A high-precision extraction method for environment elements around a power transmission channel

Abstract

The application provides a high-precision extraction method for environment elements around a power transmission channel, and belongs to the technical field of power transmission line operation and maintenance and remote sensing information extraction. The application comprises standardization processing of multi-source heterogeneous data, multi-modal feature fusion oriented to a power transmission scene, multi-class element synchronous segmentation driven by deep learning, and closed-loop optimization and precision control of extraction results. The application realizes accurate and efficient extraction of environment elements by fusing multi-source remote sensing data and constructing a deep learning model special for the power transmission scene, and provides data support for intelligent operation and maintenance of the power transmission line.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of power transmission line operation and maintenance and remote sensing information extraction, and in particular to a high-precision extraction method for environmental elements around power transmission channels, specifically a high-precision extraction method for environmental elements around power transmission channels that integrates multi-source data and deep learning. Background Technology

[0002] As a core infrastructure of the power system, the dynamic changes in the surrounding environmental factors (such as tree obstructions, buildings, geological disaster sites, and crossings) of transmission channels directly affect the operational safety of the lines. Timely and accurate extraction of these environmental factors is a prerequisite for carrying out condition-based maintenance, risk warning, and hazard mitigation of transmission lines, and is of great significance for ensuring the stable operation of the power grid.

[0003] Existing methods for extracting environmental elements from power transmission channels are mainly divided into traditional manual inspection methods, single remote sensing extraction methods, and basic machine learning methods. Traditional manual inspection methods rely on on-site surveys by maintenance personnel. Although they can directly obtain environmental information, they are inefficient, costly, and have blind spots in complex terrains such as mountainous areas and areas spanning rivers, making it difficult to meet the routine monitoring needs of large-scale power transmission channels. Single remote sensing extraction methods (such as threshold segmentation based on optical images and regular filtering based on LiDAR point clouds) have achieved automated extraction, but they have obvious limitations: optical images are easily affected by lighting and weather, and their accuracy in distinguishing between tree obstacles and buildings is insufficient; although LiDAR point clouds can provide three-dimensional spatial information, their ability to identify small hidden dangers (such as foreign objects near conductors) is weak, and they cannot effectively correlate the attribute information of environmental elements.

[0004] In recent years, the application of deep learning technology in remote sensing segmentation has driven improvements in extraction accuracy. However, existing methods still have shortcomings: First, multi-source data fusion is insufficient, simply stitching together optical, point cloud, or GIS data without exploring the complementary features of each data source; second, the model lacks adaptability to power transmission channel scenarios, failing to optimize for professional requirements such as tree growth characteristics and building safety distance thresholds, resulting in extraction results that are out of sync with actual operation and maintenance needs; and third, the detection rate of small target elements (such as foreign objects on insulators and small geological cracks) is low, easily leading to missed detections. Therefore, developing a high-precision extraction method that integrates the advantages of multi-source data and adapts to power transmission channel scenarios has become an urgent need in the current power grid operation and maintenance field. Summary of the Invention

[0005] In view of this, in order to solve the problems of low efficiency, insufficient accuracy, poor scene adaptability and low detection rate of small targets in existing methods, this invention provides a high-precision extraction method for environmental elements around power transmission channels. By integrating multi-source remote sensing data and constructing a deep learning model dedicated to power transmission scenarios, it achieves accurate and efficient extraction of environmental elements, providing data support for intelligent operation and maintenance of power transmission lines.

[0006] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for high-precision extraction of environmental elements surrounding a power transmission channel, comprising: Collect and integrate high-resolution optical remote sensing image data, LiDAR point cloud data, UAV oblique photography data, power transmission line GIS vector data and equipment attribute data around the power transmission channel; Based on the spectral texture and spatial structure features of the collected data extracted by the multi-branch network, the feature enhancement of key environmental elements is achieved by combining prior knowledge of the power transmission scenario, and an enhanced fusion feature map is generated through an adaptive fusion mechanism. The TC-UNet power transmission scenario model, which adopts an improved U-Net architecture, performs simultaneous segmentation and extraction of multiple environmental elements from the enhanced fusion feature map, and outputs initial extraction results containing element type and spatial location. The initial extraction results are adjusted through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, the extraction results are managed in a closed loop, and high-precision environmental element extraction results that meet the needs of operation and maintenance are output.

