Power cable infrared pattern temperature anomaly sensing method and device, equipment, storage medium

Abstract

This application discloses a method, apparatus, device, and storage medium for sensing infrared image temperature anomalies in power cables. The method determines the infrared feature map corresponding to the target power cable. Temperature gradient features are extracted from the infrared feature map to obtain a first enhanced feature map. Temperature gradient enhancement is applied to the first enhanced feature map to obtain a second enhanced feature map. An attention mechanism is applied to the second enhanced feature map to obtain a third enhanced feature map. Feature fusion is performed on the first, second, and third enhanced feature maps to obtain a multi-scale feature map. The anomaly detection result of the target power cable is determined based on the multi-scale feature map. This application introduces a feature extraction method using three modules: temperature gradient prior, multi-scale enhancement, and global statistical attention. This improves detection speed, response time, and data processing capabilities, while achieving high-precision, high-robustness, and real-time online sensing of power cable temperature anomalies.

Description

Technical Field

[0001] This application relates to the field of power detection, and includes, but is not limited to, a method, apparatus, device, and storage medium for sensing abnormal infrared graphic temperature of power cables. Background Technology

[0002] With the increasing proportion of renewable energy installed capacity in new power systems, thermal power plants, as the cornerstone of the power system, are playing a significantly more important role in peak and valley regulation. Power cables, as the power transmission channels of thermal power plants, are crucial for power production due to their safe and stable operation. Power cable faults can lead not only to unplanned shutdowns of generator units and plant-wide blackouts, but also to serious personal injury accidents. Power cable fires are a major cause of cable tunnel accidents. The insulation and sheathing layers of power cables are typically flammable, and once ignited, the fire spreads rapidly. Furthermore, the dense and complex cable laying in cable tunnels results in a wide-ranging fire, making repairs difficult and producing toxic gases such as hydrogen chloride; additionally, the hydrochloric acid formed by hydrogen chloride dissolving in water corrodes electrical equipment, creating safety hazards. In conclusion, as the carrier of electrical energy transmission between the power system and users, abnormal temperature rise in power cables not only affects the normal operation of the power system but can also cause fires, triggering large-scale power outages and resulting in huge economic losses. Summary of the Invention

[0003] In view of this, the infrared graphic temperature anomaly sensing method, apparatus, equipment, and storage medium for power cables provided in the embodiments of this application significantly improve the efficiency, accuracy, and reliability of the detection process.

[0004] The infrared graphic temperature anomaly sensing method, device, equipment, and storage medium for power cables provided in this application embodiment are implemented as follows: One aspect of this application provides a method for sensing abnormal infrared temperature patterns in power cables, the method comprising: Determine the infrared signature image corresponding to the target power cable; Temperature gradient features are extracted from the infrared feature map to obtain the first enhanced feature map; A temperature gradient enhancement is applied to the first enhanced feature map to obtain a second enhanced feature map; The second enhanced feature map is enhanced using an attention mechanism to obtain the third enhanced feature map; Feature fusion is performed on the first enhanced feature map, the second enhanced feature map, and the third enhanced feature map to obtain a multi-scale feature map; Anomaly detection results for target power cables are determined based on multi-scale feature maps.

[0005] In one possible implementation, temperature gradient feature extraction is performed on the infrared feature map to obtain a first enhanced feature map, including: The gradients of the infrared feature map in the horizontal and vertical directions are calculated using different gradient operators to obtain the first gradient feature map and the second gradient feature map. After concatenating the first and second gradient feature maps, a convolution process is performed to obtain the third gradient feature map. The third gradient feature map and the infrared feature map are extracted and concatenated to obtain the depth features of the image, resulting in the first enhanced feature map.

[0006] In one possible implementation, a temperature gradient enhancement is performed on the first enhanced feature map to obtain a second enhanced feature map, including: The fourth gradient feature map of the first enhanced feature map is calculated using the gradient operator; The fourth gradient feature map is multiplied element-wise with the first enhanced feature map to obtain the second enhanced feature map.

[0007] In one possible implementation, the second enhanced feature map is enhanced using an attention mechanism to obtain a third enhanced feature map, including: Calculate the global feature histogram of the second enhanced feature map to obtain temperature distribution statistics; Temperature distribution statistics are input into a multilayer perceptron for activation processing to obtain channel attention vectors. After weighting the second enhanced feature map using channel attention vectors, feature extraction is performed again to obtain the third enhanced feature map.

[0008] In one possible implementation, feature fusion is performed on the first enhanced feature map, the second enhanced feature map, and the third enhanced feature map to obtain a multi-scale feature map, including: Interpolation, concatenation, convolution, and attention mechanisms are applied to the first, second, and third enhanced feature maps to complete feature fusion and obtain three feature maps as multi-scale feature maps.

[0009] In one possible implementation, the anomaly detection results of the target power cable are determined based on multi-scale feature maps, including: Local features are extracted by performing convolution operations on multi-scale feature maps to obtain local feature maps. Classification and bounding box regression based on local feature maps are used to identify abnormal temperature regions of the target power cable and obtain anomaly detection results.

[0010] In one possible implementation, determining the infrared feature map corresponding to the target power cable includes: Acquire the infrared image of the receiving power cable corresponding to the target power cable; The infrared image of the power cable is resized and normalized to obtain an infrared feature map.

