Telegraph pole area segmentation positioning method and related equipment

Abstract

The invention discloses a telegraph pole area segmentation positioning method and related equipment, and the method comprises the steps: obtaining a to-be-segmented telegraph pole image, and inputting the to-be-segmented telegraph pole image into a trained telegraph pole image segmentation network; the network comprises a down-sampling module, an up-sampling module, a core continuous boundary extraction module CBEM and a global telegraph pole positioning module GUPLM. The CBEM extracts multi-scale local features by connecting a plurality of BEMs in series, and feature fusion is carried out on each BEM after operations such as deep expansion convolution and the like. And the GUPLM is used for splicing the middle feature map and the up-sampling feature map, global and local features are fused through calculation such as depth separable asymmetric convolution and the like, finally, a telegraph pole segmentation result grey-scale map is output, and a telegraph pole region is positioned according to a grey-scale value. According to the method, the problem of edge blurring caused by single traditional feature scale is solved; in complex background segmentation, wrong segmentation is avoided, the segmentation accuracy is improved, the adaptability to complex scenes is greatly enhanced, and the defects in the prior art are effectively overcome.

Description

Technical Field

[0001] The present application relates to the field of image processing, and more specifically, to a method for segmenting and locating a utility pole area and related equipment. Background Art

[0002] When it comes to monitoring and maintaining distribution network infrastructure, the stability and safety of utility poles directly impact power system reliability and public safety. Traditional manual inspections rely on visual inspections by on-site personnel or simple instrument measurements. These methods suffer from low efficiency, limited coverage, and poor real-time performance. This is especially true in the face of frequent natural disasters (such as strong winds, ice, and snow), making it difficult to detect the risk of utility pole tilt or displacement in a timely manner.

[0003] While existing automated monitoring technologies based on image processing can partially replace manual inspections, they still suffer from significant drawbacks due to complex environmental interference and algorithmic performance bottlenecks. First, the accuracy of utility pole image segmentation is insufficient: affected by factors such as lighting variations, weather conditions (such as rain and fog), and background interference (such as obstructions by trees and buildings), traditional segmentation algorithms are prone to edge blurring, loss of detail, or mis-segmentation, leading to cumulative errors in subsequent tilt detection. Second, existing deep learning segmentation models lack the ability to integrate multi-scale features and global contextual information. For example, while U-Net-based architectures can extract local features, they are poor at modeling long-range dependencies between utility poles and have difficulty distinguishing utility poles from similar structures (such as streetlight poles) in complex backgrounds. Furthermore, mainstream segmentation loss functions (such as cross-entropy loss and Dice loss) focus excessively on pixel-level classification accuracy and ignore the optimization of edge structure continuity. This results in jagged or irregular segmentation boundaries, directly affecting the accuracy of tilt angle calculation.

[0004] The above-mentioned defects make it difficult for existing technologies to meet the urgent needs of the power system for high-precision and real-time monitoring. Based on this, the present application provides a pole area segmentation and positioning solution that avoids the defects of existing methods, has strong robustness, retains complete edge details and is adaptable to complex scenarios. Summary of the Invention

[0005] This application provides a method and related equipment for segmenting and positioning a utility pole area, which can more effectively ensure the accuracy and integrity of edge information, and the output segmentation results are more accurate and have smooth edges. At the same time, it fully utilizes the local self-similarity of deep feature maps to further improve positioning accuracy and robustness, and has higher practical application value.

[0006] A method for segmenting and locating a utility pole region, comprising:

[0007] Obtain the telephone pole image to be segmented;

[0008] The utility pole image is input into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM, the continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations, the global utility pole positioning module GUPLM splices the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features;

[0009] The utility pole area is located according to the grayscale value of each pixel in the grayscale image of the utility pole segmentation result.

[0010] Optionally, the training process of the utility pole image segmentation network includes:

[0011] Constructing a utility pole image segmentation dataset, wherein the utility pole image segmentation dataset includes original images and corresponding real label images in single-channel digital form;

[0012] Based on the utility pole image segmentation dataset, the utility pole image segmentation network is trained using an edge similarity loss function, wherein the edge similarity loss function optimizes network parameters by calculating the Laplace filter difference and gradient difference between the segmentation result and the true label image.

