are used as a large sample set, and the sampling interval is set to N. First, 1 data point is randomly selected from the first N data points of the sample set, and then, starting from this data point, the next data point is selected every N data points until all the data points in the original data are traversed. The new data point set composed of the selected data points is the result obtained based on the system thinning algorithm.

[0101] In one embodiment, preprocessing the updated point cloud dataset according to the point cloud registration algorithm to obtain a target point cloud dataset includes: determining two pieces of point cloud from the updated point cloud dataset and respectively calculating the depth images of the two pieces of point cloud; extracting the features of the input depth images through a convolutional neural network, using the feature difference vector as the input of a regression model to output registration parameters; and performing point cloud registration on the updated point cloud dataset through the registration parameters to obtain a target point cloud dataset.

[0102] Exemplarily, in this embodiment, a point cloud registration technique based on a convolutional neural network is used for point cloud registration. Specifically, given two pieces of point cloud P1 and P2, their depth images X1 and X2 are calculated, and taking the depth images as inputs, the model needs to output registration parameters t x 、t y 、t z 、t α 、t β 、t θ , so as to minimize the registration error as much as possible.

[0103] It should be noted that, different from the traditional method, in this embodiment, the convolutional neural network is not used for classification, but for regression. By calculating the features of the input depth images X1 and X2 extracted by the convolutional neural network and using the feature difference vector as the input of the regression model, the registration parameters can be effectively and quickly output.

[0104] Specifically, depth image calculation. Given a point cloud P, where the depth of the point (x, y) at the observation view is D, the pixel value c of this point in the depth image can be expressed as:

[0105]

[0106] Among them, F and N are set to appropriate values so that the calculated depth image has sufficient contrast, which is beneficial for subsequent calculations. By calculating the depth image of the point cloud, the three-dimensional data is projected onto a two-dimensional plane, which is beneficial for the convolutional neural network to process the data.

[0107] Specifically, the network structure of the convolutional neural network. The overall model of the point cloud registration method based on the convolutional neural network proposed in this embodiment is as Figure 6 shown. This model is roughly divided into two parts. First, the convolutional neural network is used to calculate the features of the depth image of the point cloud, and then the obtained features are used to calculate the registration parameters. In order to enable the model to work properly in different registration parameter intervals, the model can adopt 20 groups of parallel convolutional neural networks to independently extract 20 different features of the depth image, and each feature is a 128-dimensional vector. Subtract the first feature vector of image X1 from the first feature vector of image X2 to obtain the differential vector d1 of the first feature. The 20 different differential vectors d1, d2,..., d 20 are connected in parallel as the overall differential vector for registration. The overall differential vector passes through a fully connected layer F2 with ReLU as the activation function to obtain a 256-dimensional vector. Finally, 6 different registration parameters are obtained through the fully connected layer F3. By training the entire network end-to-end, each part of the network can be coordinated so that the network can effectively register within various registration parameter ranges.

[0108] Figure 7 Each group of convolutional neural network structures shown. The input is a 64×64 depth image, which passes through a convolutional layer with a convolutional kernel size of 5×5, a stride of 1, and 20 convolutional kernels to obtain 20 groups of 60×60 images. After passing through a max-pooling layer with a window size of 2×2 and a stride of 2, 20 groups of 30×30 images are obtained. Then, it passes through a convolutional layer with a convolutional kernel size of 5×5, a stride of 1, and 20 convolutional kernels to obtain 20 groups of 26×26 images. The obtained images pass through a max-pooling layer with a window size of 2×2 and a stride of 2 to obtain 20 groups of 13×13 images. The finally output images pass through a fully connected layer to obtain 128-dimensional feature vectors. Due to the particularity of the regression problem, a larger receptive field is required to extract the overall features. In this embodiment, it is found through experiments that using a convolutional kernel with a size of 5×5 has a better effect, rather than using small convolutional kernels that have been more popular in recent years.