[0011] Another aspect of this application embodiment provides a power cable infrared graphic temperature anomaly sensing device, the device comprising: The feature map determination module is used to determine the infrared feature map corresponding to the target power cable; The first feature enhancement module is used to extract temperature gradient features from the infrared feature map to obtain the first enhanced feature map. The second feature enhancement module is used to perform temperature gradient enhancement on the first enhanced feature map to obtain the second enhanced feature map. The third feature enhancement module is used to enhance the second enhanced feature map using an attention mechanism to obtain the third enhanced feature map; The feature fusion module is used to fuse the first enhanced feature map, the second enhanced feature map and the third enhanced feature map to obtain a multi-scale feature map; The anomaly detection module is used to determine the anomaly detection results of the target power cable based on multi-scale feature maps.

[0012] In one possible implementation, the first feature enhancement module is further used for: The gradients of the infrared feature map in the horizontal and vertical directions are calculated using different gradient operators to obtain the first gradient feature map and the second gradient feature map. After concatenating the first and second gradient feature maps, a convolution process is performed to obtain the third gradient feature map. The third gradient feature map and the infrared feature map are extracted and concatenated to obtain the depth features of the image, resulting in the first enhanced feature map.

[0013] In one possible implementation, the second feature enhancement module is further used for: The fourth gradient feature map of the first enhanced feature map is calculated using the gradient operator; The fourth gradient feature map is multiplied element-wise with the first enhanced feature map to obtain the second enhanced feature map.

[0014] In one possible implementation, the third feature enhancement module is further used for: Calculate the global feature histogram of the second enhanced feature map to obtain temperature distribution statistics; Temperature distribution statistics are input into a multilayer perceptron for activation processing to obtain channel attention vectors. After weighting the second enhanced feature map using channel attention vectors, feature extraction is performed again to obtain the third enhanced feature map.

[0015] In one possible implementation, the feature fusion module is further used for: Interpolation, concatenation, convolution, and attention mechanisms are applied to the first, second, and third enhanced feature maps to complete feature fusion and obtain three feature maps as multi-scale feature maps.

[0016] In one possible implementation, the anomaly detection module is further used for: Local features are extracted by performing convolution operations on multi-scale feature maps to obtain local feature maps. Classification and bounding box regression based on local feature maps are used to identify abnormal temperature regions of the target power cable and obtain anomaly detection results.

[0017] In one possible implementation, the feature map determination module is further used for: Acquire the infrared image of the receiving power cable corresponding to the target power cable; The infrared image of the power cable is resized and normalized to obtain an infrared feature map.

[0018] The electronic device provided in this application includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the method described in this application.

[0019] The computer-readable storage medium provided in this application embodiment stores a computer program thereon, which, when executed by a processor, implements the method provided in this application embodiment.

[0020] In some embodiments, the method determines the infrared feature map corresponding to the target power cable. Temperature gradient features are extracted from the infrared feature map to obtain a first enhanced feature map. Temperature gradient enhancement is applied to the first enhanced feature map to obtain a second enhanced feature map. Attention mechanism enhancement is applied to the second enhanced feature map to obtain a third enhanced feature map. Feature fusion is performed on the first, second, and third enhanced feature maps to obtain a multi-scale feature map. The anomaly detection result of the target power cable is determined based on the multi-scale feature map. This application embodiment introduces a feature extraction method using three modules: temperature gradient prior, multi-scale enhancement, and global statistical attention. This improves detection speed, response time, and data processing capabilities, while achieving high-precision, high-robustness, and real-time online power cable temperature anomaly detection. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart is shown for a method for sensing abnormal temperature in an infrared graphic of a power cable according to an embodiment of this application. Figure 2 A schematic diagram of an infrared graphic temperature anomaly sensing system for power cables according to an embodiment of this application is shown. Figure 3 A schematic diagram of a temperature gradient feature extraction module according to an embodiment of this application is shown; Figure 4 A schematic diagram of a temperature gradient enhanced spatial pyramid pooling module according to an embodiment of this application is shown; Figure 5 A schematic diagram of an attention mechanism module for statistical anomalies according to an embodiment of this application is shown; Figure 6 A schematic diagram of a neck network module according to an embodiment of this application is shown; Figure 7 A schematic diagram of a detection head module according to an embodiment of this application is shown; Figure 8 A schematic diagram of an infrared graphic temperature anomaly sensing device for a power cable according to an embodiment of this application is shown. Figure 9 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0025] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0026] It should be noted that the terms "first, second, third" used in the embodiments of this application are used to distinguish similar or different objects and do not represent a specific order of objects. It can be understood that "first, second, third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0027] The infrared graphic temperature anomaly sensing method for power cables according to this application embodiment can be executed by any electronic device, including but not limited to mobile phones, wearable devices (such as smartwatches, smart bracelets, smart glasses, etc.), tablet computers, laptops, vehicle terminals, PCs (Personal Computers), etc. The functions implemented by this method can be achieved by a processor in the electronic device calling program code. Of course, the program code can be stored in a computer storage medium. Therefore, the electronic device includes at least a processor and a storage medium.