[0007] Preferably, the device attribute data includes: Conductor type is used to specify the specifications of the transmission line conductors, providing a basis for determining safe distances in subsequent environmental factor extraction. The safe distance threshold, as a key constraint indicator in power transmission scenarios, is used to identify potential risks that exceed the safe distance.

[0008] In this invention, the collected data undergoes unified correction, registration, and format conversion to construct a standardized dataset with spatial consistency, providing a unified benchmark for subsequent feature extraction; wherein, Standardization processing includes: optical images undergoing beam adjustment geometric correction and Retinex texture enhancement; LiDAR point clouds undergoing statistical filtering for noise reduction and ICP registration; all data are unified to the WGS-84 coordinate system and the effective monitoring area of ​​the power transmission channel is cropped.

[0009] In this invention, the multi-branch network includes a spectral texture branch constructed by ResNet and a spatial structure branch constructed by PointConv; the prior knowledge of the power transmission scenario includes the conductor safety distance threshold and the tower coordinate range, and feature selection optimization is achieved by constructing a binary constraint mask; the adaptive fusion mechanism achieves weighted fusion by calculating the weight ratio of multi-source features to obtain an enhanced fusion feature map.

[0010] Preferably, the encoder of the TC-UNet power transmission scenario model embeds a dilated convolution module to expand the feature receptive field, and the decoder adds a spatial attention gating unit to enhance the features of key regions; the output layer uses the Softmax activation function to achieve classification output of multiple environmental elements.

[0011] Preferably, the rule correction and iterative optimization strategy is based on the power transmission operation and maintenance rules to remove invalid elements outside the conductor safety distance, uses the CRF algorithm to optimize the element segmentation boundary, and iteratively adjusts the element coordinate accuracy through the gradient descent method.

[0012] Preferably, the multi-dimensional accuracy verification metrics are intersection-over-union ratio, precision, recall, and small target detection rate.

[0013] Secondly, the present invention provides an electronic device, including a memory storing computer-executable instructions and a processor, wherein when the computer-executable instructions are executed by the processor, the device performs the above-mentioned high-precision extraction method for environmental elements surrounding the power transmission channel.

[0014] Thirdly, a readable storage medium stores a computer-executable program that, when executed, enables the aforementioned method for high-precision extraction of environmental elements surrounding the power transmission channel.

[0015] Fourthly, a three-dimensional modeling device for the terrain surrounding a power grid, used to perform the aforementioned high-precision extraction method for environmental elements surrounding power transmission channels, includes: The acquisition module is used to collect and integrate high-resolution optical remote sensing image data, LiDAR point cloud data, UAV oblique photography data, power transmission line GIS vector data and equipment attribute data around the power transmission channel; The fusion enhancement module is used to extract the spectral texture and spatial structure features of the data collected by the acquisition module based on the multi-branch network, combine the prior knowledge of the power transmission scenario to enhance the features of key environmental elements, and generate an enhanced fusion feature map through an adaptive fusion mechanism. The modeling module adopts the TC-UNet power transmission scenario model with an improved U-Net architecture. It performs synchronous segmentation and extraction of multiple environmental elements on the enhanced fusion feature map obtained by the fusion enhancement module, and outputs the initial extraction results containing element type and spatial location. The correction and optimization module adjusts the initial extraction results of the modeling module through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, it achieves closed-loop management of the extraction results and outputs high-precision environmental element extraction results that meet the needs of operation and maintenance.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention integrates multi-source data such as high-resolution optics, LiDAR point clouds, and UAV imagery, giving full play to the spectral and textural advantages of optical data and the three-dimensional spatial advantages of point cloud data. Combined with prior information on power transmission scenarios provided by GIS vector data, it effectively solves the problem of insufficient extraction accuracy from a single data source and achieves a comprehensive feature characterization of environmental elements.