[0109] The optimization objective of the entire network is to minimize the mean square loss with the true registration parameters, which is defined as:

[0110]

[0111] where M is the number of training samples, and y i is the true registration parameter of the i-th group of samples, X i is the i-th group of input depth image pairs, and θ is the parameter to be trained in the model. The model initializes the network parameters using the Xavier method to improve the training efficiency of the network. Training uses the method of stochastic gradient descent for parameter optimization. Among them, the batch size of gradient update is set to 100, the momentum of gradient update is m = 0.9, and the weight decay rate d = 0.0001. The update formula for the parameter θ can be expressed as:

[0112]

[0113] θ i+1 = θ i ± μ i+1

[0114] η i = 0.002(1 + 0.0001i) -0.75

[0115] where i is the number of gradient updates, μ is the momentum of the gradient, and η i is the learning rate at the i-th gradient update, is the partial derivative of the objective function with respect to the network parameters at time i calculated based on the backpropagation algorithm. The learning rate decays with training according to the above formula of η i to make the network training stable.

[0116] Specifically, the iterative point cloud registration method. It should be noted that due to the complexity of the spatial position transformation of the point cloud, it is difficult to effectively estimate the registration parameters accurately at one time. In the experiment, the in-plane parameters only involve the translation and in-plane rotation of the point cloud and do not change the shape of the image, so they are relatively easy to estimate. While the out-of-plane parameters will cause changes in the object contour and are more difficult to estimate. Therefore, this embodiment adopts iterative point cloud registration, and divides the parameters to be registered into three groups, namely t x 、t y 、t θ 、t α 、t β 、t z 。First, register the registration parameters that are easy to estimate, and then register the registration parameters that are more difficult to estimate, making the training easier to carry out. Exemplarily, first obtain the registration parameters t x 、t y 、t θ through the registration network model, use the obtained parameters to register the point cloud, and capture the registered depth image. Then, obtain the registration parameters t α 、t β through the registration network model again, and use t α 、tβ Register the point cloud to obtain a new depth image. Finally, obtain t through the registration network z , register the point cloud again, and update the depth image. Repeat the above steps k times, where k is a constant, usually set to within 10, so that the registration error is acceptable. The specific algorithm process is as Figure 8 shown.

[0117] In one embodiment, perform point cloud classification on the target point cloud dataset based on progressive morphological filtering and Euclidean distance clustering to obtain power line point cloud and tower point cloud, including: extracting single-span point cloud data based on the target point cloud dataset; processing the single-span point cloud data according to progressive morphological ground filtering to separate ground point cloud data and non-ground point cloud data; extracting and processing power facility point cloud from the non-ground point cloud data according to Euclidean distance clustering to obtain power line point cloud; extracting tower point cloud according to the ground point cloud data.

[0118] Specifically, classification of transmission line and tower point cloud. Point cloud data is an unstructured true three-dimensional data. For transmission line corridor point cloud data, the method of segmentation first and then recognition can be used for classification by utilizing the spatial characteristics of towers and transmission lines. The point cloud classification operation of this embodiment is as Figure 9 shown, mainly including three steps: ① Extract power facility point cloud by progressive morphological filtering and Euclidean clustering; ② Calculate the tower center coordinates by voxel downsampling and suspension analysis; ③ Fine extraction of power line and tower point cloud.

[0119] It can be understood that the purpose of transmission line point cloud classification is to extract power line and tower point cloud. In transmission line corridor point cloud data, non-target point clouds such as ground and vegetation account for the majority. Therefore, in this embodiment, the progressive morphological ground filtering algorithm is first used to separate the ground point cloud, and then the Euclidean clustering algorithm is used to extract the power facility point cloud to prepare for subsequent power line and tower extraction.

[0120] Specifically, calculation of tower center coordinates. The purpose of calculating the tower center coordinates is: ① Coarsely separate the power line and tower point cloud according to the tower center coordinates; ② Calculate the line direction L, and rotate the line data around the z-axis so that the line direction is parallel to the x-axis.

