Method and system for dynamic monitoring of external damage to power transmission lines
By deploying multi-dimensional front-end monitoring devices on transmission line towers and utilizing laser array ranging, visible light and infrared image acquisition, and AI multimodal fusion technology, the system has achieved accurate identification and early warning linkage of external damage targets on transmission lines. This solves the problems of limited monitoring coverage and delayed response in existing technologies, and improves the real-time performance and accuracy of monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INNER MONGOLIA ELECTRIC POWER (GRP) CO LTD BAYANNAOER POWER SUPPLY BRANCH
- Filing Date
- 2026-03-19
- Publication Date
- 2026-07-21
AI Technical Summary
Existing monitoring systems for external damage to transmission lines suffer from limited coverage, delayed response, poor environmental adaptability, and difficulty in timely detection and identification of external intrusions. In particular, monitoring blind spots are prone to occur at night or in severe weather, affecting the safe and stable operation of the lines.
Multi-dimensional front-end monitoring devices are deployed on the towers. An adaptive laser stealth protection zone is constructed through a laser array ranging unit. Image acquisition is performed by combining visible light cameras and infrared monitoring units, and intelligent AI multi-modal fusion is carried out to achieve target recognition and early warning linkage.
It enables real-time, accurate, and three-dimensional monitoring and early warning of external damage risks to transmission lines, improving the real-time performance and accuracy of external damage prevention monitoring.
Smart Images
Figure CN122437231A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power transmission line technology, specifically to a method and system for dynamic monitoring of power transmission lines against external damage. Background Technology
[0002] Existing monitoring methods for external damage to transmission lines largely rely on manual inspections, video surveillance, or single-sensor approaches. These methods suffer from limited monitoring coverage, delayed response times, and poor environmental adaptability, making it difficult to promptly detect and identify potential risks such as external intrusion, excessively high-rise construction, and floating foreign objects. Traditional monitoring methods cannot achieve dynamic, three-dimensional perception and accurate early warning of line protection zones. In particular, blind spots are prone to appear at night, in severe weather, or in complex terrain, leading to a high frequency of external damage incidents and untimely response, which seriously affects the safe and stable operation of transmission lines. Summary of the Invention
[0003] This application provides a method and system for dynamic monitoring of external damage to transmission lines, which is used to address the technical problem that existing technologies are unable to monitor and provide early warning of external damage risks to transmission lines in real time, accurately, and in a three-dimensional manner.
[0004] In view of the above problems, this application provides a method and system for dynamic monitoring of transmission lines against external damage.
[0005] The first aspect of this application provides a method for dynamic monitoring of transmission lines against external damage, the method comprising: A multi-dimensional front-end monitoring device is deployed on the tower of the target transmission line. An adaptive laser stealth protection zone is constructed using the laser array ranging unit of the multi-dimensional front-end monitoring device. The laser array ranging unit is used to measure distances and obtain a laser ranging data sequence. Simultaneously, the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device are invoked to acquire visible light image sequences and infrared image sequences. Intelligent AI multi-modal fusion is performed on the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multi-modal fusion data features. Target recognition is performed based on the multi-modal fusion data features. The target recognition result is compared with the adaptive laser stealth protection zone to obtain a first warning message. Based on the first warning message, audible and visual warnings and voice broadcasts are activated.
[0006] A second aspect of this application provides a dynamic monitoring system for preventing external damage to transmission lines, the system comprising: The system comprises the following modules: a deployment module for installing multi-dimensional front-end monitoring devices on the towers of the target transmission line, constructing an adaptive laser stealth protection zone using the laser array ranging unit of the multi-dimensional front-end monitoring device; an image acquisition module for using the laser array ranging unit to perform ranging, obtaining a laser ranging data sequence, and simultaneously calling the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device to acquire visible light image sequences and infrared image sequences; a data fusion module for performing intelligent AI multimodal fusion of the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multimodal fused data features; and an early warning module for performing target recognition based on the multimodal fused data features, comparing the target recognition result with the adaptive laser stealth protection zone to obtain a first early warning message, and activating audible and visual warnings and voice broadcasts based on the first early warning message.
[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application deploys a multi-dimensional front-end monitoring device on the towers of the target transmission line. An adaptive laser stealth protection zone is constructed using the laser array ranging unit of the multi-dimensional front-end monitoring device. The laser array ranging unit is used to measure distances, obtaining a laser ranging data sequence. Simultaneously, the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device are invoked to acquire images, obtaining visible light image sequences and infrared image sequences. Intelligent AI multi-modal fusion is performed on the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multi-modal fused data features. Based on the multi-modal fused data features, target recognition is performed. The target recognition result is compared with the adaptive laser stealth protection zone to obtain a first warning message. Based on the first warning message, audible and visual warnings and voice broadcasts are activated. This invention addresses the technical problem of existing technologies' inability to monitor and warn of external damage risks to transmission lines in a real-time, accurate, and three-dimensional manner. By deploying multi-dimensional front-end monitoring devices on towers and utilizing laser array ranging, visible light and infrared image acquisition, and AI multimodal fusion technology, it achieves accurate identification and early warning linkage of external damage targets, thereby improving the real-time performance and accuracy of external damage monitoring of transmission lines. Attached Figure Description
[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0009] Figure 1 This is a schematic flowchart of the dynamic monitoring method for preventing external damage to transmission lines provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the dynamic monitoring system for preventing external damage to transmission lines provided in an embodiment of this application.