[0017] The constructed TC-UNet model, dedicated to power transmission scenarios, improves the extraction efficiency of large-scale elements (such as clusters of buildings) and enhances the feature representation of small target elements (such as foreign objects on conductors) through the collaborative design of dilated convolution and attention gating units. The detection rate of small targets is more than 20% higher than that of the traditional U-Net model, which can meet the refined needs of power transmission channel hidden danger monitoring.

[0018] By introducing power transmission operation and maintenance rules and scenario constraints, the extracted results are deeply integrated with actual operation and maintenance needs. Invalid data is eliminated through rule correction, and structured results containing attributes such as element type, coordinate location, and safety distance are output. This can directly support subsequent risk assessment and hazard management, significantly improving the engineering applicability of the method.

[0019] This invention enables automated and high-precision extraction of environmental elements in power transmission channels. The data processing efficiency is more than 80% higher than that of manual inspection, and the extraction accuracy is 15%-20% higher than that of single remote sensing methods. It can adapt to complex terrain scenarios such as mountainous areas and crossing rivers, and provides an efficient technical means for routine monitoring of power transmission lines. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, this invention provides a method for high-precision extraction of environmental elements surrounding power transmission channels, including: Standardization of Multi-Source Heterogeneous Data: High-resolution optical remote sensing imagery, LiDAR point cloud data, UAV oblique photogrammetry data, power transmission line GIS vector data, and equipment attribute data from the vicinity of power transmission channels are collected and integrated. Through unified correction, registration, and format conversion, a spatially consistent standardized dataset is constructed, providing a unified benchmark for subsequent feature extraction. For example, multi-source raw data from the vicinity of power transmission channels are collected, such as high-resolution optical remote sensing imagery, LiDAR point cloud data, UAV oblique photogrammetry data, power transmission line GIS vector data, and equipment attribute data. The multi-source raw data undergoes standardization processing to obtain a standardized dataset in a unified coordinate system. The processing includes image correction and enhancement, point cloud denoising and registration, data format conversion, and outlier removal.

[0023] Multimodal feature fusion guided by power transmission scenarios: This approach extracts spectral texture and spatial structure features from multi-source data using a multi-branch network, enhances key environmental elements by incorporating prior knowledge of power transmission scenarios, and generates an enhanced fusion feature map through an adaptive fusion mechanism. For example, a multi-scale feature enhancement network is constructed, inputting high-resolution optical remote sensing images and LiDAR point cloud data from a standardized dataset into the spectral texture branch and spatial structure branch of the network, respectively. An attention mechanism is used to strengthen key features such as tree canopy layers and building edges. Prior knowledge of power transmission scenarios (such as conductor safety distances and tower coordinate ranges) is introduced to construct a feature constraint layer, which filters and optimizes the extracted features. Finally, an adaptive fusion module is used to achieve multi-dimensional feature fusion, outputting an enhanced fusion feature map.

[0024] Deep learning-driven simultaneous segmentation of multiple environmental elements: The TC-UNet power transmission scenario model, based on an improved U-Net architecture, performs simultaneous segmentation and extraction of multiple environmental elements from the fused feature map, outputting initial extraction results containing element type and spatial location. For example, a dedicated extraction model for power transmission channels (TC-UNet) based on the improved U-Net is constructed, with a dilated convolution module embedded in its encoder to expand the receptive field, and a spatial attention gating unit added to its decoder. The enhanced fused feature map is input into the TC-UNet model, and combined with equipment attribute data, simultaneous segmentation and extraction of multiple environmental elements such as tree obstacles, buildings, and geological disaster points are achieved, outputting initial extraction results.

[0025] Closed-loop optimization and accuracy control of extraction results: Initial results are finely adjusted through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, closed-loop control of the extraction results is achieved, outputting high-precision environmental element extraction results that meet operational requirements. For example, the initial extraction results are optimized using a "rule correction + iterative optimization" strategy. A verification dataset is constructed by combining on-site inspection data and LiDAR measurement data. Accuracy is verified using four indicators: Intersection over Union (IoU), precision, recall, and small target detection rate. If the preset accuracy threshold is not reached, feature enhancement is performed to adjust network parameters until a final extraction result that meets the requirements is output.