[0028] The infrared graphic temperature anomaly sensing method for power cables according to the embodiments of this application can be used in any application scenario that requires anomaly detection of power cables. For example, the embodiments of this application can be applied to monitor cable tunnels or mezzanines in thermal power plants, to monitor the surface temperature of power cables and control cables in the tunnel in real time, so as to detect abnormal temperature rise caused by overload, aging or poor contact at an early stage. Alternatively, it can also be applied to monitor cable terminations in substations, so as to identify local overheating caused by stress cone damage, insulation moisture, etc. by monitoring the temperature distribution of high-voltage cable terminations in real time.

[0029] Existing methods for detecting abnormal temperatures in power cables mainly include oscillating wave partial discharge detection and fiber optic sensing technology. However, oscillating wave partial discharge detection is only suitable for offline detection and has a long detection time. Fiber optic sensing technology suffers from issues such as detecting localized hotspots, slow response, and weak data processing capabilities.

[0030] Therefore, the technical problem solved by the embodiments of this application is how to improve detection efficiency and the accuracy of detection results, and reduce the workload of maintenance personnel.

[0031] The following describes in detail the infrared graphic temperature anomaly sensing scheme for power cables according to embodiments of this application, with reference to the accompanying drawings.

[0032] Figure 1 A flowchart illustrating a method for sensing abnormal temperature in an infrared graphic of a power cable according to an embodiment of this application is shown. Figure 1 As shown, the infrared graphic temperature anomaly sensing method for power cables in this application embodiment may include the following steps S10-S60.

[0033] For ease of description, the infrared graphic temperature anomaly sensing method for power cables according to embodiments of this application is described using an electronic device as the execution subject. It should be understood that the execution subject in embodiments of this application can also be a processor or chip in an electronic device, and this application does not impose any limitations.

[0034] Step S10: Determine the infrared feature map corresponding to the target power cable.

[0035] In one possible implementation, an infrared feature map of the target power cable requiring anomaly detection is acquired via an electronic device, and anomaly detection is performed on the target power cable based on the infrared feature map. Specifically, the electronic device acquires the infrared feature map of the target power cable when an anomaly detection mechanism is triggered, such as periodic anomaly detection or upon receiving an anomaly detection notification. The infrared feature map includes the temperature characteristics of the target power cable, which can be obtained by extracting the infrared image features of the corresponding power cable. That is, the electronic device can first acquire the infrared image of the corresponding power cable, and then perform image resizing and normalization processing on the power cable infrared image to obtain the infrared feature map.

[0036] Optionally, the infrared image of the power cable acquired by the electronic device can be obtained by an infrared thermal imager fixedly installed at key monitoring points (such as cable tunnels, joint wells, and substations). Further, the electronic device can adjust the size of the power cable infrared image to obtain an image of a preset size (e.g., 640×640 pixels) to match the input size of the anomaly detection algorithm. Furthermore, based on the temperature distribution characteristics of the power cable infrared image, the pixel temperature value T is mapped to the [0, 1] interval, and the normalization formula can be... , among which, T norm The pixel value representing the normalized pixel temperature value T, T min and T max These are used to characterize the lowest and highest temperatures in infrared images of power cables, respectively.

[0037] Step S20: Extract temperature gradient features from the infrared feature map to obtain the first enhanced feature map.

[0038] In one possible implementation, after determining the infrared feature map of the target power cable, the electronic device can perform temperature gradient feature extraction on the infrared feature map to obtain a first enhanced feature map. This temperature gradient feature extraction process is used to introduce temperature gradient information as prior knowledge into the entire anomaly detection process. Optionally, this temperature gradient feature extraction process may include calculating the horizontal and vertical gradients of the infrared feature map using different gradient operators to obtain a first gradient feature map and a second gradient feature map. The first and second gradient feature maps are then convolved to obtain a third gradient feature map. The third gradient feature map and the infrared feature map are extracted and concatenated to obtain the depth features of the image, resulting in the first enhanced feature map.

[0039] Step S30: Perform temperature gradient enhancement on the first enhanced feature map to obtain a second enhanced feature map.

[0040] In one possible implementation, after extracting temperature gradient features to obtain a first enhanced feature map, the electronic device can further enhance the first enhanced feature map with temperature gradients to obtain a second enhanced feature map. This temperature gradient enhancement process is more sensitive to regions with drastic temperature changes and integrates multi-scale information, ensuring that the anomaly detection algorithm can accurately capture anomaly boundaries and local hot spots. Optionally, this temperature gradient enhancement process may include calculating a fourth gradient feature map of the first enhanced feature map using a gradient operator. The fourth gradient feature map is then multiplied element-wise with the first enhanced feature map to obtain the second enhanced feature map.

[0041] Step S40: Apply attention mechanism enhancement to the second enhanced feature map to obtain the third enhanced feature map.

[0042] In one possible implementation, after the electronic device enhances the temperature gradient of the first enhanced feature map to obtain the second enhanced feature map, it further enhances the second enhanced feature map using an attention mechanism to obtain the third enhanced feature map. This attention mechanism enhancement process dynamically adjusts the weights through global statistical analysis of the feature maps, ensuring that the resulting third enhanced feature map is more capable of representing abnormal situations. Optionally, this attention mechanism enhancement process may include calculating the global feature histogram of the second enhanced feature map to obtain temperature distribution statistics. The temperature distribution statistics are then input into a multilayer perceptron for activation processing to obtain channel attention vectors. After weighting the second enhanced feature map using these channel attention vectors, feature extraction is performed again to obtain the third enhanced feature map.