[0010] Explanation of reference numerals in the attached diagram: Deployment module 11, Image acquisition module 12, Data fusion module 13, Early warning module 14. Detailed Implementation
[0011] This application provides a dynamic monitoring method and system for preventing external damage to transmission lines. It addresses the technical problem that existing technologies are unable to monitor and warn of external damage risks to transmission lines in real time, accurately, and in a three-dimensional manner. By deploying multi-dimensional front-end monitoring devices on the towers and utilizing laser array ranging, visible light and infrared image acquisition, and AI multimodal fusion technology, it achieves accurate identification and early warning linkage of external damage targets, thereby improving the real-time performance and accuracy of monitoring for external damage to transmission lines.
[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.
[0014] Example 1, as Figure 1 As shown, this application provides a dynamic monitoring method for preventing external damage to transmission lines, the method comprising: Step S100: Deploy multi-dimensional front-end monitoring devices on the towers of the target transmission line, and construct an adaptive laser stealth protection zone through the laser array ranging unit of the multi-dimensional front-end monitoring devices.
[0015] In this embodiment of the application, a multi-dimensional front-end monitoring device is first deployed at a predetermined position on the tower of the target transmission line. This multi-dimensional front-end monitoring device is a composite monitoring terminal that integrates laser array ranging, visible light acquisition and infrared monitoring, and is used to conduct all-round and multi-angle real-time monitoring of the transmission line protection zone.
[0016] Next, an adaptive laser stealth protection zone is constructed using the laser array ranging unit of the multi-dimensional front-end monitoring device. In this process, the initial protection range is first determined based on the correlation indicators of the primary protection zone, forming the primary laser stealth protection zone to be deployed. Then, historical ranging data from the laser array ranging unit is retrieved, and combined with ranging log features, adaptive identification of the secondary protection zone is performed to obtain the secondary laser stealth protection zone to be deployed. The secondary laser stealth protection zone is then superimposed on the primary laser stealth protection zone to form a dynamically adjustable adaptive deployment zone. Finally, the zone deployment is completed through the control module, resulting in the adaptive laser stealth protection zone.
[0017] Furthermore, in the method provided in the application embodiments, a multi-dimensional front-end monitoring device is deployed on the tower of the target transmission line, and an adaptive laser stealth protection area is constructed through the laser array ranging unit of the multi-dimensional front-end monitoring device, which further includes: The process involves: acquiring primary protection zone correlation indicators; identifying primary protection zones based on these indicators to obtain primary laser stealth protection zones to be deployed; retrieving the ranging log set of the laser array ranging unit within a preset historical window; adaptively identifying secondary protection zones based on the ranging log set to obtain secondary laser stealth protection zones to be deployed; deploying the secondary laser stealth protection zones outside the primary laser stealth protection zones to obtain adaptive laser stealth protection zones to be deployed; and feeding back the adaptive laser stealth protection zones to the control module of the laser array ranging unit for area deployment to obtain the adaptive laser stealth protection zones.
[0018] Furthermore, the method provided in the application embodiments also includes: The relevant indicators for the primary protection zone include the voltage level of the transmission line, clearance height, tower spacing, and laser equipment performance.
[0019] In this embodiment, firstly, the primary protection zone association indicators are obtained and the primary protection zone identification is completed. The voltage level refers to the rated voltage of the transmission line, such as 110 kV or 220 kV. The clearance height is the vertical or oblique safe distance between the conductor and ground-based construction machinery to meet insulation requirements. The tower spacing is the horizontal projected distance between two adjacent towers. The laser equipment performance is a set of key parameters related to laser array ranging, including emission power, ranging accuracy, field of view, and beam divergence angle. Based on the correspondence between different voltage levels and clearance requirements for enterprises or industries, the voltage level, clearance height, and tower spacing are used for spatial geometric constraint calculations. Combined with the laser equipment performance, the coverage area and boundary errors are checked. A strip-shaped foundation area is generated along the tower connection axis, with a bandwidth set to 10 to 50 meters to meet the safety distance warning requirements, thus obtaining the primary laser stealth protection zone to be deployed.
[0020] Next, the ranging log set of the laser array ranging unit within the preset historical window is retrieved. The ranging log set is a time series dataset formed during the laser ranging process, containing indicators such as timestamp communication response delay, ranging confidence, and relative velocity trigger frequency. Each indicator is aligned and normalized according to the timestamp, and obvious outliers are removed to form a time series dataset that can be used for stable statistical analysis, resulting in the sorted ranging log set.