[0121] Specifically, voxel downsampling. When performing suspension analysis, it is necessary to traverse each point for analysis, and the density of transmission line point cloud data is relatively high, resulting in a large amount of calculation and a very high operation time consumption. Therefore, in this embodiment, voxel downsampling is performed on the power facility point cloud cloud_power. This method will not change the overall characteristics of the point cloud data. The main process is: regularize cloud_power into a three-dimensional grid {g i} according to a predetermined voxel size, and calculate each g iThe centroid point P in c , and use it to replace g i and the other points in

[0122] Specifically, the suspension analysis. The suspension analysis aims to roughly segment the tower point cloud from the downsampled cloud_power to calculate the center coordinates of the tower. This process is based on the spatial distribution characteristics of the power lines and towers: ① In the z-axis direction, the tower point cloud has elevation continuity; ② The power line point cloud has suspension. The specific analysis process is as follows: ① Take a point P as the center, with the tower cross-arm width w cs as the side length, and a predetermined height h c as the height, to create a bounding box B c ; ② Take the upper surface of B c as the bottom surface, and h up as the height, to create a bounding box B up , take the lower surface of B c as the top surface, and h d as the height, to create a bounding box B d ; ③ Query whether there are points in B up and B d . If (‖=true), P∈cloud_pylon_downsampling, else, P∈cloud_line_downsampling. After the query, traverse the next point (as shown in Figure 10 ).

[0123] Specifically, the slicing method is used to calculate the tower center. Using the suspension analysis, the tower point cloud cloud_pylon_downsampling after downsampling can be obtained. At this time, the slicing method is used to calculate the tower center coordinates, with the aims of: ① Rotating the point cloud of a single-span transmission line so that its direction is parallel to the x-axis; ② Roughly separating the towers and power lines from the non-downsampled power facility point cloud cloud_power according to the tower center coordinates. Since there may be missing parts at the upper end of the tower or connected vegetation at the bottom of the tower, the centroid of each layer of the tower point cloud is calculated using the slicing method, and the centroid points with large deviations are deleted, and the average value of the centroid points is retained as the tower center coordinates (the specific algorithm flow is as shown in Figure 11 ).

[0124] Specifically, the power lines and towers are finely extracted. After obtaining the tower center coordinates {Pylon c} through the suspension analysis, with Pylon c as the center of the circle and the cross-arm width w csTaking the diameter to make a circle C parallel to the horizontal plane. If the horizontal projection of a certain point p falls within C, then this point is classified as the tower cloud_pylon_rough; otherwise, it is classified as the power line cloud_line_rough.

[0125] After the rough separation using the tower center coordinates is completed, incomplete power line point clouds can be obtained. In the next step, the spatial curve parameters are calculated using the incomplete power line point clouds, and then the parameters are used for model growth to finely extract the power lines and towers.

[0126] Specifically, the power lines are finely extracted based on the fitting of a straight line and a parabola. Under natural conditions, a power line can be regarded as a catenary suspended at both ends on the towers. It can be regarded as a straight line model on the horizontal plane and conforms to the catenary model on the vertical plane. Since the catenary equation is relatively complex, it can be approximately represented by a parabola. Therefore, the fitting process can be divided into two steps: ① Straight line fitting; ② Parabola fitting. The connection line L between the tower center coordinates obtained during the calculation of the tower center coordinates is used as the direction of a single-span transmission line. Calculate the angle θ between L and the x-axis, and then rotate the point cloud data around the z-axis by θ to make the direction of the point cloud parallel to the x-axis. The Euclidean distance clustering algorithm is used to divide the power line data into strands, so that a single power line is clustered into one strand. Project a single power line onto the o-xy plane, and use the least squares fitting to obtain the straight line parameters for the straight line equation y = kx + b. Then, project this power line onto the o-xz plane, and use the least squares fitting to obtain the curve parameters for the parabola model z = Ax 2 + Bx + C (k * , b * , A * , B * , C * ).

[0127] It can be understood that after obtaining the straight line and parabola model parameters (k * , b * , A * , B * , C * ) of a single power line, model growth is performed on the roughly separated tower point cloud cloud_pylon_rough. The distance between each point and the power line model that conforms is judged. If it is less than the predetermined threshold, it is classified as a power line point; otherwise, it is a tower point. This operation is performed for each power line to finely separate the power line point cloud from cloud_pylon_rough.

[0128] Specifically, the ground wire and the conductor are marked. The power lines of the high-voltage transmission line are divided into two ground wires and several transmission conductors. For general high-voltage lines, there are two main differences between the ground wire and the conductor: ① The conductor is a multi-split wire. The 500kV high-voltage line usually has 4 split conductors, and the single ground wire has no split wire; ② The ground wire is located at the top of the line, and the conductor is located below the line. Based on these two differences, this embodiment proposes a marking method based on elevation features and residuals with the spatial parameter model (such as Figure 12 shown).