[0043] Step S50: Perform feature fusion on the first enhanced feature map, the second enhanced feature map, and the third enhanced feature map to obtain a multi-scale feature map.

[0044] In one possible implementation, after determining the first, second, and third enhanced feature maps, the electronic device further fuses these three enhanced feature maps to obtain a multi-scale feature map. This process may include interpolation, concatenation, convolution, and attention mechanisms on the first, second, and third enhanced feature maps to achieve feature fusion and obtain three feature maps as the multi-scale feature map. This feature fusion process adjusts parameters such as the number of channels and size of the feature maps based on the three types of feature maps, providing the most ideal input conditions for anomaly detection in the algorithm.

[0045] The multi-scale feature map includes three parts, corresponding to the first enhanced feature map, the second enhanced feature map, and the third enhanced feature map, respectively.

[0046] Step S60: Determine the anomaly detection result of the target power cable based on the multi-scale feature map.

[0047] In one possible implementation, after the electronic device completes feature fusion to obtain a multi-scale feature map, it extracts and identifies feature information of various defects based on the multi-scale feature map to determine the anomaly detection result of the target power cable. This process can involve performing a convolution operation on the multi-scale feature map to extract local features, resulting in a local feature map. Then, based on the local feature map, classification and bounding box regression are performed to identify temperature anomaly regions of the target power cable and obtain the anomaly detection result.

[0048] Figure 2 A schematic diagram of an infrared graphic temperature anomaly sensing system for power cables according to an embodiment of this application is shown. Figure 2 As shown, in one possible implementation, the power cable infrared image temperature anomaly sensing method of this application embodiment can be integrated into a power cable infrared image temperature anomaly sensing system. This system can be a convolutional neural network including multiple algorithm processing modules. The convolutional neural network can include a backbone network module for feature extraction, a neck network module for feature fusion, and a detection head module for anomaly detection.

[0049] Optionally, the backbone network module may also include a temperature gradient feature extraction module, a temperature gradient enhancement spatial pyramid pooling module, and a statistical anomaly attention mechanism module, which are used to extract temperature gradient features, enhance temperature gradients, and enhance attention mechanisms, respectively, to obtain a first enhanced feature map, a second enhanced feature map, and a third enhanced feature map.

[0050] For example, the input layer receives infrared images of power cables and performs image resizing and normalization. The temperature gradient feature extraction module introduces temperature gradient information as prior knowledge into the network model. The temperature gradient enhanced spatial pyramid pooling module enables the network model to fuse multi-scale information and has higher sensitivity to areas with drastic temperature changes. The attention mechanism module for statistical anomalies dynamically adjusts weights through global statistical analysis of the feature maps. The neck network module receives the feature maps obtained from the backbone network module and performs feature fusion, adjusting parameters such as the number and size of the feature maps to provide the optimal input conditions for the detection head module. The detection head model then learns from a large amount of labeled data, extracts and identifies feature information of various defects, and obtains anomaly detection results.

[0051] Optionally, the input layer is responsible for receiving infrared images of power cables and performing image resizing, normalization, and other processing. The specific process includes processing the original infrared image of the power cable to obtain a 640×640 pixel image to match the input size of the pre-trained convolutional neural network; and using the min-max method to normalize the infrared feature map according to the temperature distribution characteristics of the infrared image.

[0052] In some embodiments, the temperature gradient feature extraction module introduces temperature gradient information as prior knowledge into the network model. Specifically, the process may involve using a parallel thermal gradient extractor with Sobel and Prewitt gradient operators to calculate the horizontal and vertical gradients of the input feature map, generating a first gradient feature map and a second gradient feature map. These are then concatenated along the channel dimension and processed using a 1×1 convolutional layer to match the number of channels C of the original feature map, resulting in a third gradient feature map. Finally, the standard C2f network uses the extracted third gradient feature map to directly access and utilize the boundary information of temperature changes, thus focusing more on key regions during the learning process to obtain the first enhanced feature map. The pyramid pooling module based on the temperature gradient enhancement space enables the network model to fuse multi-scale information and has higher sensitivity to regions with drastic temperature changes. Its temperature gradient enhancement process may include using the same thermal gradient extractor to calculate a fourth gradient feature map of the first enhanced feature map, using an activation function to ensure its non-negativity, and then modulating the first enhanced image through element-wise multiplication to obtain the second enhanced feature map. Based on the attention mechanism module of statistical anomalies, the embodiments of this application can perform global statistical analysis on the second enhanced feature map, dynamically adjust the weights, ensure that statistically deviated hot spots receive more attention, and improve the detection performance of the model in complex environments.

[0053] In some embodiments, the neck network module receives a first enhanced feature map, a second enhanced feature map, and a third enhanced feature map, and performs feature fusion. By adjusting parameters such as the number of channels and size of each feature map, it provides the most ideal input conditions for the detection head module. This process uses the neck network module as a key bridge connecting the backbone network module and the detection head module. The neck network module receives feature maps from the backbone network module and performs feature fusion. Through the FPN+PAN structure, it not only transfers deep semantic information to the shallow layer, enhancing the detection of small targets, but also transfers detailed features from the shallow layer to the deep layer, optimizing the localization accuracy of large targets. Finally, it outputs multi-scale feature maps for subsequent prediction by the detection head module.