[0026] This technical solution achieves high-precision extraction through a four-stage process: "data standardization - feature fusion - deep learning segmentation - closed-loop optimization". Multi-source data standardization: Collect and integrate high-resolution optical imagery, LiDAR point cloud, UAV imagery, GIS vector data and equipment attribute data, and construct a spatially consistent standardized dataset through correction, registration and format conversion; Multimodal feature fusion: Multi-branch networks extract spectral texture and spatial structure features respectively, combine them with prior knowledge of power transmission scenarios to enhance key element features, and generate enhanced feature maps through adaptive weighted fusion; Deep learning segmentation: The TC-UNet model with an improved U-Net architecture is used to simultaneously segment multiple environmental elements on the fused feature map, and the output is an initial result containing the element type and spatial location; Closed-loop optimization and accuracy control: By correcting rules and iteratively optimizing the initial results, and verifying the accuracy with multi-dimensional indicators, a closed-loop control mechanism is formed to output the final result that meets the operation and maintenance requirements.

[0027] The technical solution of this invention integrates the complementary advantages of multi-source data, overcoming the limitations of single data sources in terms of illumination, resolution, and dimensionality. It constructs a dedicated model and optimized process for power transmission scenarios, ensuring that the extracted results are deeply adapted to the professional needs of power grid operation and maintenance. This achieves automated and large-scale extraction, significantly improving operational efficiency and adapting to complex terrain scenarios such as mountainous areas and areas spanning rivers.

[0028] In this invention, the device attribute data includes: Conductor type is used to specify the specifications of the transmission line conductors, providing a basis for determining safe distances in subsequent environmental factor extraction. The safe distance threshold, as a key constraint indicator in power transmission scenarios, is used to identify potential risks that exceed the safe distance.

[0029] This technical solution provides precise safety constraints for environmental element extraction, ensuring that the extraction results comply with power grid safety operation and maintenance standards. It establishes an attribute correlation between environmental elements and transmission equipment, enabling the extraction results to directly support practical risk assessment.

[0030] In this invention, the standardization process includes: optical image beam adjustment geometric correction and Retinex texture enhancement. Beam adjustment geometric correction eliminates spatial deviations, while Retinex texture enhancement improves the texture recognition of tree obstacles and buildings. LiDAR point clouds undergo statistical filtering for denoising and ICP registration. Statistical filtering removes outliers from the LiDAR point clouds, and ICP registration enables spatial stitching of multi-site point clouds. All data is unified to the WGS-84 coordinate system, and the effective monitoring area of ​​the transmission channel is cropped to ensure data spatial consistency and validity.

[0031] This technical solution eliminates spatial bias and quality differences in multi-source data, providing a unified benchmark for subsequent feature extraction. Cropping effective regions reduces interference from invalid data, improving the efficiency of subsequent feature extraction and model training.

[0032] In this invention, the multi-branch network includes a spectral texture branch constructed by ResNet and a spatial structure branch constructed by PointConv; the prior knowledge of the power transmission scenario includes the conductor safety distance threshold and the tower coordinate range, and feature selection optimization is achieved by constructing a binary constraint mask; the adaptive fusion mechanism achieves weighted fusion by calculating the weight ratio of multi-source features to obtain an enhanced fusion feature map.

[0033] In this technical solution, multimodal feature optimization is achieved through "branch extraction - prior constraints - adaptive fusion": Multi-branch feature extraction: The spectral texture branch constructed by ResNet extracts the spectral and texture features of optical images, and the spatial structure branch constructed by PointConv extracts the three-dimensional spatial and normal vector features of LiDAR point clouds. Scenario prior constraints: Combining prior knowledge such as conductor safety distance threshold and tower coordinate range, a binary constraint mask is constructed to filter and optimize the extracted features; Adaptive fusion: Weighted fusion is achieved by calculating the weight ratio of multi-source features to generate an enhanced fusion feature map.