[0021] Then, adaptive identification of the secondary protection zone is performed based on the ranging log set. This process begins by extracting communication response delay, ranging confidence level, and relative velocity to construct a multi-dimensional protection reliability index set. Next, mean-shift filtering is applied to this set to determine the selected multi-dimensional protection reliability indices. Then, the trigger frequency of the ranging log set is extracted and weighted with the selected multi-dimensional protection reliability indices to calculate the secondary protection coefficient. Finally, based on this secondary protection coefficient, the bandwidth outside the primary zone is adaptively expanded or contracted to obtain the secondary laser stealth protection zone to be deployed.
[0022] Next, the secondary laser stealth protection area to be deployed is placed outside the primary laser stealth protection area to be deployed. Through spatial superposition, an overall protection boundary is formed, thus completing the generation of the adaptive laser stealth protection area to be deployed.
[0023] Finally, the boundary parameters of the adaptive laser stealth protection area to be deployed are fed back to the control module of the laser array ranging unit. This control module receives the boundary parameters and configures and enables the ranging threshold, scanning angle, and linkage alarm strategy. During operation, it automatically updates the boundary parameters based on newly generated ranging logs, achieving continuous dynamic adaptation and ultimately obtaining the adaptive laser stealth protection area.
[0024] Furthermore, in the method provided in the application embodiments, the method for adaptively identifying secondary protection areas based on the ranging log set to obtain the secondary laser stealth protection area to be deployed also includes: The communication response delay, ranging confidence level, and relative speed of the ranging log set are extracted and traversed to construct a multi-dimensional protection reliability index set. The multi-dimensional protection reliability index set is filtered by mean shift to determine the selected multi-dimensional protection reliability indexes. The trigger frequency of the ranging log set is extracted and weighted with the selected multi-dimensional protection reliability indexes to obtain the secondary protection coefficient. The secondary protection area is adaptively identified based on the secondary protection coefficient to obtain the secondary laser stealth protection area to be deployed.
[0025] In this embodiment, communication response delay, ranging confidence level, and relative velocity are first extracted to construct a multi-dimensional set of protection reliability indicators. Communication response delay is the time difference between the laser array ranging unit issuing a ranging command and receiving the echo signal, reflecting the timeliness and stability of the ranging link. Ranging confidence level is the degree of reliability calculated based on the echo signal quality and intensity, used to measure the reliability of the ranging results. Relative velocity is calculated from the ranging change over adjacent time steps. By extracting these three types of indicators and synchronizing them by timestamp, and using standardization processes such as Z-score standardization to eliminate the influence of dimensions, a multi-dimensional set of protection reliability indicators is formed.
[0026] Next, mean drift filtering is performed on the set of multidimensional protection reliability indicators. In this process, the initial mean drift filtering center is first determined by calculating the mean of the set of multidimensional protection reliability indicators. Then, the initial mean drift filtering center is iteratively filtered according to a preset filtering bandwidth, ultimately converging to obtain the selected multidimensional protection reliability indicators.
[0027] Next, the trigger frequency of the ranging log set is extracted and weighted with the selected multi-dimensional protection reliability indicators to calculate the secondary protection coefficient. The trigger frequency is the number of effective triggers of the laser ranging signal per unit time, reflecting the frequency of activity within the protected area. Since the trigger frequency has different dimensions than other indicators, it is standardized to maintain a magnitude consistent with communication response delay, ranging confidence level, and relative speed. In the weighted analysis, a linear weighted model is used to combine communication response delay, ranging confidence level, relative speed, and trigger frequency according to weights, for example, setting weight coefficients W1, W2, W3, and W4. The calculation formula is: Secondary protection coefficient = W1 × ranging confidence level. W2 × Communication Response Delay + W3 × Relative Velocity + W4 × Trigger Frequency. When the trigger frequency is high or the ranging confidence level remains stable, the secondary protection coefficient increases, indicating increased external activity. Conversely, when the communication response delay increases or the relative velocity change decreases, the secondary protection coefficient decreases, indicating that external activity is stabilizing. This weighted approach integrates multi-source dynamic information to obtain a secondary protection coefficient reflecting the current protection pressure and risk level. W1, W2, W3, and W4 are preset by technical experts and sum to 1.
[0028] Finally, the secondary protection zone is adaptively identified based on the secondary protection coefficient. When the secondary protection coefficient is higher than the set upper threshold, the protection bandwidth outside the primary protection zone is expanded according to the expansion rule, for example, from 30 meters to 60 meters, to cover a larger area. When the secondary protection coefficient is lower than the set lower threshold, the outer bandwidth is reduced according to the contraction rule, for example, from 60 meters to 30 meters. Through this adaptive expansion and contraction process, the secondary protection zone can dynamically respond to changes in the intensity of target activity and external risks, thereby obtaining the secondary laser stealth protection zone to be deployed.