[0054] Furthermore, the detection head module learns from a large amount of labeled data and extracts and identifies the feature information of various defects based on the input multi-scale feature map. Specifically, the detection head module is responsible for classifying, locating and optionally extending the multi-scale feature map output by the neck network to realize the perception of abnormal temperature of the power cable and obtain the abnormal detection result of the target power cable.

[0055] Based on the aforementioned deep neural network model structure, this application employs a deep neural network structure that integrates temperature gradient prior, multi-scale enhancement, and global statistical attention. This fundamentally addresses the problems of response lag, low detection efficiency, and weak data processing capabilities inherent in traditional detection methods. The temperature gradient feature extraction module accurately captures temperature change boundaries, while the temperature gradient enhancement spatial pyramid pooling module strengthens multi-scale sensitivity to local hot spots. Furthermore, the statistical anomaly attention mechanism dynamically focuses on anomalous features from the global distribution. These three elements synergistically significantly improve the model's accuracy and robustness in locating temperature anomalies in complex environments. Ablation experiments demonstrate that this structure improves many key detection indicators by approximately 10% compared to the baseline model, achieving high-precision, real-time online, and non-contact intelligent sensing. This significantly reduces the operational burden and provides reliable technical support for early warning of potential fire hazards in power cables.

[0056] Figure 3 A schematic diagram of a temperature gradient feature extraction module according to an embodiment of this application is shown. Figure 3 As shown, in this embodiment of the application, the temperature gradient feature extraction module calculates the input infrared feature map using the Sobel and Prewitt gradient operators respectively. The gradients in the horizontal and vertical directions yield the first and second gradient feature maps. Further, the first and second gradient feature maps are concatenated along the channel dimension and processed using a 1×1 convolutional layer to match the number of channels C of the original infrared feature map, resulting in the third gradient feature map. This third gradient feature map can be represented as... G represents the calculated third gradient feature map, Conv1×1 indicates a 1×1 convolutional layer, and Concat is channel concatenation. Sobel G is the first gradient feature map generated by the Sobel gradient operator. Prewitt The second gradient feature map generated by the Prewitt gradient operator is referred to as the Thermal Gradient Extractor. G is then concatenated with the original input infrared feature map along the channel dimension to form an enhanced feature representation. The expression can be Finally, using F through a standard C2f network. aug The extracted deep features are used to obtain the first enhanced feature map, which directly accesses and utilizes the boundary information of temperature changes. This allows the model to focus more on key areas during the learning process and improve the model's accuracy in locating the contours of temperature anomaly regions.

[0057] Based on the structure of the temperature gradient feature extraction module, the temperature gradient can be introduced into the network as explicit prior knowledge through parallel computation using Sobel and Prewitt operators. This allows the model to accurately focus on boundary regions with drastic temperature changes in the early stages of feature extraction, effectively improving the localization accuracy of abnormal contours. At the same time, the gradient information naturally suppresses background thermal noise, enhances the robustness of features, and provides high-quality gradient enhancement input for subsequent multi-scale pooling and attention modules. This synergistic effect with the overall network lays a key foundation for achieving high-precision temperature anomaly detection.

[0058] Figure 4 A schematic diagram of a temperature gradient enhanced spatial pyramid pooling module according to an embodiment of this application is shown. Figure 4 As shown, the temperature gradient enhancement spatial pyramid pooling module differs from the module that directly feeds the input first enhanced feature map into the pooling layer. Instead, it utilizes the same thermal gradient extractor as the temperature gradient feature extraction module to calculate the fourth gradient feature map G of the first enhanced feature map, accurately adjusting the weights of the input first enhanced feature map. Specifically, the fourth gradient feature map G is ensured to be non-negative using the activation function of the Rectified Linear Unit (ReLU), and then modulates the first enhanced feature map through element-wise multiplication to obtain the second enhanced feature map. , This indicates element-wise multiplication. This is the second enhanced feature map after gradient enhancement.

[0059] Based on the structural design of the temperature gradient-enhanced spatial pyramid pooling module, a gradient modulation mechanism is introduced before traditional multi-scale pooling. The gradient map is calculated using the same operator as the temperature gradient feature extraction module, and the original feature map is weighted element-wise. This makes the network more sensitive to areas with drastic temperature changes during multi-scale feature fusion. This design not only retains the advantages of spatial pyramid pooling in fusing multi-scale information, but also ensures that the model can accurately capture subtle features of local thermal anomalies under different receptive fields. It effectively solves the problem of missed detection caused by scale changes. Ablation experiments show that when used in conjunction with the attention mechanism, it can significantly improve various detection indicators and provide high-quality feature inputs with clear boundaries and prominent hotspots for the subsequent neck network and detection head.