[0034] This technical solution fully leverages the complementary features of multi-source data to enhance the feature representation capabilities of key elements such as tree canopies and building edges. It incorporates prior knowledge of power transmission scenarios to strengthen the feature weights of critical operation and maintenance areas, making the fusion results more aligned with actual needs. The adaptive fusion mechanism avoids information redundancy from simple splicing, achieving efficient integration of multi-dimensional features.

[0035] In this invention, the high-resolution optical remote sensing image features in the multi-source features have a weight of 0.45, prioritizing their rich spectral and textural information. The LiDAR point cloud data features have a weight of 0.4, focusing on utilizing their three-dimensional spatial structure information. The constraint mask features have a weight of 0.15, strengthening the filtering effect of power transmission scene constraints on features.

[0036] This technical solution rationally allocates the weight of each data source, fully leverages the core value of different data, balances the role of multi-source data and scenario constraints, optimizes the quality of fused feature maps, and improves subsequent segmentation accuracy.

[0037] In this invention, the encoder of the TC-UNet power transmission scenario model embeds a dilated convolution module to expand the feature receptive field without sacrificing resolution, capturing features of a wide range of environmental elements. The decoder adds a spatial attention gating unit to enhance key region features, and uses tower coordinate masks to further strengthen the representation of these key region features. The output layer employs a Softmax activation function to achieve classification output of multiple environmental elements.

[0038] In this technical solution, dilated convolution expands the receptive field, improving the extraction efficiency of large-scale elements such as clusters of buildings. Spatial attention gating units enhance features in key areas such as the perimeter of towers and under conductors, improving the detection rate of small target elements such as foreign objects on conductors. The improved architecture is deeply adapted to the needs of power transmission scenarios, significantly improving the segmentation accuracy of multiple types of elements.

[0039] In this invention, the simultaneous segmentation and extraction of six core environmental elements of the power transmission channel are achieved through Softmax classification in the output layer of the TC-UNet model. These environmental elements include trees, shrubs, buildings, roads, geological fissures, and foreign objects on the conductors.

[0040] This technical solution comprehensively covers the main environmental elements of power transmission channels, meeting the monitoring needs of various risks such as tree obstructions, buildings, geological hazards, and foreign objects on conductors during operation and maintenance; multiple elements are segmented simultaneously, reducing subsequent data processing steps and improving overall extraction efficiency.

[0041] In this invention, the rule correction and iterative optimization strategy removes invalid elements such as low shrubs outside the conductor safety distance based on power transmission operation and maintenance rules. The CRF (Conditional Random Field) algorithm is used to optimize the segmentation boundaries of elements (buildings, roads, etc.) to eliminate adhesion errors. The coordinate accuracy of elements (geological cracks, etc.) is iteratively adjusted using the gradient descent method.

[0042] In this technical solution, rule correction removes invalid data, making the results more closely match actual operation and maintenance needs. The CRF algorithm optimizes feature segmentation boundaries, improving the granularity of the results. Gradient descent adjusts coordinates, further improving the spatial positioning accuracy of the results.

[0043] In this invention, the multi-dimensional accuracy verification index is: Intersection over Union (IoU) measures the degree of overlap between the predicted region and the actual region. Precision is a measure of the proportion of true positive samples among the results that were predicted as positive. Recall rate measures the proportion of real samples that are correctly predicted; Small target detection rate, specifically targeting the detection capability of small target elements such as foreign objects in conductors and small geological cracks.

[0044] This technical solution comprehensively evaluates the accuracy of the extraction results using multiple dimensions, avoiding the limitations of a single indicator. Addressing the characteristics of small-target hazards in power transmission channels, a small-target detection rate indicator is added to ensure the monitoring accuracy of critical hazards. Closed-loop accuracy control of the extraction results is achieved, ensuring stable output that meets operational and maintenance requirements.

[0045] The invention will be further described in detail below with reference to a specific example of a 500kV transmission line channel monitoring.

