Power service safety management method, device, equipment, medium and program

Abstract

The invention discloses an electric power service safety management method, device and equipment, a medium and a program. The method comprises the following steps: processing a received original image of the power service to obtain a corresponding filtering image; determining the attention weight of the filtered image, and based on the attention weight, determining an enhanced feature related to the power behavior in the filtered image; performing feature fusion on the enhanced features to obtain fused features; and performing service security detection on the electric power service based on the fusion feature, and reporting risk information when an obtained detection result is that a security risk exists. According to the method, the problems of low detection accuracy and high safety management cost of the power service in the existing scheme can be solved only by focusing on the enhanced features related to the power behavior, the detection accuracy of the power service is improved, and the calculation cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power safety, and in particular to a power service safety management method, device, equipment, medium and program. BACKGROUND

[0002] In the field of power safety, the safe and normative implementation of substation equipment operation and maintenance, line repair, and user end service can ensure the stable operation of the power system and the safety of personnel life. Therefore, safety management for power service is a necessary means to ensure the safety of the power system.

[0003] The safety management of power service relies on the detection of power service violations (such as operating without a ticket, not wearing insulating equipment, and violating high-voltage equipment, etc.). However, in the prior art, a general detection model is usually used to identify power service violations, which not only has low detection accuracy for power service violations, but also has the problems of large model parameter quantity and high computational complexity, resulting in high cost of safety management of power service.

[0004] Therefore, how to improve the detection accuracy and reduce the cost of power service safety management is a problem to be solved at present. SUMMARY

[0005] The present application provides a power service safety management method, device, equipment, medium and program, which can focus on the enhanced features of the filtered image by filtering the attention weight of the image, thereby improving the detection accuracy and reducing the computational cost.

[0006] According to a first aspect of the present application, a power service safety management method is provided, the method comprising:

[0007] processing a received original image of power service to obtain a corresponding filtered image;

[0008] determining the attention weight of the filtered image, and based on the attention weight, determining the enhanced features related to power behavior in the filtered image;

[0009] performing feature fusion on the enhanced features to obtain a fused feature;

[0010] performing service safety detection on the power service based on the fused feature, and when the obtained detection result is that there is a safety risk, reporting risk information.

[0011] According to a second aspect of the present application, a power service safety management device is provided, the device comprising:

[0012] a processing module configured to process a received original image of power service to obtain a corresponding filtered image;

[0013] determining a focus weight of the filtered image, and determining an enhanced feature related to the power behavior in the filtered image based on the focus weight;

[0014] fusing the enhanced feature to obtain a fused feature;

[0015] detecting service security of the power service based on the fused feature, and reporting risk information when a detection result obtained is that there is a security risk.

[0016] According to a third aspect of the present application, an electronic device is provided, which comprises:

[0017] at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the power service security management method according to any one of the embodiments of the present application.

[0018] According to a fourth aspect of the present application, a computer readable storage medium is provided, which stores computer instructions for enabling a processor to execute the power service security management method according to any one of the embodiments of the present application when executed by the processor.

[0019] According to a fifth aspect of the present application, a computer program product is provided, which comprises a computer program for enabling a processor to execute the power service security management method according to any one of the embodiments of the present application when executed by the processor.

[0020] The technical solution of the present application can process the received original image of the power service to obtain a corresponding filtered image, then determine the focus weight of the filtered image, and determine the enhanced feature related to the power behavior in the filtered image based on the focus weight, finally fuse the enhanced feature to obtain a fused feature, and detect the service security of the power service based on the fused feature, and report risk information when a detection result obtained is that there is a security risk. The present application can focus on the enhanced feature related to the power behavior to solve the problems of low detection accuracy and high security management cost of the existing solution for the power service, improve the detection accuracy of the power service, and reduce the calculation cost.