[0029] Furthermore, in the method provided in the application embodiments, the mean shift screening of the multidimensional protection reliability index set to determine the selected multidimensional protection reliability index further includes: Calculate the mean of the set of multidimensional protection reliability indicators to determine the initial mean drift screening center; according to the preset screening bandwidth, iteratively screen the initial mean drift screening center in the set of multidimensional protection reliability indicators to determine the screened multidimensional protection reliability indicators.
[0030] In this embodiment, the mean of the multidimensional protection reliability index set is first calculated to determine the initial mean drift screening center. The multidimensional protection reliability index set consists of three standardized feature vectors: communication response delay, ranging confidence level, and relative velocity. The arithmetic mean of each dimension is calculated to obtain the center point located in the feature space, and this center point is defined as the initial mean drift screening center.
[0031] Next, the initial mean drift screening center is iteratively screened within the multidimensional protection reliability index set according to a preset screening bandwidth. The preset screening bandwidth is a distance threshold used to control the screening range. In each iteration, the Euclidean distance between each data point in the multidimensional protection reliability index set and the current mean drift screening center is calculated. Data points whose distance does not exceed the preset screening bandwidth are selected for re-averaging, and this result is used to update the mean drift screening center. The iterative process continues until the offset of the center position between two calculations is less than the convergence threshold, or the maximum number of iterations is reached, thus completing the iterative screening process and obtaining the converged mean drift screening center.
[0032] Finally, using the converged mean drift center as a benchmark, data within the preset filtering bandwidth are considered high-density areas and retained, while data outside the filtering bandwidth are discarded. This process yields the filtered multi-dimensional protection reliability index.
[0033] Step S200: Use the laser array ranging unit to measure the distance, obtain the laser ranging data sequence, and call the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device to simultaneously acquire images, and obtain the visible light image sequence and infrared image sequence.
[0034] In this embodiment, when using a laser array ranging unit for distance measurement, a laser beam is first emitted from a laser emitter towards the target area. After the laser beam strikes the surface of the target object, a reflected signal is generated, which is received by the receiver array of the laser array ranging unit. The time-of-flight method is used to calculate the time difference between laser emission and reception. Multiplying this time difference by half the speed of light yields the distance between the target object and the laser array ranging unit. By simultaneously calculating the distances at multiple measurement points and recording the results in chronological order, continuous distance measurement information is formed, thus obtaining a laser ranging data sequence.
[0035] Simultaneously, the visible light camera unit of the multi-dimensional front-end monitoring device is invoked to acquire images synchronously. The visible light camera unit operates based on a CCD image sensor. When it receives external ambient light, it converts the light signal into an electrical signal, which is then converted into digital image frames by the imaging component and arranged in the order of acquisition time to form a visible light image sequence.
[0036] In addition, the infrared monitoring unit of the multi-dimensional front-end monitoring device synchronously acquires images. The infrared monitoring unit operates based on the principle of thermal radiation detection. It receives the infrared radiation energy emitted by the target area, converts it into electrical signals, and generates infrared image frames. These frames are recorded in the order of acquisition time to form an infrared image sequence.
[0037] Step S300: Perform intelligent AI multimodal fusion on the laser ranging data sequence, visible light image sequence and infrared image sequence to obtain multimodal fusion data features.
[0038] Furthermore, the method provided in the application embodiments, which performs intelligent AI multimodal fusion on the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multimodal fusion data features, further includes: The visible light image sequence and the infrared image sequence are semantically consistent and integrated to obtain a fused image semantic feature sequence; point cloud feature analysis is performed based on the laser ranging data sequence to obtain a laser point cloud feature sequence; intelligent AI multimodal linkage fusion is performed on the fused image semantic feature sequence and the laser point cloud feature sequence to obtain the multimodal fused data features.
[0039] In this embodiment, when performing intelligent AI multimodal fusion on laser ranging data sequences, visible light image sequences, and infrared image sequences, the visible light image sequences and infrared image sequences are first integrated for semantic consistency. In this process, a semantic segmentation network layer is first used to perform semantic feature recognition on the visible light image sequences and infrared image sequences respectively, extracting semantic category information, edge features, and thermal radiation features of the target region to form visible light semantic feature map sequences and infrared semantic feature map sequences. Subsequently, through a mapping attention weighting method, the visible light semantic feature map sequences and infrared semantic feature map sequences are aligned and weighted and integrated in the same semantic space to achieve complementarity between texture information and thermal information, obtaining a fused image semantic feature sequence.
[0040] Next, point cloud feature analysis is performed based on the laser ranging data sequence. First, the laser ranging data is converted into spatial point cloud coordinates using coordinate reconstruction methods. Specifically, the distance information in the laser ranging data is combined with preset scanning angle information, and the three-dimensional coordinates of each ranging point are calculated using the polar coordinate to Cartesian coordinate conversion formula to obtain the initial point cloud data. Second, the initial point cloud data is sparsified using a voxel filtering method. This involves dividing the point cloud into a fixed-size voxel grid and replacing all points within the grid with voxel centers, reducing redundant data and improving computational efficiency. Next, statistical filtering is used to remove noise from the sparse point cloud data. By calculating the average distance between each point and its neighbors, isolated and outlier points are removed according to preset thresholds to obtain denoised point cloud data. Then, a region growing method is used to extract point cloud boundary features. By calculating the angle between the normal vectors and the curvature difference between points, the boundary regions are clustered and segmented to identify the structural contour information of the point cloud. Finally, the depth distribution features, boundary features, and spatial geometric features are combined in chronological order to form a laser point cloud feature sequence.