[0013] Optionally, the boundary extraction submodule BEM processes the input feature map, including:

[0014] Perform deep dilation convolution on the input feature map to obtain a dilated feature map;

[0015] Performing batch normalization and convolution operations with a convolution kernel of 1 on the expanded feature map in sequence to obtain a normalized feature map;

[0016] After activating and convolving the normalized feature map using a SMU activation function, an activation feature map is generated;

[0017] The activation feature map is element-wise added to the input feature map to obtain a local feature map.

[0018] Optionally, the calculation expression for a fusion extraction of the boundary extraction submodule BEM is:

[0019]

[0020]

[0021] in, is the input feature map, is the activation feature map, is the local feature map, of Represents a convolution operation with a convolution kernel of 1, is the SMU activation function, is batch normalization, The expansion rate is Depth-wise dilated convolution.

[0022] Optionally, the global utility pole positioning module GUPLM processes the input intermediate feature map and the upsampled feature map, including:

[0023] Concatenate the intermediate feature map and the upsampled feature map along the channel dimension, and reduce the dimension by convolution to obtain a reduced-dimensional feature map;

[0024] Performing depth-separable asymmetric convolution of a first preset convolution kernel and a second preset convolution kernel on the dimensionality reduction feature map respectively, adding the results and activating them through SMU to generate a separable convolution feature map;

[0025] Performing SimAM attention mechanism calculation on the separated convolution feature map, and fusing it with the dimensionality reduction feature map by element addition to obtain a local fused feature map;

[0026] The local fused feature map, the intermediate feature map and the up-sampled feature map are fused element-wise to obtain a global fused feature map.

[0027] Optionally, the edge similarity loss function is calculated as follows:

[0028]

[0029] in, The pixel matrix of the output image of the utility pole image segmentation network, is the pixel matrix of the true label image, is the absolute value, is the Laplace filtering operation, is the stability coefficient, for and The gradient difference in the horizontal direction, for and The gradient difference in the vertical direction.

[0030] A device for segmenting and positioning a utility pole area, comprising:

[0031] An image acquisition unit, used for acquiring an image of the electric pole to be segmented;

[0032] A segmentation network unit is used to input the utility pole image into a trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM. The continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, and each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations. The global utility pole positioning module GUPLM combines the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features.

[0033] The area positioning unit is used to locate the electric pole area according to the grayscale value of each pixel in the grayscale image of the electric pole segmentation result.

[0034] A utility pole area segmentation and positioning device, comprising a memory and a processor;

[0035] The memory is used to store programs;

[0036] The processor is used to execute the program to implement each step of the utility pole area segmentation and positioning method as described in any one of the above items.

[0037] A readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, each step of the utility pole area segmentation and positioning method as described in any one of the above items is implemented.

[0038] A computer program product includes a computer program, wherein when the computer program is run by a processor, the computer program executes each step of any of the above-mentioned methods for segmenting and locating utility pole areas.

[0039] It can be seen from the above technical solutions that the method for segmenting and locating a utility pole area and the related equipment provided in the embodiments of the present application have significant beneficial effects in many aspects and can effectively avoid many defects of the existing technology.

[0040] In terms of accurate multi-scale edge extraction, this application connects multiple levels of boundary extraction sub-modules (BEM) in series through a continuous boundary extraction module (CBEM), and combines deep dilated convolution with feature fusion operations. Deep dilated convolution can expand the receptive field of the convolution kernel without adding too many parameters, thereby fully capturing the local detail information of the utility pole. At the same time, through the feature fusion operation, feature maps of different scales are integrated, so that the global contour features of the utility pole can also be effectively preserved. This innovative design significantly improves the edge recognition accuracy in complex backgrounds, such as vegetation occlusion and poor lighting conditions, and successfully solves the edge blurring problem caused by the single feature scale of traditional methods.

[0041] For robust segmentation in complex backgrounds, the Global Pole Localization Module (GUPLM) fully utilizes image information at different resolutions by fusing the CBEM output with multi-level feature maps from the upsampling stage. Furthermore, it introduces depthwise separable asymmetric convolution and the SimAM attention mechanism. The depthwise separable asymmetric convolution effectively captures the directional characteristics of long, rectangular objects like utility poles, while the SimAM attention mechanism enhances the model's focus on the main area of ​​the pole, effectively suppressing background noise interference. This avoids mis-segmentation of poles with similar objects, significantly improving segmentation accuracy in complex backgrounds.