[0129] Specifically, the marking process is:

[0130] (1) Cut off a part of the power line in the middle of the line and divide it into clusters {PL i}, i = 1, 2…n, n is the number of shares;

[0131] (2) From PL i Calculate each point P j (x j ,y j , z j ) in the straight line model The residual error Δy in the y direction j , and the parabolic model The residual error Δz in the z direction j , using the formula Calculate point P j The residual v of the power line model j ;

[0132] (3) Using the following formula:

[0133]

[0134] According to the residual {v j}Calculate the PL of a single power line i The mean error δ i After the calculation of n electric lines is completed, δ i The two smallest power lines are ground wires, and the others are marked as conductors.

[0135] Specifically, the tower is finely extracted. The plane coordinates of the tower center obtained in the process of calculating the tower center using the above slicing method are Pylon cTaking (x, y) as the center, a rectangle R parallel to the horizontal plane is made with the width of the tower base as the side length. The points in cloud_ground whose horizontal projections fall within R are classified as candidate point clouds cloud_candidate. The cloud_candidate is sliced and stratified with a layer height of h. Since there is a large gap between the point cloud densities of the ground and the tower, in this embodiment, the point cloud densities of each layer are counted, and the layer with the largest density is classified as the ground points, and the other points are classified as the tower points. Thus, the entire classification process is completed, and the classified tower body point cloud, conductor point cloud, and ground wire point cloud are obtained.

[0136] In one embodiment, simulating the target distances between a cylindrical operator and a target conductor, a target ground wire, and a tower based on the target point cloud points includes: simulating the position of the operator according to the target point cloud points; simulating the operator with a cylinder; simulating the distances between the operator and the target conductor, the target ground wire, and the tower respectively through the cylinder based on the position; and obtaining the target distance according to the sum of the distances between the operator and the target conductor, the distance between the operator and the target ground wire, and the distance between the operator and the tower.

[0137] Specifically, after determining the point cloud classification results (tower body point cloud, conductor point cloud, and ground wire point cloud), a point cloud point is selected at an arbitrary position in the point cloud, and the position where the operator is located is simulated according to the point cloud point, and the distances from the simulated cylinder to the above three types of point clouds are calculated. Using a cylinder to simulate the human body and calculating the distances to the conductor, the ground wire, and the tower, the specific method is as follows:

[0138] To find the shortest distance from a point on a cylinder (simulating the human body) to a set of discrete points, the following steps can be used:

[0139] 1. Cylinder parameters: Obtain the parameters of the cylinder, including the coordinates of the center point of the bottom surface (C x , C y , C z ), radius R, and height H.

[0140] 2) Point cloud data: Obtain the point cloud data (tower body point cloud or conductor point cloud or ground wire point cloud), that is, a set of coordinate sets of points. These points represent the positions of discrete points, and the points on the cylinder with the shortest distance to these points need to be found.

[0141] 3) Calculate the shortest distance: For each point, perform the following steps:

[0142] a. Coordinate transformation: Transform the coordinates of the point from the global coordinate system to the cylinder coordinate system. This can be achieved by subtracting the coordinates of the center point of the cylinder bottom surface from the point coordinates. The coordinates of the point in the cylinder coordinate system are obtained (P x ', P y ', P z ').

[0143] b. Horizontal distance calculation: Calculate the coordinates of the horizontal projection point of the point on the bottom surface of the cylinder (P x ', P y ', 0), where the Z coordinate is 0. The horizontal distance is equal to the two-dimensional Euclidean distance from the point to the projection point, that is

[0144]

[0145] c. Vertical distance calculation: Calculate the vertical distance from the point to the horizontal projection point on the bottom surface of the cylinder. This can be achieved by calculating the Euclidean distance between the two points, that is

[0146] H′ = abs(P z ' - C z ')

[0147] d. Shortest distance point calculation: The shortest distance point is the projection point of the point on the surface of the cylinder, which can be calculated by adding the horizontal distance R and the vertical distance to the coordinates of the center point of the bottom surface of the cylinder. The coordinates of the shortest distance point are (C x , C y , 0) + (R, 0, H′).