[0041] Finally, intelligent AI multimodal linkage fusion is performed on the semantic feature sequence of the fused image and the laser point cloud feature sequence. In this process, the semantic branch is used to extract the semantic category trend and edge contour information trend of the semantic feature sequence of the fused image, and the geometric branch is used to extract the point cloud depth trend and spatial position trend of the laser point cloud feature sequence. The extraction results are then input into the linkage fusion unit to obtain multimodal fused data features.
[0042] Furthermore, in the method provided in the application embodiments, the semantic consistency integration of the visible light image sequence and the infrared image sequence to obtain a fused image semantic feature sequence further includes: Semantic feature map recognition is performed on the visible light image sequence and the infrared image sequence using a semantic segmentation network layer to obtain a visible light semantic feature map sequence and an infrared semantic feature map sequence; the visible light semantic feature map sequence and the infrared semantic feature map sequence are then integrated by mapping attention weighted semantic consistency to obtain the fused image semantic feature sequence.
[0043] In this embodiment, when performing semantic feature map recognition on visible light image sequences and infrared image sequences using a semantic segmentation network layer, the visible light image sequences and infrared image sequences are first input into a pre-trained semantic segmentation network layer, and the pixels of each frame are semantically categorized. The semantic segmentation network layer is a feature extraction structure based on a deep convolutional neural network. Through layer-by-layer convolution and upsampling operations, it expresses the semantic information corresponding to different regions in the image in the form of feature maps. For visible light image sequences, this process extracts texture features, contour features, and object semantic labels; for infrared image sequences, this process extracts thermal radiation distribution features and semantic labels for thermal anomaly regions. After recognition, visible light semantic feature map sequences and infrared semantic feature map sequences are formed respectively.
[0044] Subsequently, the visible light semantic feature map sequences and infrared semantic feature map sequences are integrated using a mapping attention-weighted semantic consistency method. In this process, the visible light and infrared semantic feature map sequences are first synchronized frame-by-frame based on timestamps, ensuring that each frame of visible light semantic feature map is paired with its corresponding infrared semantic feature map, guaranteeing temporal consistency between the two sequences. Next, spatial mapping relationships are calculated through feature point matching, mapping the infrared semantic feature map sequence to the coordinate reference of the visible light semantic feature map sequence, ensuring that corresponding pixel positions point to the same target region. This achieves frame-by-frame alignment in the spatial dimension, resulting in spatially aligned visible light and infrared semantic feature map sequences.
[0045] Then, the spatially aligned visible light semantic feature map sequence and infrared semantic feature map sequence are processed frame-by-frame and pixel-by-pixel to measure the semantic closeness of the two types of features. Attention weights are generated based on similarity, with higher weights for pixels with high similarity and lower weights for pixels with low similarity. The visible light and infrared semantic features of each frame are then weighted according to these attention weights, enhancing features in highly consistent regions and suppressing conflicting information in low-consistency regions, thus achieving dynamic weight allocation at the feature level. Finally, all frames processed by the mapping attention-weighted semantic consistency integration are combined in chronological order to form a fused image semantic feature sequence.
[0046] Furthermore, in the method provided in the application embodiments, the intelligent AI multimodal linkage fusion of the fused image semantic feature sequence and the laser point cloud feature sequence to obtain the multimodal fused data features further includes: The semantic branch is used to extract the semantic category trend and edge contour information trend of the semantic feature sequence of the fused image, and the geometric branch is used to extract the point cloud depth trend and spatial position trend of the laser point cloud feature sequence. The extraction results are then input into the linkage fusion unit to obtain the multimodal fused data features.
[0047] In this embodiment, a semantic branch is first used to extract semantic category trends and edge contour information trends from the semantic feature sequence of the fused image. The semantic branch is pre-trained and its parameters optimized on a dataset containing a large number of image samples, enabling it to stably extract temporal semantic features. When processing the semantic feature sequence of the fused image, the category probability distribution and edge response information are first extracted for each frame. Then, in the time dimension, the category changes and contour changes of adjacent frames are differentially processed and temporally smoothed to enhance the feature representation of continuously changing regions. Through this process, semantic category trends reflecting the changes of the target category over time and edge contour information trends reflecting the changes of edge morphology over time are obtained.