[0042] Furthermore, the utility pole image segmentation network is trained using the edge similarity loss function (ESLoss), which constrains the smoothness of the network output boundaries through gradient differences. Unlike traditional loss functions, which focus on a single optimization objective, the edge similarity loss function not only focuses on pixel-level classification accuracy but also emphasizes the continuity of edge structures. This ensures that the segmentation results more closely match the physical structure of real utility poles.

[0043] Compared with existing technologies, this application can more effectively ensure the accuracy and integrity of edge information, and the output segmentation results are more accurate and have smooth edges. At the same time, the present invention can fully utilize the local self-similarity of deep feature maps, further improving positioning accuracy and robustness, and has higher practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0045] Figure 1 This is a flow chart of a method for segmenting and locating a utility pole area disclosed in an embodiment of the present application;

[0046] Figure 2 A schematic diagram of the structure of a utility pole image segmentation network disclosed in an embodiment of the present application;

[0047] Figure 3 A schematic diagram of a boundary extraction submodule BEM disclosed in an embodiment of the present application;

[0048] Figure 4 This is a structural diagram of a global utility pole positioning module GUPLM disclosed in an embodiment of the present application;

[0049] Figure 5 A schematic diagram of a device for segmenting and locating a utility pole area disclosed in an embodiment of the present application;

[0050] Figure 6 This is a hardware structure block diagram of a utility pole area segmentation and positioning device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0052] The present application can be used in a variety of general or special computing device environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multi-processor devices, and distributed computing environments including any of the above devices or devices.

[0053] Next, we will introduce the application scheme. This application proposes the following technical scheme, please see below for details.

[0054] Figure 1 This is a flowchart of a method for segmenting and locating a utility pole area disclosed in an embodiment of the present application.

[0055] like Figure 1 As shown, the method may include:

[0056] Step S1: Acquire an image of a utility pole to be segmented.

[0057] Specifically, the utility pole images used for subsequent segmentation operations are obtained. These images can come from a variety of sources, such as:

[0058] Drone photography: During power inspections, drones are often used to photograph utility poles from the air, thereby obtaining images of utility poles at different angles and positions.

[0059] Surveillance cameras: Surveillance cameras installed around utility poles can capture images of the utility poles in real time, which can be used as objects for segmentation processing.

[0060] Historical image data: Select images containing utility poles from the existing image database. These images may have been accumulated from past inspections or other related work.

[0061] After acquiring the image, some preprocessing operations may be required, such as adjusting the image size, color correction, noise removal, etc., in order to ensure the image quality and prepare for subsequent segmentation work.

[0062] Step S2: input the utility pole image into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result.

[0063] Specifically, the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM. The continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction sub-modules BEM in series. Each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations. The global utility pole positioning module GUPLM fuses global and local features by splicing the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performing deep separable asymmetric convolution, SMU activation function and SimAM attention calculation.

[0064] Downsampling module: This module consists of a conventional downsampling operation, including a convolutional layer, a batch normalization layer, and an activation function layer. The convolutional layer reduces the spatial resolution of the feature map and increases the number of channels. The batch normalization layer helps promote model convergence, and the activation function layer increases the model's nonlinear expression capabilities.

[0065] Upsampling module: This module consists of conventional upsampling operations, including a transposed convolution layer, a batch normalization layer, and an activation function layer. The transposed convolution layer is used to increase the spatial resolution of the feature map and reduce the number of channels. The batch normalization layer and activation function layer have similar functions to those in the downsampling module.

[0066] Continuous Boundary Extraction Module (CBEM): This extracts multi-scale local features by chaining multiple Boundary Extraction Modules (BEMs). Each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation, and feature fusion, effectively capturing the boundary information of utility poles.

[0067] Global Utility Pole Localization Module (GUPLM): It concatenates the intermediate feature maps output by CBEM and the upsampled feature maps obtained by the upsampling module, and then performs depth-wise separable asymmetric convolution, SMU activation function, and SimAM attention calculation to fuse global and local features, thereby more accurately locating the position of utility poles in the image.