[0046] Multi-source data collaborative acquisition and standardized processing Data Acquisition: A 20km area along a 500kV transmission line was selected as the research scope. 0.5m resolution optical images (including four bands: blue, green, red, and near-infrared) were acquired using WorldView-3 satellite. Point cloud data was collected using a RIEGL VZ-6000 lidar device with a point cloud density of 150 points / m². Oblique photogrammetry images and thermal infrared data were acquired using a DJI Matrice 350 RTK drone equipped with a Zenmuse H20T camera. Vector and attribute data, such as tower coordinates, conductor type, and safety distance thresholds, were exported from the power grid GIS system.

[0047] Standardization processing: Geometric correction was performed on optical images using a beam adjustment algorithm, with the error controlled within 1 pixel. The Retinex algorithm was used to enhance the texture contrast between tree barriers and buildings. Statistical filtering (18 neighboring points, standard deviation threshold of 1.8) was used to remove noise from LiDAR point clouds. Point cloud registration was completed for 5 stations using the ICP algorithm. Coordinate matching between optical images and point clouds was achieved using POS data from UAV imagery. All data were uniformly converted to the WGS-84 coordinate system, and effective data within a 500m range along the passage was cropped, removing interference data such as cloud shadows and abnormally high points in the point cloud.

[0048] Enhancement and Fusion of Power Transmission Scenarios Feature extraction branch construction: The spectral texture branch uses a pre-trained ResNet-101 network, removes the last two fully connected layers, adds a CAM attention module, and outputs a 2048-dimensional image feature map; the spatial structure branch uses a PointConv structure, sets 4 convolutional layers (64, 128, 256, 512 kernels), extracts the 3D spatial features and normal vector features of the point cloud, and outputs a 512-dimensional point cloud feature map.

[0049] Scene feature constraints and fusion: Based on the tower coordinates and conductor direction in the GIS data, a constraint zone with a radius of 50m around the tower and a safety constraint zone 10m below the conductor are generated, and a binary constraint mask is constructed. The image feature map and point cloud feature map are input into the feature fusion module, and the feature weights are calculated in combination with the constraint mask (optical feature weight 0.45, point cloud feature weight 0.4, constraint feature weight 0.15). A 1024-dimensional enhanced fusion feature map is obtained through matrix weighting.

[0050] Refined extraction of environmental elements TC-UNet model construction: Based on U-Net architecture, the encoder adopts a "3×3 convolution + dilated convolution (2 / 4 dilation alternating)" structure with 5 encoding blocks; the decoder adds a spatial attention gating unit to enhance key area features through tower coordinate masks; the output layer uses the Softmax activation function to achieve classification output of 6 types of elements: trees, shrubs, buildings, roads, geological cracks, and wire foreign objects.

[0051] Model training and initial extraction: The enhanced fusion feature map is divided into training, validation and test sets in a 7:2:1 ratio. The AdamW optimizer (initial learning rate 0.0006, cosine annealing strategy T_max=60) is used to train the model for 60 rounds. After training, standardized data is input and the initial extraction results containing the types and coordinate ranges of each environmental element are output.

[0052] Results optimization and accuracy verification Results optimization: Low shrubs (height < 2m) outside the safety distance of the conductor were removed based on operation and maintenance rules; the CRF algorithm was used to optimize the separation boundary between buildings and roads to eliminate adhesion errors; the coordinate accuracy of geological cracks was iteratively adjusted by gradient descent method, with the iteration step size dynamically reduced from 0.05 to 0.01 until the residual converged.

[0053] Accuracy verification: 200 field inspection samples (including 50 small target foreign object samples) were selected to construct a verification dataset. The model extraction accuracy was calculated: the overall IoU was 90.2%, the precision was 91.5%, the recall was 90.1%, and the detection rate of small targets such as wires and foreign objects was 87.3%, all of which met the preset threshold requirements. The final extraction results were output.