[0048] Next, a geometric branch is used to extract geometric information from the laser point cloud feature sequence. This geometric branch, also trained and optimized on point cloud samples with depth information and spatial annotations, possesses the ability to extract stable geometric features. When processing the laser point cloud feature sequence, the depth statistics and spatial position of the point cloud in each frame are first encoded. Then, in the temporal dimension, a pose change estimation method is used to extract the depth changes and positional offsets of adjacent frames, refining continuous geometric change features. Through this process, the point cloud depth trend reflecting the target distance over time and the spatial position trend reflecting the target's spatial pose over time are obtained.
[0049] Finally, a joint fusion processor is used to jointly process the features output from the semantic and geometric branches. The joint fusion processor obtains stable feature fusion parameters through joint training with multi-source data. Correspondences are established between semantic category trends, edge contour information trends, point cloud depth trends, and spatial location trends at the same timestamp and corresponding spatial location. Fusion weights are calculated based on the consistency between semantic and geometric features, and features from different sources are weighted and fused. Common features are strengthened in regions of high consistency, while conflicting features are suppressed in regions of low consistency, thus obtaining multimodal fused data features with dual expressive power of semantic category trends and spatial location trends.
[0050] Step S400: Based on the multimodal fusion data features, target recognition is performed, the target recognition result is compared with the adaptive laser stealth protection area to obtain the first warning information, and the sound and light warning and voice broadcast are activated according to the first warning information.
[0051] In this embodiment, when performing target recognition based on multimodal fusion data features, the multimodal fusion data features are first identified using a pre-trained target recognizer to obtain the target recognition result. The target recognition result is then compared with an adaptive laser stealth protection area. The comparison process determines whether the identified target has entered the area based on spatial location correspondence. When the target recognition result is located within the adaptive laser stealth protection area, it is determined to be a potential intrusion or threat event, and a first warning message is generated.
[0052] Upon receiving the first warning information, the external alarm device is triggered to provide real-time alerts to the surrounding area through sound and light warnings and voice broadcasts.
[0053] Furthermore, in the method provided in the application embodiments, target recognition is performed based on the multimodal fusion data features, and the target recognition result is compared with the adaptive laser stealth protection area to obtain first warning information, which further includes: A pre-built target recognizer identifies the features of the multimodal fusion data to obtain a target recognition result; it is then determined whether the target recognition result has entered the adaptive laser stealth protection area, and if so, a first warning message is obtained.
[0054] In this embodiment, when performing target recognition based on multimodal fusion data features, the multimodal fusion data features are first identified using a pre-built target recognizer. The target recognizer is a recognition model trained using deep learning methods. During training, multimodal data containing visible light images, infrared images, and laser point cloud features are used as input, along with labeled target categories and spatial location trends as output. Feature extraction and parameter optimization are employed during training to enable the target recognizer to identify the target's category features and capture its spatial positional variation patterns.
[0055] During the identification phase, multimodal fusion data features are input into the target recognizer. The target recognizer performs feature analysis on the input data based on semantic category trends and spatial location trends, and outputs a target recognition result containing target category labels and target spatial location trends. Target category labels are used to clarify the type of target, and spatial location trends are used to describe the spatial distribution and temporal changes of the target relative to the protected area.
[0056] After obtaining the target identification result, it is compared with the adaptive laser stealth protection area. During the comparison, based on the spatial correspondence between the target's spatial position trend and the boundary of the protection area, it is determined whether the target has entered the adaptive laser stealth protection area. When the spatial position trend overlaps with the protection area, it is determined as an intrusion event, and a first warning message is generated.
[0057] In summary, the embodiments of this application have at least the following technical effects: This application deploys a multi-dimensional front-end monitoring device on the towers of the target transmission line. An adaptive laser stealth protection zone is constructed using the laser array ranging unit of the multi-dimensional front-end monitoring device. The laser array ranging unit is used to measure distances, obtaining a laser ranging data sequence. Simultaneously, the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device are invoked to acquire images, obtaining visible light image sequences and infrared image sequences. Intelligent AI multi-modal fusion is performed on the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multi-modal fused data features. Based on the multi-modal fused data features, target recognition is performed. The target recognition result is compared with the adaptive laser stealth protection zone to obtain a first warning message. Based on the first warning message, audible and visual warnings and voice broadcasts are activated. This invention addresses the technical problem of existing technologies' inability to monitor and warn of external damage risks to transmission lines in a real-time, accurate, and three-dimensional manner. By deploying multi-dimensional front-end monitoring devices on towers and utilizing laser array ranging, visible light and infrared image acquisition, and AI multimodal fusion technology, it achieves accurate identification and early warning linkage of external damage targets, thereby improving the real-time performance and accuracy of external damage monitoring of transmission lines.