[0068] like Figure 2 As shown in Figure 2, the processing process of the utility pole image segmentation network is as follows:

[0069] Image input: The telephone pole image to be segmented with a spatial resolution of H×W and 3 channels is input into the telephone pole image segmentation network. After a convolution layer with a convolution kernel size of 3×3 and a SMU activation function, a telephone pole feature map with a spatial resolution of H×W and 3 channels is obtained.

[0070] Downsampling and CBEM processing: After three downsampling modules and a continuous boundary extraction module (CBEM), the BEM module is repeatedly executed in the CBEM module each time to obtain utility pole feature maps with spatial resolutions of H / 2×W / 2, H / 4×W / 4, and H / 8×W / 8, and channel numbers of 2C, 4C, and 8C, respectively.

[0071] Intermediate processing: The down-sampled feature map is used as input. After passing through a convolution layer with a convolution kernel size of 3×3 and a SMU activation function, an intermediate feature map of the utility pole with a spatial resolution of H / 8×W / 8 and a channel number of 8C is obtained.

[0072] Global positioning and upsampling processing: The upsampled feature map and the feature map obtained by intermediate processing are used as input, passing through the global pole positioning module GUPLM, and then passing through the continuous boundary extraction module CBEM and the upsampling module in sequence. This is repeated multiple times to finally obtain a pole feature map with a spatial resolution of H×W and a channel number of C.

[0073] Output processing: The intermediate feature map is passed through a convolution layer with a convolution kernel size of 3×3 and a SMU activation function to obtain a grayscale image of the utility pole image segmentation result with a spatial resolution of H×W and a channel number of 1. The grayscale value of each pixel represents the probability or confidence that the pixel belongs to the utility pole.

[0074] Step S3: locating the electric pole area according to the grayscale value of each pixel in the grayscale image of the electric pole segmentation result.

[0075] Specifically, after obtaining the grayscale image of the utility pole segmentation result, the utility pole area can be located according to the grayscale value of each pixel in the image.

[0076] By setting an appropriate grayscale threshold, pixels with grayscale values ​​greater than the threshold are considered to belong to the utility pole, while pixels with grayscale values ​​less than the threshold are considered to belong to the background. This allows the utility pole to be separated from the background, forming a binary image. Connected region analysis is performed on this binary image to identify all interconnected sets of pixels, each representing a possible utility pole region. Based on prior knowledge of the utility pole, such as its area and shape, the connected regions are screened, eliminating those that do not meet the criteria, and ultimately determining the exact location of the utility pole.

[0077] Through the above three steps, the segmentation and positioning of the utility pole image can be achieved, providing strong support for subsequent power inspections, fault detection and other tasks.

[0078] It can be seen from the above technical solutions that the method for segmenting and locating a utility pole area and the related equipment provided in the embodiments of the present application have significant beneficial effects in many aspects and can effectively avoid many defects of the existing technology.

[0079] In terms of accurate multi-scale edge extraction, this application connects multiple levels of boundary extraction sub-modules (BEM) in series through a continuous boundary extraction module (CBEM), and combines deep dilated convolution with feature fusion operations. Deep dilated convolution can expand the receptive field of the convolution kernel without adding too many parameters, thereby fully capturing the local detail information of the utility pole. At the same time, through the feature fusion operation, feature maps of different scales are integrated, so that the global contour features of the utility pole can also be effectively preserved. This innovative design significantly improves the edge recognition accuracy in complex backgrounds, such as vegetation occlusion and poor lighting conditions, and successfully solves the edge blurring problem caused by the single feature scale of traditional methods.

[0080] For robust segmentation in complex backgrounds, the Global Pole Localization Module (GUPLM) fully utilizes image information at different resolutions by fusing the CBEM output with multi-level feature maps from the upsampling stage. Furthermore, it introduces depthwise separable asymmetric convolution and the SimAM attention mechanism. The depthwise separable asymmetric convolution effectively captures the directional characteristics of long, rectangular objects like utility poles, while the SimAM attention mechanism enhances the model's focus on the main area of ​​the pole, effectively suppressing background noise interference. This avoids mis-segmentation of poles with similar objects, significantly improving segmentation accuracy in complex backgrounds.