[0060] Figure 5 This diagram illustrates an attention mechanism module for statistical anomalies according to an embodiment of this application. Figure 5As shown, the attention mechanism module for statistical anomalies first calculates the global feature histogram (Histogram) of the input second enhanced feature map, discretizes the continuous feature values ​​representing different temperature levels into preset intervals (bins), and calculates the global feature distribution using a one-dimensional vector, represented as follows: This includes statistical information on the overall temperature distribution of the image. Subsequently, a multilayer perceptron (MLP) learns h through training to identify "temperature anomaly" signals from specific distribution patterns, and generates channel attention vectors via a sigmoid activation function σ. , represented as Each value in this vector represents the importance of the corresponding channel; the higher the weight, the better the match between the channel's feature distribution pattern and the temperature anomaly signal learned by the MLP. Finally, this attention vector is applied to the input second enhanced feature map, adaptively enhancing the feature channels for temperature anomaly perception through channel weighting. Subsequently, the output feature map is fed into a C2f structure for processing, enabling the model to transcend the limitations of local receptive fields and, based on global statistical information, focus on the most suspicious features to obtain a third enhanced feature map, improving the model's ability to detect various forms of thermal anomalies.

[0061] Based on the structural design of the attention mechanism module for statistical anomalies, this embodiment of the application performs global histogram statistical analysis on the feature map, enabling the network to identify abnormal feature channels that deviate from the normal pattern from the perspective of the overall temperature distribution. It also uses a multilayer perceptron to dynamically generate channel attention weights, achieving adaptive enhancement of temperature anomaly-related features. This design breaks through the limitations of local receptive fields, enabling the model to accurately focus on statistically significant hotspot areas even in complex backgrounds (such as multiple intersecting cables and environmental thermal radiation interference), effectively suppressing background noise and significantly improving various detection indicators. Ablation experiments have shown that its contribution to detection accuracy and robustness is most prominent when used in conjunction with the gradient enhancement module, providing a feature selection capability under global statistical guidance for ultimately achieving high-confidence temperature anomaly perception.

[0062] Figure 6 A schematic diagram of a neck network module according to an embodiment of this application is shown. Figure 6 As shown, the neck network module in this embodiment receives a first enhanced feature map, a second enhanced feature map, and a third enhanced feature map, and performs feature fusion. By adjusting parameters such as the number of channels and size of each input feature map, a multimodal feature map corresponding to each feature map is obtained, providing the most ideal input conditions for the detection head.

[0063] Optionally, the backbone network module outputs three enhanced feature maps—a first enhanced feature map, a second enhanced feature map, and a third enhanced feature map—corresponding to different receptive fields and semantic levels, respectively. Using an FPN+PAN structure, through mathematical operations such as interpolation, concatenation, convolution, and attention mechanisms, the hierarchical features output by the backbone network module are reconstructed into a multi-scale feature set more suitable for the detection task, represented as follows: , Where O(x, y) is the pixel value of the corresponding output feature map, and I( f( ) is any input feature map pixel value, f( ) is the interpolation kernel, and s is the upsampling scaling factor. The upsampled corresponding feature map is concatenated with the enhanced feature map output by the backbone network at the same resolution level, fusing features from different levels, represented as... ,in, This is the spliced ​​multimodal feature map. and These are two feature maps that participate in the stitching process. H and W are the image dimensions, and C1 and C2 are the number of channels.

[0064] Based on the structure of the neck network module, a bidirectional fusion structure of FPN+PAN can be used as a bridge connecting the backbone network and the detection head. Its core advantage lies in achieving deep complementarity and enhancement of multi-scale features: the top-down path transmits deep semantic information to the shallow layer, improving the detection capability of small targets such as local hotspots; the bottom-up path feeds back shallow detailed features to the deep layer, optimizing the localization accuracy of large-area temperature rise regions. This bidirectional fusion mechanism, through upsampling, concatenation, and convolution operations, reconstructs feature maps that are both rich in semantic information and retain fine spatial structure, providing the detection head with the most ideal input conditions. This allows it to meet the perception needs of temperature anomalies at different scales in complex scenarios, and is a key support for improving the overall performance of the model.

[0065] Figure 7 A schematic diagram of a detection head module according to an embodiment of this application is shown. Figure 7 As shown, the detection head module of this application embodiment can extract and identify feature information of various defects by learning from a large amount of labeled data. This process extracts local features from the three input multimodal feature maps through convolution operations. The convolution operation uses a set of learnable convolution kernels to slide across the input feature maps, performing weighted summation and adding a bias to generate new feature maps. This is represented as... , where x is the input multimodal feature map, w is the convolution kernel, b is the bias, y is the output local feature map, (i, j) are the spatial coordinates of the output local feature map, k is the output channel index, (M, N) are the spatial dimensions of the convolution kernel, and C is the number of input channels.

[0066] Furthermore, standard convolution is decomposed into depthwise convolution and pointwise convolution. Depthwise convolution performs convolution on each input channel independently, while pointwise convolution combines the output channels of the depthwise convolution. This is represented as... The detection head module extracts features through a series of convolution operations, thereby realizing the anomaly detection of the target power cable in the image and obtaining the anomaly detection result.

[0067] In some embodiments, the performance of the system model can be evaluated using three metrics: MAP50, MAP75, and MAP5095. The calculation formulas for these metrics include... , and N represents the number of categories. Average precision is calculated for a single category, while mean average precision (MAP) is the average precision calculated for all categories. MAP50 is the average detection precision of the mean when the intersection-over-union (IoU) threshold is 0.5. It is one of the most commonly used evaluation metrics for object detection, with a relatively lenient evaluation standard. MAP75 is the average detection precision of the mean when the IoU threshold is 0.75. It requires a high degree of overlap between the predicted and ground truth boxes and is a more stringent evaluation metric. MAP5095 represents the average MAP calculated when the IoU threshold is between 0.5 and 0.95, with a step size within the range of 0.005. As one of the most stringent standards in the field of object detection, it can comprehensively reflect the performance of the system model.