[0148] 4) Find the minimum distance point: In step 3) for calculating the shortest distance, the shortest distance from each point to the cylinder is calculated. Finally, find the point with the minimum distance among these calculation results. This point is the point on the cylinder closest to the point in the point cloud, and also the point on the cylinder closest to the point cloud.

[0149] It should be noted that the above calculation for simulating the distance from the cylinder to the three types of point clouds can effectively find the point on the cylinder with the minimum distance from a set of points, regardless of how these points are distributed. After obtaining the distances from the cylinder to the conductor, the ground wire, and the tower, combine the distance from the cylinder to the conductor (the first shortest distance), the distance from the cylinder to the ground wire (the second shortest distance), and the distance from the cylinder to the tower (the third shortest distance), that is, sum the first shortest distance, the second shortest distance, and the third shortest distance to obtain the target distance. By comparing with the threshold value, it can be determined whether the operator is safe, that is, determine whether the operator is safe according to the comparison result between the target distance and the preset safety threshold. Among them, the preset safety threshold can be set according to the actual situation, and this embodiment does not limit it.

[0150] In this embodiment, by obtaining the point cloud data set of the extra-high voltage power transmission channel, removing the outlier noise from the point cloud data set based on the improved k-nearest neighbor search algorithm to obtain the initial point cloud data set; preprocessing the initial point cloud data set based on the thinning algorithm and the point cloud registration algorithm to obtain the target point cloud data set; classifying the point cloud of the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain the power line point cloud and the tower point cloud; selecting the target point cloud points according to the power line point cloud and the tower point cloud; simulating the target distances between the cylindrical operator and the target conductor, the target ground wire and the tower based on the target point cloud points; and judging whether the operator is safe according to the comparison result between the target distance and the preset safety threshold. In this embodiment, through outlier noise removal, thinning, and registration, a certain or certain points can be quickly located in the massive point cloud data, the calculation amount is significantly reduced, and the search efficiency is also greatly improved, which is convenient for subsequent processing. Based on progressive morphological filtering and Euclidean distance clustering, classification is carried out using the spatial characteristics of the tower and the transmission line to accurately and finely extract the power line and tower point cloud, and then the operation simulation is carried out through the cylinder to make the distance calculation more accurate, thereby further improving the distance judgment accuracy, so as to improve the accuracy of safety judgment, and solve the technical problems that there are potential safety hazards in the existing maintenance process of overhead extra-high voltage power transmission lines, and the existing safety assessment methods have low accuracy and misjudgment.

[0151] In addition, an embodiment of the present invention further provides a storage medium, on which an extra-high voltage maintenance operator safety assessment program is stored. When the extra-high voltage maintenance operator safety assessment program is executed by a processor, the steps of the extra-high voltage maintenance operator safety assessment method described above are implemented.

[0152] Refer to Figure 13 , Figure 13 which is the structural block diagram of an embodiment of the extra-high voltage maintenance operator safety assessment device of the present invention.

[0153] As Figure 13 shown, the extra-high voltage maintenance operator safety assessment device includes:

[0154] The point cloud denoising module 10 is used to obtain the point cloud data set of the extra-high voltage power transmission channel, and remove the outlier noise from the point cloud data set based on the improved k-nearest neighbor search algorithm to obtain the initial point cloud data set;

[0155] The point cloud processing module 20 is used to preprocess the initial point cloud data set based on the thinning algorithm and the point cloud registration algorithm to obtain the target point cloud data set;

[0156] The point cloud classification module 30 is used to classify the point cloud of the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain the power line point cloud and the tower point cloud;

[0157] A point cloud selection module 40, configured to select target point cloud points according to the power line point cloud and the tower point cloud;

[0158] A distance calculation module 50, configured to simulate the target distances between a cylindrical operator and a target conductor, a target ground wire, and a tower based on the target point cloud points;

[0159] A safety judgment module 60, configured to judge whether the operator is safe according to the comparison result between the target distance and a preset safety threshold.