[0058] Example 2, based on the same inventive concept as the dynamic monitoring method for preventing external damage to transmission lines in the aforementioned examples, such as... Figure 2 As shown, this application provides a dynamic monitoring system for preventing external damage to transmission lines. The system and method embodiments in this application are based on the same inventive concept. The system includes: The deployment module 11 is used to deploy a multi-dimensional front-end monitoring device on the tower of the target transmission line, and construct an adaptive laser stealth protection area through the laser array ranging unit of the multi-dimensional front-end monitoring device; the image acquisition module 12 is used to perform ranging using the laser array ranging unit to obtain a laser ranging data sequence, and call the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device to simultaneously perform image acquisition, and obtain a visible light image sequence and an infrared image sequence; the data fusion module 13 is used to perform intelligent AI multi-modal fusion of the laser ranging data sequence, the visible light image sequence and the infrared image sequence to obtain multi-modal fusion data features; the early warning module 14 is used to perform target recognition based on the multi-modal fusion data features, compare the target recognition result with the adaptive laser stealth protection area, obtain a first early warning information, and activate audible and visual warnings and voice broadcasts according to the first early warning information.
[0059] Furthermore, the system is also used to implement the following functions: The process involves: acquiring primary protection zone correlation indicators; identifying primary protection zones based on these indicators to obtain primary laser stealth protection zones to be deployed; retrieving the ranging log set of the laser array ranging unit within a preset historical window; adaptively identifying secondary protection zones based on the ranging log set to obtain secondary laser stealth protection zones to be deployed; deploying the secondary laser stealth protection zones outside the primary laser stealth protection zones to obtain adaptive laser stealth protection zones to be deployed; and feeding back the adaptive laser stealth protection zones to the control module of the laser array ranging unit for area deployment to obtain the adaptive laser stealth protection zones.
[0060] Furthermore, the system is also used to implement the following functions: The relevant indicators for the primary protection zone include the voltage level of the transmission line, clearance height, tower spacing, and laser equipment performance.
[0061] Furthermore, the system is also used to implement the following functions: The communication response delay, ranging confidence level, and relative speed of the ranging log set are extracted and traversed to construct a multi-dimensional protection reliability index set. The multi-dimensional protection reliability index set is filtered by mean shift to determine the selected multi-dimensional protection reliability indexes. The trigger frequency of the ranging log set is extracted and weighted with the selected multi-dimensional protection reliability indexes to obtain the secondary protection coefficient. The secondary protection area is adaptively identified based on the secondary protection coefficient to obtain the secondary laser stealth protection area to be deployed.
[0062] Furthermore, the system is also used to implement the following functions: Calculate the mean of the set of multidimensional protection reliability indicators to determine the initial mean drift screening center; according to the preset screening bandwidth, iteratively screen the initial mean drift screening center in the set of multidimensional protection reliability indicators to determine the screened multidimensional protection reliability indicators.
[0063] Furthermore, the system is also used to implement the following functions: The visible light image sequence and the infrared image sequence are semantically consistent and integrated to obtain a fused image semantic feature sequence; point cloud feature analysis is performed based on the laser ranging data sequence to obtain a laser point cloud feature sequence; intelligent AI multimodal linkage fusion is performed on the fused image semantic feature sequence and the laser point cloud feature sequence to obtain the multimodal fused data features.
[0064] Furthermore, the system is also used to implement the following functions: Semantic feature map recognition is performed on the visible light image sequence and the infrared image sequence using a semantic segmentation network layer to obtain a visible light semantic feature map sequence and an infrared semantic feature map sequence; the visible light semantic feature map sequence and the infrared semantic feature map sequence are then integrated by mapping attention weighted semantic consistency to obtain the fused image semantic feature sequence.
[0065] Furthermore, the system is also used to implement the following functions: The semantic branch is used to extract the semantic category trend and edge contour information trend of the semantic feature sequence of the fused image, and the geometric branch is used to extract the point cloud depth trend and spatial position trend of the laser point cloud feature sequence. The extraction results are then input into the linkage fusion unit to obtain the multimodal fused data features.
[0066] Furthermore, the system is also used to implement the following functions: A pre-built target recognizer identifies the features of the multimodal fusion data to obtain a target recognition result; it is then determined whether the target recognition result has entered the adaptive laser stealth protection area, and if so, a first warning message is obtained.
[0067] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0068] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
[0069] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.
Claims
1. A dynamic monitoring method for preventing external damage to transmission lines, characterized in that, The method includes: Multi-dimensional front-end monitoring devices are deployed on the towers of the target transmission line, and an adaptive laser stealth protection zone is constructed through the laser array ranging unit of the multi-dimensional front-end monitoring devices. The laser array ranging unit is used to measure the distance and obtain the laser ranging data sequence. The visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device are called to simultaneously acquire images and obtain visible light image sequence and infrared image sequence. Intelligent AI multimodal fusion is performed on the laser ranging data sequence, visible light image sequence, and infrared image sequence to obtain multimodal fusion data features; Target recognition is performed based on the multimodal fusion data features. The target recognition result is compared with the adaptive laser stealth protection area to obtain the first warning information. Based on the first warning information, sound and light warnings and voice broadcasts are activated.
2. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 1, characterized in that, Multi-dimensional front-end monitoring devices are deployed on the towers of the target transmission line. An adaptive laser stealth protection zone is constructed using the laser array ranging unit of these devices, including: Obtain the primary protection area association index, identify the primary protection area based on the primary protection area association index, and obtain the primary laser stealth protection area to be deployed. Retrieve the range measurement log set of the laser array ranging unit within a preset historical window; Based on the ranging log set, adaptive identification of the secondary protection area is performed to obtain the secondary laser stealth protection area to be deployed. The secondary laser stealth protection area to be deployed is placed outside the primary laser stealth protection area to be deployed, thereby obtaining an adaptive laser stealth protection area to be deployed. The adaptive laser stealth protection area to be deployed is fed back to the control module of the laser array ranging unit for area deployment, thereby obtaining the adaptive laser stealth protection area.
3. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 2, characterized in that, The relevant indicators for the primary protection zone include the voltage level of the transmission line, clearance height, tower spacing, and laser equipment performance.
4. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 2, characterized in that, Based on the ranging log set, adaptive identification of secondary protection areas is performed to obtain the secondary laser stealth protection areas to be deployed, including: The communication response delay, ranging confidence, and relative speed of the ranging log set are extracted by traversing the set to construct a multi-dimensional set of protection reliability indicators; The set of multidimensional protection reliability indicators is subjected to mean drift screening to determine the selected multidimensional protection reliability indicators; The trigger frequency of the ranging log set is extracted and weighted with the selected multi-dimensional protection reliability index to obtain the secondary protection coefficient. Based on the secondary protection coefficient, the secondary protection area is adaptively identified to obtain the secondary laser stealth protection area to be deployed.
5. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 4, characterized in that, The set of multidimensional protection reliability indicators is subjected to mean shift filtering to determine the selected multidimensional protection reliability indicators, including: Calculate the mean of the multidimensional protection reliability index set and determine the initial mean drift screening center; According to the preset screening bandwidth, the initial mean drift screening center is iteratively screened in the set of multidimensional protection reliability indicators to determine the screened multidimensional protection reliability indicators.
6. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 1, characterized in that, The laser ranging data sequence, visible light image sequence, and infrared image sequence are subjected to intelligent AI multimodal fusion to obtain multimodal fused data features, including: The visible light image sequence and the infrared image sequence are semantically consistent to obtain a fused image semantic feature sequence. Point cloud feature analysis is performed based on the laser ranging data sequence to obtain a laser point cloud feature sequence. The semantic feature sequence of the fused image and the laser point cloud feature sequence are subjected to intelligent AI multimodal linkage fusion to obtain the multimodal fused data features.
7. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 6, characterized in that, The visible light image sequence and the infrared image sequence are semantically consistent and integrated to obtain a fused image semantic feature sequence, including: The semantic segmentation network layer is used to perform semantic feature map recognition on the visible light image sequence and the infrared image sequence respectively to obtain the visible light semantic feature map sequence and the infrared semantic feature map sequence. The visible light semantic feature map sequence and the infrared semantic feature map sequence are integrated by mapping attention weighted semantic consistency to obtain the fused image semantic feature sequence.
8. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 7, characterized in that, The fused image semantic feature sequence and the laser point cloud feature sequence are subjected to intelligent AI multimodal linkage fusion to obtain the multimodal fused data features, including: The semantic branch is used to extract the semantic category trend and edge contour information trend of the semantic feature sequence of the fused image, and the geometric branch is used to extract the point cloud depth trend and spatial position trend of the laser point cloud feature sequence. The extraction results are then input into the linkage fusion unit to obtain the multimodal fused data features.
9. The dynamic monitoring method for preventing external damage to transmission lines as described in claim 1, characterized in that, Target recognition is performed based on the multimodal fusion data features. The target recognition result is compared with the adaptive laser stealth protection area to obtain first warning information, including: A pre-built target recognizer identifies the features of the multimodal fusion data to obtain target recognition results; Determine whether the target identification result has entered the adaptive laser stealth protection area; if so, obtain the first warning information.
10. A dynamic monitoring system for preventing external damage to transmission lines, characterized in that, The system is used to execute the dynamic monitoring method for preventing external damage to transmission lines as described in any one of claims 1-9, and the system includes: The deployment module is used to deploy multi-dimensional front-end monitoring devices on the towers of the target transmission line, and to construct an adaptive laser stealth protection zone through the laser array ranging unit of the multi-dimensional front-end monitoring device. The image acquisition module is used to measure distance using the laser array ranging unit, obtain laser ranging data sequence, and call the visible light camera unit and infrared monitoring unit of the multi-dimensional front-end monitoring device to simultaneously acquire images, obtain visible light image sequence and infrared image sequence; The data fusion module is used to perform intelligent AI multimodal fusion on the laser ranging data sequence, visible light image sequence and infrared image sequence to obtain multimodal fused data features; The early warning module is used to perform target recognition based on the multimodal fusion data features, compare the target recognition result with the adaptive laser stealth protection area, obtain first early warning information, and activate sound and light warnings and voice broadcasts according to the first early warning information.