[0081] Furthermore, the utility pole image segmentation network is trained using the edge similarity loss function (ESLoss), which constrains the smoothness of the network output boundaries through gradient differences. Unlike traditional loss functions, which focus on a single optimization objective, the edge similarity loss function not only focuses on pixel-level classification accuracy but also emphasizes the continuity of edge structures. This ensures that the segmentation results more closely match the physical structure of real utility poles.

[0082] Compared with existing technologies, this application can more effectively ensure the accuracy and integrity of edge information, and the output segmentation results are more accurate and have smooth edges. At the same time, the present invention can fully utilize the local self-similarity of deep feature maps, further improving positioning accuracy and robustness, and has higher practical application value.

[0083] In some embodiments of the present application, the training process of the utility pole image segmentation model is introduced, which may specifically include:

[0084] The first step is to construct a utility pole image segmentation dataset, which contains the original image and the corresponding real label image in single-channel digital form;

[0085] The second step is to train the utility pole image segmentation network based on the utility pole image segmentation dataset using an edge similarity loss function. The edge similarity loss function optimizes the network parameters by calculating the Laplace filter difference and gradient difference between the segmentation result and the true label image.

[0086] Specifically, the training of the utility pole image segmentation model includes dataset construction, loss function design, and network training verification.

[0087] To construct a utility pole image segmentation dataset, we first captured images from videos recording utility poles tilting, then collected images of utility poles in different states to enrich the data diversity. After manually annotating the images, we converted the RGB color labels into single-channel digital labels, cropped them to a uniform size, and paired them to form the dataset. This means that the utility pole image segmentation dataset contains a large number of original images and their corresponding single-channel digitally labeled images.

[0088] In utility pole tilt monitoring, the boundaries of segmentation results are crucial. The edge similarity loss function optimizes network parameters by calculating the Laplace filter difference and gradient difference between the segmentation results and the true labeled image, increasing the network's sensitivity to boundary information and making the segmentation boundaries more accurate and smooth.

[0089] The edge similarity loss function is calculated as follows:

[0090]

[0091] in, The pixel matrix of the output image of the utility pole image segmentation network, is the pixel matrix of the true label image, is the absolute value, is the Laplace filtering operation, is the stability coefficient, for and The gradient difference in the horizontal direction, for and The gradient difference in the vertical direction.

[0092]

[0093]

[0094] in, Represents the horizontal gradient matrix of the calculated image pixel matrix, Represents the vertical gradient matrix of the pixel matrix of the image.

[0095]

[0096]

[0097] in, represents the input image pixel matrix, is the filtering operation used to calculate the gradient, is a horizontal filter that extracts horizontal edges and linear features in the image. It is a vertical filter that extracts vertical edges and linear features in the image.

[0098] In the specific implementation of the present invention, the stability coefficient in the loss function Can be set to 0.002, and The possible values ​​are as follows:

[0099]

[0100]

[0101] In some embodiments of the present application, Figure 3 This section introduces the processing of the input feature map by the boundary extraction submodule BEM, which can specifically include:

[0102] Perform deep dilation convolution on the input feature map to obtain a dilated feature map;

[0103] Performing batch normalization and convolution operations with a convolution kernel of 1 on the expanded feature map in sequence to obtain a normalized feature map;

[0104] After activating and convolving the normalized feature map using a SMU activation function, an activation feature map is generated;

[0105] The activation feature map is element-wise added to the input feature map to obtain a local feature map.

[0106] Specifically, such as Figure 3 As shown in the figure, the BEM module is designed to extract multi-scale local features. The input is a utility pole feature map X1 with a spatial resolution of H×W and a number of channels C. A depthwise dilated convolution (DDWConv) with a dilation rate of r and a kernel size of 3×3 is first performed on X1a. Batch normalization (BatchNorm) is then performed on X1a, followed by a convolution with a kernel size of 1×1 to obtain X1b. The SMU activation function is then applied to X1b, followed by another 1×1 convolution to obtain X1c. Finally, X1c is element-wise added to X1 to obtain X2, the boundary feature extraction result. If the number of executions n has not been reached, the loop continues; if it has, the result is output.