[0160] This embodiment provides a safety assessment device for ultra-high voltage maintenance operators. In this embodiment, by removing outlier noise, thinning, and registration, certain points or some points can be quickly located in a large amount of point cloud data, the calculation amount is significantly reduced, and the search efficiency is also greatly improved, which is convenient for subsequent processing. Based on progressive morphological filtering and Euclidean distance clustering, the spatial characteristics of the tower and the transmission line are used for classification to accurately and finely extract the power line and tower point cloud. Then, the operation simulation is carried out through a cylinder to make the distance calculation more accurate, thereby further improving the distance judgment accuracy, and thus improving the accuracy of safety judgment, solving the technical problems of potential safety hazards existing in the maintenance process of existing overhead ultra-high voltage transmission lines and inaccurate safety assessment methods with misjudgment.

[0161] In addition, for the technical details not described in detail in this embodiment of the safety assessment device for ultra-high voltage maintenance operators, reference can be made to the ultra-high voltage maintenance operator safety assessment method provided in any embodiment of the present invention as described above, which will not be elaborated here.

[0162] It should be understood that the above is only an example for illustration and does not constitute any limitation to the technical solution of the present invention. In specific applications, those skilled in the art can set according to needs, and the present invention does not limit this.

[0163] It should be noted that the above-described work process is only illustrative and does not limit the protection scope of the present invention. In actual applications, those skilled in the art can select some or all of them according to actual needs to achieve the purpose of the solution of this embodiment, and this is not limited here.

[0164] In addition, it should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or system. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or system including that element.

[0165] The serial numbers of the above-described embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments.

[0166] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as a read-only memory (ROM) / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0167] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied to other related technical fields, is equally included in the patent protection scope of the present invention.

Claims

1. A safety assessment method for ultra-high voltage maintenance operators, characterized in that, Adopt a method of ultra-high voltage maintenance based on the single-point method. The safety assessment method for ultra-high voltage maintenance personnel includes: Obtain the point cloud data set of the ultra-high voltage transmission channel, and remove the outlier noise from the point cloud data set based on the improved k-nearest neighbor search algorithm to obtain the initial point cloud data set; Preprocess the initial point cloud data set based on the downsampling algorithm and the point cloud registration algorithm to obtain the target point cloud data set; Classify the point cloud of the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain the power line point cloud and the tower point cloud; Select the target point cloud points according to the power line point cloud and the tower point cloud; Simulate the target distances between the cylindrical operator and the target conductor, the target ground wire and the tower based on the target point cloud points; Judge whether the operator is safe according to the comparison result between the target distance and the preset safety threshold; Among them, simulating the target distances between the cylindrical operator and the target conductor, the target ground wire and the tower based on the target point cloud points includes: simulating the position of the operator according to the target point cloud points; simulating the operator with a cylinder; based on the position, simulating the distances between the operator and the target conductor, the target ground wire and the tower respectively through the cylinder; obtaining the target distance according to the sum of the distance between the operator and the target conductor, the distance between the operator and the target ground wire, and the distance between the operator and the tower; Calculate the distances from the simulated cylinder to the three types of point clouds of the conductor, the ground wire and the tower, and find the point on the cylinder with the minimum distance from a group of points: 1) Cylinder parameters: Obtain the parameters of the cylinder, including the coordinates of the center point of the bottom surface (C x , C y , C z ), radius R, and height H; 2) Point cloud data: Obtain the point cloud data (tower body point cloud, or conductor point cloud, or ground wire point cloud), that is, a set of coordinates of points, and these points represent the positions of discrete points. It is necessary to find the points on the cylinder with the shortest distance to these points; 3) Calculate the shortest distance: For each point, perform the following steps: a. Coordinate transformation: Convert the coordinates of a point from the global coordinate system to the cylindrical coordinate system. This is achieved by subtracting the coordinates of the center point of the cylinder base from the point coordinates, obtaining the coordinates of the point in the cylindrical coordinate system (P x ＇ , P y ＇ , P z ＇ ); b. Horizontal distance calculation: Calculate the coordinates of the horizontal projection point of the point on the bottom surface of the cylinder (P x ＇ , P y ＇ , 0), where the Z coordinate is 0, and the horizontal distance is equal to the two-dimensional Euclidean distance from the point to the projection point, that is ； c. Vertical distance calculation: Calculate the vertical distance from the point to the horizontal projection point on the bottom surface of the cylinder, which is achieved by calculating the Euclidean distance between two points, that is ； d. Shortest distance point calculation: The shortest distance point is the projection point of the point on the surface of the cylinder, which is calculated by adding the horizontal distance D and the vertical distance to the coordinates of the center point of the bottom surface of the cylinder. The coordinates of the shortest distance point are (C x , C y , 0) + (D, 0, H ＇ ); 4) Find the point with the minimum distance: In step 3) of calculating the shortest distance, the shortest distance points from each point to the cylinder are calculated. Finally, find the point with the minimum distance among these calculation results. This point is the point on the cylinder closest to the points in the point cloud, that is, the point on the cylinder closest to the point cloud.