[0068] This application embodiment can use the dataset to conduct 7 ablation experiments. Each experiment is performed to verify the dataset under the same parameters, and the comparison results of each model's indicators are obtained, as shown in Table 1. "√" indicates that the module is added to the network.

[0069] Table 1 Ablation Experiment Results

[0070] As shown in Table 1, the network model corresponding to the anomaly detection scheme in this application has more balanced performance, and the experimental results prove the advancement of the proposed algorithm.

[0071] Based on the aforementioned technical features, the anomaly detection scheme for power cables in this application introduces three innovative modules: temperature gradient feature extraction, temperature gradient enhanced spatial pyramid pooling, and statistical anomaly attention. This fundamentally solves the pain points of traditional detection methods, such as long detection time, slow response, and weak data processing capabilities. The method utilizes temperature gradient priors to accurately locate anomaly boundaries, combines a multi-scale enhancement mechanism to focus on local hot spots, and leverages global statistical attention to dynamically enhance anomaly feature channels. These three elements synergistically significantly improve the model's detection accuracy and robustness in complex environments. Simultaneously, this method achieves real-time online, non-contact intelligent sensing, greatly reducing the workload of maintenance personnel and providing reliable technical support for early warning of power cable fire hazards. It has significant engineering application value for ensuring the safe operation of power systems.

[0072] It should be understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0073] Based on the foregoing embodiments, this application provides an infrared graphic temperature anomaly sensing device for power cables. The device includes various modules and units included in each module, which can be implemented by a processor; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), microprocessor (MPU), digital signal processor (DSP), or field programmable gate array (FPGA), etc.

[0074] Figure 8 A schematic diagram of an infrared graphic temperature anomaly sensing device for power cables according to an embodiment of this application is shown. Figure 8 As shown, the infrared graphic temperature anomaly sensing device for power cables in this embodiment of the application includes: The feature map determination module 80 is used to determine the infrared feature map corresponding to the target power cable; The first feature enhancement module 81 is used to extract temperature gradient features from the infrared feature map to obtain the first enhanced feature map. The second feature enhancement module 82 is used to perform temperature gradient enhancement on the first enhanced feature map to obtain the second enhanced feature map; The third feature enhancement module 83 is used to enhance the second enhanced feature map using an attention mechanism to obtain the third enhanced feature map; The feature fusion module 84 is used to fuse the first enhanced feature map, the second enhanced feature map and the third enhanced feature map to obtain a multi-scale feature map. Anomaly detection module 85 is used to determine the anomaly detection results of the target power cable based on multi-scale feature maps.

[0075] In one possible implementation, the first feature enhancement module 81 is further used for: The gradients of the infrared feature map in the horizontal and vertical directions are calculated using different gradient operators to obtain the first gradient feature map and the second gradient feature map. After concatenating the first and second gradient feature maps, a convolution process is performed to obtain the third gradient feature map. The third gradient feature map and the infrared feature map are extracted and concatenated to obtain the depth features of the image, resulting in the first enhanced feature map.

[0076] In one possible implementation, the second feature enhancement module 82 is further used for: The fourth gradient feature map of the first enhanced feature map is calculated using the gradient operator; The fourth gradient feature map is multiplied element-wise with the first enhanced feature map to obtain the second enhanced feature map.

[0077] In one possible implementation, the third feature enhancement module 83 is further used for: Calculate the global feature histogram of the second enhanced feature map to obtain temperature distribution statistics; Temperature distribution statistics are input into a multilayer perceptron for activation processing to obtain channel attention vectors. After weighting the second enhanced feature map using channel attention vectors, feature extraction is performed again to obtain the third enhanced feature map.

[0078] In one possible implementation, the feature fusion module 84 is further used for: Interpolation, concatenation, convolution, and attention mechanisms are applied to the first, second, and third enhanced feature maps to complete feature fusion and obtain three feature maps as multi-scale feature maps.

[0079] In one possible implementation, the anomaly detection module 82 is further used for: Local features are extracted by performing convolution operations on multi-scale feature maps to obtain local feature maps. Classification and bounding box regression based on local feature maps are used to identify abnormal temperature regions of the target power cable and obtain anomaly detection results.

[0080] In one possible implementation, the feature map determination module 80 is further used for: Acquire the infrared image of the receiving power cable corresponding to the target power cable; The infrared image of the power cable is resized and normalized to obtain an infrared feature map.

[0081] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0082] It should be noted that, in the embodiments of this application... Figure 8 The module division of the infrared graphic temperature anomaly sensing device for power cables shown is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, exist as separate physical units, or be integrated into one unit with two or more units. The integrated units can be implemented in hardware, as software functional units, or a combination of both.

[0083] It should be noted that, in the embodiments of this application, if the above-described methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0084] Figure 9 A schematic diagram of an electronic device according to an embodiment of this application is shown. For example... Figure 9 As shown in the figure, this application provides an electronic device, which can be a server, and its internal structure diagram can be as follows. Figure 9As shown, the electronic device includes a processor 920, a memory, and a transceiver 940 connected via a system bus 910. The processor 920 provides computing and control capabilities. The memory includes a non-volatile storage medium 931 and internal memory 932. The non-volatile storage medium 931 stores an operating system, computer programs, and a database. The internal memory 932 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium 931. The database stores data. The transceiver 940 communicates with an external terminal via a network connection. The computer program is executed by the processor 920 to implement the aforementioned methods.