[0107] The calculation expression of the one-time fusion extraction of the boundary extraction submodule BEM is:

[0108]

[0109]

[0110] in, is the input feature map, is the activation feature map, is the local feature map, of Represents a convolution operation with a convolution kernel of 1, is the SMU activation function, is batch normalization, The expansion rate is Depth-wise dilated convolution.

[0111] Deep dilated convolution can increase the receptive field without increasing the amount of computation or reducing the resolution. This is achieved by setting the dilation rate parameter. The SMU (Smooth Maximum Unit) activation function is an activation function based on the smooth maximum technology, and its expression is:

[0112]

[0113] in, and is a hyperparameter that can be learned via gradient descent.

[0114] In some embodiments of the present application, Figure 4 The processing process of the global utility pole positioning module GUPLM on the input intermediate feature map and the upsampled feature map is introduced, which may specifically include:

[0115] Concatenate the intermediate feature map and the upsampled feature map along the channel dimension, and reduce the dimension by convolution to obtain a reduced-dimensional feature map;

[0116] Performing depth-separable asymmetric convolution of a first preset convolution kernel and a second preset convolution kernel on the dimensionality reduction feature map respectively, adding the results and activating them through SMU to generate a separable convolution feature map;

[0117] Performing SimAM attention mechanism calculation on the separated convolution feature map, and fusing it with the dimensionality reduction feature map by element addition to obtain a local fused feature map;

[0118] The local fused feature map, the intermediate feature map and the up-sampled feature map are fused element-wise to obtain a global fused feature map.

[0119] Specifically, utility poles may overlap with trees or buildings in the background when viewed from certain angles. Furthermore, global information is diluted due to the long path of propagation from high-level feature maps to low-level feature maps, resulting in only partial identification and segmentation of utility poles. Therefore, this application utilizes the Global Utility Pole Localization Module (GUPLM) to more effectively utilize global information in high-level feature maps, accurately locate utility pole targets, and enhance the network's ability to accurately identify and segment utility poles in complex backgrounds.

[0120] like Figure 4 As shown, the feature map output by the adjacent CBEM module after the downsampling module is used as an input feature map of GUPLM, denoted as .set up The size of is C×H×W, where C represents the number of channels, H represents the height, and W represents the width. The feature map output by the corresponding upsampling module is used as another input feature map of GUPLM, denoted as . The size and The specific implementation process of this module is as follows:

[0121] ① and Concatenate along the channel dimension to obtain a feature map with a spatial resolution of H×W and a channel number of 2C. Perform a convolution operation with a convolution kernel size of 1×1 on the feature map to reduce the number of channels of the feature map for full fusion. and The feature information of the pole is obtained with a spatial resolution of H×W and a channel number of C. .

[0122] ② General After a pair of depth-wise separable asymmetric convolutions DWConv, where the kernel size of one depth-wise separable asymmetric convolution is 1×3, a feature map of the electric pole with a spatial resolution of H×W and a number of channels C is obtained. Another depth-wise separable asymmetric convolution has a kernel size of 3×1, which results in a pole feature map with a spatial resolution of H×W and a number of channels of C. .

[0123] ③ General and Add them element by element, then perform a SMU activation function calculation to enhance the nonlinear ability of the model, and obtain the pole feature map with a spatial resolution of H×W and a channel number of C .

[0124] ④ General Input the SimAM module for attention calculation to obtain the pole feature map with adaptive spatial feature enhancement , its spatial resolution is H×W and the number of channels is C.

[0125] ⑤ General and Add element by element to get the pole feature map with spatial resolution of H×W and number of channels C .

[0126] ⑥ General 、 and Add element by element to get the pole feature map with spatial resolution of H×W and number of channels C , This is the output of the GUPLM module.

[0127] The following describes a device for dividing and locating a utility pole area provided by an embodiment of the present application. The device for dividing and locating a utility pole area described below and the method for dividing and locating a utility pole area described above can refer to each other.

[0128] See also Figure 5 , Figure 5 This is a schematic diagram of a utility pole area segmentation and positioning device disclosed in an embodiment of the present application.