[0054] This invention also provides a three-dimensional modeling device for the terrain surrounding a power grid, characterized in that it is used to perform the above-mentioned high-precision extraction method for environmental elements surrounding power transmission channels, comprising: The acquisition module is used to collect and integrate high-resolution optical remote sensing image data, LiDAR point cloud data, UAV oblique photography data, power transmission line GIS vector data and equipment attribute data around the power transmission channel; The fusion enhancement module is used to extract the spectral texture and spatial structure features of the data collected by the acquisition module based on the multi-branch network, combine the prior knowledge of the power transmission scenario to enhance the features of key environmental elements, and generate an enhanced fusion feature map through an adaptive fusion mechanism. The modeling module adopts the TC-UNet power transmission scenario model with the improved U-Net architecture. It performs synchronous segmentation and extraction of multiple environmental elements on the enhanced fusion feature map obtained by the fusion enhancement module, and outputs the initial extraction results containing element type and spatial location. The correction and optimization module adjusts the initial extraction results of the modeling module through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, it achieves closed-loop management of the extraction results and outputs high-precision environmental element extraction results that meet the needs of operation and maintenance.

[0055] The methods and related apparatus mentioned in the above embodiments are described with reference to the flowcharts provided in the embodiments of the present invention. Each step of the method flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to create a machine, such that the instructions executable by the processor of the computer or other programmable data processing device generate instructions for implementing the process. Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operations to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 A process or multiple processes and / or structures illustrate the functions specified in one or more boxes.

[0056] The following embodiments illustrate the application of this method to a computer device. It is understood that the computer device can be any device with computing and processing capabilities, including but not limited to servers or personal laptops. In one embodiment, the computer device can be an application server, which can be a server used to run the application under test.

[0057] The present invention provides an electronic device, including a memory storing computer-executable instructions and a processor, wherein when the computer-executable instructions are executed by the processor, the device performs the above-mentioned high-precision extraction method for environmental elements surrounding the power transmission channel.

[0058] This invention provides a readable storage medium storing a computer-executable program, which, when executed, enables the aforementioned method for high-precision extraction of environmental elements surrounding power transmission channels.

[0059] The invention has been described in particular detail above with respect to possible scenarios, and those skilled in the art will recognize that the invention can be practiced through other embodiments. Specific naming of components, capitalization of terms, attributes, data structures, or any other programming or structural aspects are not mandatory or important, and the mechanisms or features of implementing the invention may have different names, forms, or procedures. The system can be implemented through a combination of hardware and software (as described), entirely through hardware elements, or entirely through software elements. The specific division of functions among the various system components described herein is merely exemplary and not mandatory; rather, the functions performed by a single system component can be performed by multiple components, or the functions performed by multiple components can be performed by a single component.

[0060] Those skilled in the art should understand that the methods disclosed above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using device-executable program code, thereby allowing them to be stored in a storage device for execution by the computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules can be fabricated as a single integrated circuit module. Thus, the embodiments disclosed in this invention are not limited to any specific hardware and software combination.

[0061] The programs (also referred to as programs, software, software applications, or code) executable by these computing devices include machine instructions of a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0062] Certain aspects of this invention include the processes and instructions described herein in algorithmic form. It should be noted that the processes and instructions of this invention can be implemented in software, firmware, and / or hardware; when implemented in software, they can be downloaded, stored on various operating systems used on different platforms, and operated from said platforms.

[0063] Those skilled in the art will understand that the structures shown in the figures are merely block diagrams of some structures related to the present application and do not constitute a limitation on the terminal device to which the present application is applied. Specific terminal devices may include more or fewer components than those shown in the figures, or combine certain components, or have different component arrangements.

[0064] The above description is merely a preferred embodiment of the present invention. However, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for high-precision extraction of environmental elements surrounding a power transmission channel, characterized in that, include: Collect and integrate high-resolution optical remote sensing image data, LiDAR point cloud data, UAV oblique photography data, power transmission line GIS vector data and equipment attribute data around the power transmission channel; Based on the spectral texture and spatial structure features of the collected data extracted by the multi-branch network, the feature enhancement of key environmental elements is achieved by combining prior knowledge of the power transmission scenario, and an enhanced fusion feature map is generated through an adaptive fusion mechanism. The TC-UNet power transmission scenario model, which adopts an improved U-Net architecture, performs simultaneous segmentation and extraction of multiple environmental elements from the enhanced fusion feature map, and outputs initial extraction results containing element type and spatial location. The initial extraction results are adjusted through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, the extraction results are managed in a closed loop, and high-precision environmental element extraction results that meet the needs of operation and maintenance are output.