[0085] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor 920, implements the steps of the method provided in the above embodiments.

[0086] This application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the steps in the method provided in the above-described method embodiments.

[0087] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0088] In one possible implementation, the apparatus provided in this application can be implemented as a computer program, which can be configured as follows: Figure 9 The device operates on the electronic device shown. The memory of the electronic device can store various program modules that make up the above-described apparatus. The computer program composed of the various program modules causes the processor 920 to execute the steps of the methods in the various embodiments of this application described in this specification.

[0089] It should be noted that the descriptions of the storage medium and device embodiments above are similar to the descriptions of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium, storage medium, and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0090] It should be understood that the phrases "one embodiment," "an embodiment," or "some embodiments" mentioned throughout the specification mean that a specific feature, structure, or characteristic related to an embodiment is included in at least one embodiment of this application. Therefore, phrases such as "in one possible implementation," "in one embodiment," or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments; their similarities or commonalities can be referred to mutually, and for the sake of brevity, they will not be repeated here.

[0091] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three kinds of relationships. For example, object A and / or object B can represent three situations: object A exists alone, object A and object B exist simultaneously, and object B exists alone.

[0092] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or modules can be electrical, mechanical, or other forms.

[0094] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0095] In addition, each functional module in the various embodiments of this application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the integrated modules can be implemented in hardware or in the form of hardware plus software functional units.

[0096] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0097] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0098] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0099] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0100] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0101] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A power cable infrared pattern temperature anomaly sensing method, characterized by, The method includes: Determine the infrared signature image corresponding to the target power cable; Temperature gradient features are extracted from the infrared feature map to obtain a first enhanced feature map; The first enhanced feature map is enhanced by a temperature gradient to obtain a second enhanced feature map. The second enhanced feature map is enhanced using an attention mechanism to obtain the third enhanced feature map; The first enhanced feature map, the second enhanced feature map, and the third enhanced feature map are fused to obtain a multi-scale feature map. The anomaly detection results of the target power cable are determined based on the multi-scale feature map.

2. The method of claim 1, wherein, The step of extracting temperature gradient features from the infrared feature map to obtain a first enhanced feature map includes: The gradients in the horizontal and vertical directions of the infrared feature map are calculated using different gradient operators to obtain the first gradient feature map and the second gradient feature map. After concatenating the first gradient feature map and the second gradient feature map, a convolution process is performed to obtain the third gradient feature map. The depth features of the image are obtained by extracting the third gradient feature map and the infrared feature map and concatenating them to obtain the first enhanced feature map.

3. The method according to claim 1, characterized in that, The step of performing temperature gradient enhancement on the first enhanced feature map to obtain the second enhanced feature map includes: The fourth gradient feature map of the first enhanced feature map is calculated using the gradient operator; The fourth gradient feature map is multiplied element-wise with the first enhanced feature map to obtain the second enhanced feature map.

4. The method according to claim 1, characterized in that, The process of enhancing the second enhanced feature map using an attention mechanism to obtain a third enhanced feature map includes: Calculate the global feature histogram of the second enhanced feature map to obtain temperature distribution statistics; The temperature distribution statistics are input into a multilayer sensor for activation processing to obtain the channel attention vector. After weighting the second enhanced feature map using the channel attention vector, feature extraction is performed again to obtain the third enhanced feature map.

5. The method according to claim 1, characterized in that, The feature fusion of the first enhanced feature map, the second enhanced feature map, and the third enhanced feature map to obtain a multi-scale feature map includes: The first, second, and third enhanced feature maps are processed by interpolation, concatenation, convolution, and attention mechanisms to complete feature fusion and obtain three feature maps as multi-scale feature maps.

6. The method according to claim 1, characterized in that, The determination of the anomaly detection result of the target power cable based on the multi-scale feature map includes: The multi-scale feature map is convolved to extract local features, resulting in a local feature map. Based on the local feature map, classification and bounding box regression are performed to identify the temperature anomaly region of the target power cable and obtain anomaly detection results.

7. The method according to claim 1, characterized in that, The determination of the infrared feature map corresponding to the target power cable includes: Acquire the infrared image of the receiving power cable corresponding to the target power cable; The infrared image of the power cable is resized and normalized to obtain an infrared feature map.

8. A power cable infrared graphic temperature anomaly sensing device, characterized in that, The device includes: The feature map determination module is used to determine the infrared feature map corresponding to the target power cable; The first feature enhancement module is used to extract temperature gradient features from the infrared feature map to obtain a first enhanced feature map; The second feature enhancement module is used to perform temperature gradient enhancement on the first enhanced feature map to obtain the second enhanced feature map; The third feature enhancement module is used to enhance the second enhanced feature map using an attention mechanism to obtain the third enhanced feature map; The feature fusion module is used to fuse the first enhanced feature map, the second enhanced feature map and the third enhanced feature map to obtain a multi-scale feature map. An anomaly detection module is used to determine the anomaly detection result of the target power cable based on the multi-scale feature map.

9. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images