2. The safety assessment method for ultra-high voltage maintenance operators according to claim 1, wherein, The obtaining of the point cloud data set of the ultra-high voltage transmission channel, removing the outlier noise from the point cloud data set based on the improved k-nearest neighbor search algorithm to obtain the initial point cloud data set includes: Collect the point cloud data set of the ultra-high voltage transmission channel based on LiDAR technology; Project the scattered point cloud data of the point cloud data set onto a two-dimensional plane for grid division, and judge whether the scattered point cloud data is a noise point based on the improved k-nearest neighbor search algorithm; Obtain the processed point cloud data after removing the noise points; Perform secondary processing on the processed point cloud data through the image median filtering method to completely remove the outlier noise points in the scattered point cloud to obtain the initial point cloud data set.

3. The safety assessment method for ultra-high voltage maintenance operators according to claim 2, characterized in that Projecting the scattered point cloud data of the point cloud data set onto a two-dimensional plane for grid division, and determining whether the scattered point cloud data is noise points based on an improved k-nearest neighbor search algorithm, including: Importing the scattered point cloud data of the point cloud data set; Projecting the scattered point cloud data onto a two-dimensional plane to obtain a two-dimensional plan; Performing grid division on the scattered point cloud data in the two-dimensional plan to obtain a plurality of squares; Traversing the point cloud points in the scattered point cloud data, and taking the point cloud point as the current measurement point; Finding k nearest neighbor points of the current measurement point based on the improved k-nearest neighbor search algorithm and the squares; Calculating the average value of the Euclidean distances between the current measurement point and the k nearest neighbor points; Comparing the average value with a preset threshold, and determining whether the current measurement point is a noise point according to the comparison result.

4. The safety assessment method for ultra-high voltage maintenance operators according to claim 1, wherein Preprocessing the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set, including: Taking the initial point cloud data set as a sample set and setting a sampling interval; Selecting a start data point from the data points of the sample set; Traversing the data points in the initial point cloud data set based on the thinning algorithm according to the sampling interval and the start data point to obtain updated data points; Constructing an updated point cloud data set according to the updated data points; Preprocessing the updated point cloud data set according to the point cloud registration algorithm to obtain a target point cloud data set.

5. The safety assessment method for ultra-high voltage maintenance operators according to claim 4, wherein, Preprocessing the updated point cloud data set according to the point cloud registration algorithm to obtain a target point cloud data set, including: Determining two pieces of point cloud from the updated point cloud data set, and respectively calculating the depth images of the two pieces of point cloud; Extracting the features of the input depth images through a convolutional neural network, and taking the feature difference vector as the input of a regression model to output registration parameters; Performing point cloud registration on the updated point cloud data set through the registration parameters to obtain a target point cloud data set.

6. The safety assessment method for ultra-high voltage maintenance operators according to claim 1, characterized in that Performing point cloud classification on the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain power line point cloud and tower point cloud, including: Extracting single-span point cloud data based on the target point cloud data set; Processing the single-span point cloud data according to progressive morphological ground filtering to separate the ground point cloud data and the non-ground point cloud data; Extracting and processing the power facility point cloud from the non-ground point cloud data according to Euclidean distance clustering to obtain a power line point cloud; Extracting tower point cloud according to the ground point cloud data.