[0129] like Figure 5 As shown, the utility pole area segmentation and positioning device may include:

[0130] An image acquisition unit 110 is used to acquire an image of a utility pole to be segmented;

[0131] The segmentation network unit 120 is used to input the utility pole image into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM. The continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, and each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations. The global utility pole positioning module GUPLM combines the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features.

[0132] The area positioning unit 130 is used to locate the electric pole area according to the grayscale value of each pixel in the grayscale image of the electric pole segmentation result.

[0133] It can be seen from the above technical solutions that the method for segmenting and locating a utility pole area and the related equipment provided in the embodiments of the present application have significant beneficial effects in many aspects and can effectively avoid many defects of the existing technology.

[0134] In terms of accurate multi-scale edge extraction, this application connects multiple levels of boundary extraction sub-modules (BEM) in series through a continuous boundary extraction module (CBEM), and combines deep dilated convolution with feature fusion operations. Deep dilated convolution can expand the receptive field of the convolution kernel without adding too many parameters, thereby fully capturing the local detail information of the utility pole. At the same time, through the feature fusion operation, feature maps of different scales are integrated, so that the global contour features of the utility pole can also be effectively preserved. This innovative design significantly improves the edge recognition accuracy in complex backgrounds, such as vegetation occlusion and poor lighting conditions, and successfully solves the edge blurring problem caused by the single feature scale of traditional methods.

[0135] For robust segmentation in complex backgrounds, the Global Pole Localization Module (GUPLM) fully utilizes image information at different resolutions by fusing the CBEM output with multi-level feature maps from the upsampling stage. Furthermore, it introduces depthwise separable asymmetric convolution and the SimAM attention mechanism. The depthwise separable asymmetric convolution effectively captures the directional characteristics of long, rectangular objects like utility poles, while the SimAM attention mechanism enhances the model's focus on the main area of ​​the pole, effectively suppressing background noise interference. This avoids mis-segmentation of poles with similar objects, significantly improving segmentation accuracy in complex backgrounds.

[0136] Furthermore, the utility pole image segmentation network is trained using the edge similarity loss function (ESLoss), which constrains the smoothness of the network output boundaries through gradient differences. Unlike traditional loss functions, which focus on a single optimization objective, the edge similarity loss function not only focuses on pixel-level classification accuracy but also emphasizes the continuity of edge structures. This ensures that the segmentation results more closely match the physical structure of real utility poles.

[0137] Compared with existing technologies, this application can more effectively ensure the accuracy and integrity of edge information, and the output segmentation results are more accurate and have smooth edges. At the same time, the present invention can fully utilize the local self-similarity of deep feature maps, further improving positioning accuracy and robustness, and has higher practical application value.

[0138] The utility pole area segmentation and positioning device provided in the embodiment of the present application can be applied to utility pole area segmentation and positioning equipment. Figure 6 The hardware structure diagram of the utility pole area segmentation and positioning device is shown. Figure 6 ,The hardware structure of the utility pole area segmentation and positioning device may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;

[0139] In the embodiment of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 communicate with each other through the communication bus 4;

[0140] The processor 1 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application;

[0141] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory;

[0142] The memory stores a program, and the processor can call the program stored in the memory, wherein the program is used to:

[0143] Obtain the telephone pole image to be segmented;

[0144] The utility pole image is input into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM, the continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations, the global utility pole positioning module GUPLM splices the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features;

[0145] The utility pole area is located according to the grayscale value of each pixel in the grayscale image of the utility pole segmentation result.

[0146] Optionally, the refined functions and extended functions of the program may refer to the above description.

[0147] The present application also provides a readable storage medium, which may store a program suitable for execution by a processor, wherein the program is used to:

[0148] Obtain the telephone pole image to be segmented;

[0149] The utility pole image is input into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM, the continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations, the global utility pole positioning module GUPLM splices the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features;

[0150] The utility pole area is located according to the grayscale value of each pixel in the grayscale image of the utility pole segmentation result.

[0151] Optionally, the refined functions and extended functions of the program may refer to the above description.

[0152] The present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the execution method is as follows:

[0153] Obtain the telephone pole image to be segmented;

[0154] The utility pole image is input into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM, the continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations, the global utility pole positioning module GUPLM splices the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features;

[0155] The utility pole area is located according to the grayscale value of each pixel in the grayscale image of the utility pole segmentation result.