2. The method for high-precision extraction of environmental elements surrounding a power transmission channel according to claim 1, characterized in that, Device attribute data includes: Conductor type is used to specify the specifications of the transmission line conductors, providing a basis for determining safe distances in subsequent environmental factor extraction. The safe distance threshold, as a key constraint indicator in power transmission scenarios, is used to identify potential risks that exceed the safe distance.

3. The method for high-precision extraction of environmental elements surrounding a power transmission channel according to claim 1, characterized in that, The collected data underwent unified correction, registration, and format conversion to construct a spatially consistent standardized dataset, providing a unified benchmark for subsequent feature extraction; among which, Standardization processing includes: optical images undergoing beam adjustment geometric correction and Retinex texture enhancement; LiDAR point clouds undergoing statistical filtering for noise reduction and ICP registration; all data are unified to the WGS-84 coordinate system and the effective monitoring area of ​​the power transmission channel is cropped.

4. The method for high-precision extraction of environmental elements surrounding a power transmission channel according to claim 1, characterized in that, The multi-branch network includes a spectral texture branch built from ResNet and a spatial structure branch built from PointConv; The prior knowledge of the power transmission scenario includes the safety distance threshold of the conductor and the coordinate range of the tower. Feature selection optimization is achieved by constructing a binary constraint mask. The adaptive fusion mechanism achieves weighted fusion by calculating the weight ratio of multi-source features to obtain an enhanced fusion feature map.

5. The method for high-precision extraction of environmental elements surrounding a power transmission channel according to claim 1, characterized in that, The encoder of the TC-UNet power transmission scenario model embeds a dilated convolution module to expand the feature receptive field, and the decoder adds a spatial attention gating unit to enhance the features of key regions. The output layer uses the Softmax activation function to achieve classified output of multiple environmental elements.

6. The method for high-precision extraction of environmental elements surrounding a power transmission channel according to claim 1, characterized in that, The rule correction and iterative optimization strategy is based on the power transmission operation and maintenance rules to remove invalid elements outside the conductor safety distance, uses the CRF algorithm to optimize the element segmentation boundary, and iteratively adjusts the element coordinate accuracy through the gradient descent method.

7. A method for high-precision extraction of environmental elements surrounding a power transmission channel according to any one of claims 1-6, characterized in that, The multi-dimensional accuracy verification metrics are intersection-over-union ratio, precision, recall, and small target detection rate.

8. An electronic device, characterized in that, The device includes a memory storing computer-executable instructions and a processor, which, when executed by the processor, causes the device to perform the high-precision extraction method for environmental elements surrounding the power transmission channel as described in any one of claims 1-7.

9. A readable storage medium, characterized in that, It stores a computer-executable program that, when executed, enables the high-precision extraction method for environmental elements surrounding the power transmission channel as described in any one of claims 1-7.

10. A three-dimensional modeling device for the terrain surrounding a power grid, characterized in that, include: The acquisition module is used to collect and integrate high-resolution optical remote sensing image data, LiDAR point cloud data, UAV oblique photography data, power transmission line GIS vector data and equipment attribute data around the power transmission channel; The fusion enhancement module is used to extract the spectral texture and spatial structure features of the data collected by the acquisition module based on the multi-branch network, combine the prior knowledge of the power transmission scenario to enhance the features of key environmental elements, and generate an enhanced fusion feature map through an adaptive fusion mechanism. The modeling module adopts the TC-UNet power transmission scenario model with an improved U-Net architecture. It performs synchronous segmentation and extraction of multiple environmental elements on the enhanced fusion feature map obtained by the fusion enhancement module, and outputs the initial extraction results containing element type and spatial location. The correction and optimization module adjusts the initial extraction results of the modeling module through rule correction and iterative optimization strategies. Combined with multi-dimensional accuracy verification indicators, it achieves closed-loop management of the extraction results and outputs high-precision environmental element extraction results that meet the needs of operation and maintenance.

Images