7. A safety assessment device for ultra-high voltage maintenance operators, characterized in that, Adopting a method for ultra-high voltage maintenance based on the single-point method, the ultra-high voltage maintenance operator safety assessment device includes: A point cloud denoising module, configured to obtain a point cloud data set of an ultra-high voltage transmission channel, and perform outlier noise removal on the point cloud data set based on an improved k-nearest neighbor search algorithm to obtain an initial point cloud data set; A point cloud processing module, configured to preprocess the initial point cloud data set based on a thinning algorithm and a point cloud registration algorithm to obtain a target point cloud data set; A point cloud classification module, configured to perform point cloud classification on the target point cloud data set based on progressive morphological filtering and Euclidean distance clustering to obtain power line point cloud and tower point cloud; A point cloud selection module, configured to select target point cloud points according to the power line point cloud and the tower point cloud; A distance calculation module, configured to simulate the target distances between a cylindrical operator and a target conductor, a target ground wire, and a tower based on the target point cloud points; A safety judgment module, configured to judge whether the operator is safe according to the comparison result between the target distance and a preset safety threshold; Among them, simulating the target distances between a cylindrical operator and a target conductor, a target ground wire, and a tower based on the target point cloud points includes: simulating the position of the operator according to the target point cloud points; simulating the operator by a cylinder; based on the position, simulating the distances between the operator and the target conductor, the target ground wire, and the tower respectively by the cylinder; and obtaining the target distance according to the sum of the distance between the operator and the target conductor, the distance between the operator and the target ground wire, and the distance between the operator and the tower; Calculate the distances from the simulated cylinder to the point clouds of the conductor, the ground wire, and the tower, and find the point on the cylinder with the minimum distance from a set of points: 1) Cylinder parameters: Obtain the parameters of the cylinder, including the coordinates of the center point of the bottom surface (C x , C y , C z ), radius R, and height H; 2) Point cloud data: Obtain point cloud data (tower body point cloud, or conductor point cloud, or ground wire point cloud), that is, a set of point coordinate sets, and these points represent the positions of discrete points. It is necessary to find the points on the cylinder with the shortest distance to these points; 3) Calculate the shortest distance: For each point, perform the following steps: a. Coordinate transformation: Convert the coordinates of a point from the global coordinate system to the cylindrical coordinate system. This is achieved by subtracting the coordinates of the center point of the cylinder base from the point coordinates, obtaining the coordinates of the point in the cylindrical coordinate system (P x ＇ , P y ＇ , P z ＇ ); b. Horizontal distance calculation: Calculate the coordinates of the horizontal projection point of the point on the bottom surface of the cylinder (P x ＇ , P y ＇ , 0), where the Z coordinate is 0, and the horizontal distance is equal to the two-dimensional Euclidean distance from the point to the projection point, that is ； c. Vertical distance calculation: Calculate the vertical distance from the point to the horizontal projection point on the bottom surface of the cylinder, which is achieved by calculating the Euclidean distance between two points, that is ； d. Shortest distance point calculation: The shortest distance point is the projection point of the point on the surface of the cylinder, which is calculated by adding the horizontal distance D and the vertical distance to the coordinates of the center point of the bottom surface of the cylinder. The coordinates of the shortest distance point are (C x , C y , 0) + (D, 0, H ＇ ); 4) Find the point with the minimum distance: In step 3) of calculating the shortest distance, the shortest distance points from each point to the cylinder are calculated. Finally, find the point with the minimum distance among these calculation results. This point is the point on the cylinder closest to the points in the point cloud, that is, the point on the cylinder closest to the point cloud.

8. An ultra-high voltage maintenance operator safety assessment device, characterized in that, The ultra-high voltage maintenance operator safety assessment device includes: a memory, a processor, and an ultra-high voltage maintenance operator safety assessment program stored on the memory and executable on the processor. The ultra-high voltage maintenance operator safety assessment program is configured to implement the ultra-high voltage maintenance operator safety assessment method according to any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores an ultra-high voltage maintenance operator safety assessment program, and the ultra-high voltage maintenance operator safety assessment program is used to cause the processor to implement the ultra-high voltage maintenance operator safety assessment method according to any one of claims 1 to 6 when executed.

Images