[0156] Optionally, the refined functions and extended functions of the program may refer to the above description.

[0157] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0158] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0159] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for segmenting and locating a utility pole region, characterized in that: include: Obtain the telephone pole image to be segmented; The utility pole image is input into the trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM, the continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations, the global utility pole positioning module GUPLM splices the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features; The utility pole area is located according to the grayscale value of each pixel in the grayscale image of the utility pole segmentation result.

2. The method according to claim 1, characterized in that The training process of the utility pole image segmentation network includes: Constructing a utility pole image segmentation dataset, wherein the utility pole image segmentation dataset includes original images and corresponding real label images in single-channel digital form; Based on the utility pole image segmentation dataset, the utility pole image segmentation network is trained using an edge similarity loss function, wherein the edge similarity loss function optimizes network parameters by calculating the Laplace filter difference and gradient difference between the segmentation result and the true label image.

3. The method according to claim 1, characterized in that The processing process of the boundary extraction submodule BEM on the input feature map includes: Perform deep dilation convolution on the input feature map to obtain a dilated feature map; Performing batch normalization and convolution operations with a convolution kernel of 1 on the expanded feature map in sequence to obtain a normalized feature map; After activating and convolving the normalized feature map using a SMU activation function, an activation feature map is generated; The activation feature map is element-wise added to the input feature map to obtain a local feature map.

4. The method according to claim 3, characterized in that The calculation expression of the one-time fusion extraction of the boundary extraction submodule BEM is: in, is the input feature map, is the activation feature map, is the local feature map, of Represents a convolution operation with a convolution kernel of 1, is the SMU activation function, is batch normalization, The expansion rate is Depth-wise dilated convolution.

5. The method according to claim 1, wherein The global utility pole positioning module GUPLM processes the input intermediate feature map and the upsampled feature map, including: Concatenate the intermediate feature map and the upsampled feature map along the channel dimension, and reduce the dimension by convolution to obtain a reduced-dimensional feature map; Performing depth-separable asymmetric convolution of a first preset convolution kernel and a second preset convolution kernel on the dimensionality reduction feature map respectively, adding the results and activating them through SMU to generate a separable convolution feature map; Performing SimAM attention mechanism calculation on the separated convolution feature map, and fusing it with the dimensionality reduction feature map by element addition to obtain a local fused feature map; The local fused feature map, the intermediate feature map and the up-sampled feature map are fused element-wise to obtain a global fused feature map.

6. The method according to claim 2, characterized in that The edge similarity loss function is calculated as follows: in, The pixel matrix of the output image of the utility pole image segmentation network, is the pixel matrix of the true label image, is the absolute value, is the Laplace filtering operation, is the stability coefficient, for and The gradient difference in the horizontal direction, for and The gradient difference in the vertical direction.

7. A device for segmenting and positioning a utility pole area, characterized in that: include: An image acquisition unit, configured to acquire an image of the electric pole to be segmented; A segmentation network unit is used to input the utility pole image into a trained utility pole image segmentation network to obtain a grayscale image of the utility pole segmentation result, wherein the utility pole image segmentation network includes a downsampling module, an upsampling module, a continuous boundary extraction module CBEM and a global utility pole positioning module GUPLM. The continuous boundary extraction module CBEM extracts multi-scale local features by connecting multiple boundary extraction submodules BEM in series, and each BEM module sequentially performs deep dilated convolution, batch normalization, SMU activation function and feature fusion operations. The global utility pole positioning module GUPLM combines the intermediate feature map output by the continuous boundary extraction module CBEM and the upsampled feature map obtained by the upsampling module, and performs deep separable asymmetric convolution, SMU activation function and SimAM attention calculation to fuse global and local features. The area positioning unit is used to locate the electric pole area according to the grayscale value of each pixel in the grayscale image of the electric pole segmentation result.

8. A utility pole area segmentation and positioning device, characterized in that: including memory and processor; The memory is used to store programs; The processor is configured to execute the program to implement each step of the utility pole area segmentation and positioning method according to any one of claims 1 to 6.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the utility pole area segmentation and positioning method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the method for segmenting and locating a utility pole area as claimed in any one of claims 1 to 6 is executed.

Images