[0021] It is to be understood that the embodiments described herein are merely exemplary of the application and that a myriad of modifications, both as to the nature and number of elements within the execution of the application and as to the modes of execution thereof, can be made by those skilled in the art, without expressly quantifying the application and without departing from the scope of the application. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiments description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0023] Figure 1 is a flow chart of a power service security management method provided by an embodiment of the present application;

[0024] Figure 2 is a flow chart of a power service security management method provided by an embodiment of the present application;

[0025] Figure 3 is a structural schematic diagram of a power service security management device provided by an embodiment of the present application;

[0026] Figure 4 is a structural schematic diagram of an electronic device for implementing a power service security management method of an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the technical personnel in the art better understand the present application, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0028] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment One

[0030] Figure 1 A flowchart of a power service safety management method provided by Embodiment One of the present application, the embodiment can be applicable to real-time monitoring and risk warning of illegal behavior in a power service scenario. The method can be executed by a power service safety management device, which can be realized in the form of hardware and / or software and can be configured in a power service center platform or an edge computing device. As shown in the figure, the method comprises: Figure 1

[0031] S101, processing the received original image of the power service to obtain a corresponding filtered image.

[0032] The original image of the power service can be an unprocessed video frame or static image obtained by an image acquisition device (such as a camera or inspection robot) in a power service scenario (such as substation inspection, power repair, user operation, etc.), which can contain visual data such as power equipment status, personnel operation behavior, and environmental information. The filtered image can be a denoised image obtained after image preprocessing.

[0033] It should be noted that the image acquisition device can transmit the video frame or static image of the power service scenario to the power service management device in real time after collecting it, so as to monitor the power behavior of the power service in real time. However, due to the complex power site environment, the original image collected by the camera often contains a large amount of noise, which will interfere with subsequent feature extraction and affect the accuracy of illegal behavior determination. In order to further ensure the detection accuracy and robustness, the original image can be denoised first to remove the noise related to the illumination fluctuation of the power scene (such as the transition from cloudy to sunny) and the shadow of the equipment.

[0034] For example, after receiving the original image of the power service sent by the image acquisition device, the original image can be scaled to a uniform size by pixel normalization, and the normalized original image can be denoised by using a preset filtering algorithm to generate a filtered image.

[0035] Optionally, the processing of the received original image of the power service to obtain a corresponding filtered image can further comprise:

[0036] standardizing the original image to obtain a standardized image;

[0037] using adaptive Gaussian filtering to denoise the standardized image to obtain the filtered image.

[0038] ​For example, after collecting video / image data from a power service scenario, the set of original images can be obtained as: D = {I1, I2, ..., I...} N}, I i ∈R H×W×3 Among them, I i It can be the original image of the i-th frame; I i ∈R H×W×3 This can represent that each original image in the set of original images belongs to a three-dimensional array with dimensions H×W×3, where H can be the number of pixels in the height direction, W can be the number of pixels in the width direction, 3 can represent the number of RGB channels, and N can be the total amount of data in the original image.

[0039] Furthermore, by standardizing the original image, a standardized image can be obtained, thereby eliminating the difference in pixel value magnitude. Specifically:

[0040] ;

[0041] in, It can represent the pixel values ​​of a standardized image; It can represent the pixel values ​​of the original image; (x, y) can be pixel coordinates; c∈{1, 2, 3} can represent color channel values, μ c It can be the global mean of channel c; σ c It can be the global standard deviation of channel c.

[0042] Finally, adaptive Gaussian filtering can be used to denoise the normalized image, resulting in the filtered image I. i denoise :

[0043] ;

[0044] in, It can represent the pixel value of the i-th frame of the filtered image; k can be the size of the filter kernel; It can be an adaptive Gaussian kernel; σ(x,y) is the adaptive standard deviation of the neighborhood of pixel (x,y); It can be the pixel value of the normalized image of the i-th frame in the neighborhood range; (x+u, y+v) can represent the neighborhood range of the pixel (x,y).

[0045] S102. Determine the attention weights of the filtered image, and based on the attention weights, determine the enhancement features related to power behavior in the filtered image.

[0046] Attention weights can be used to automatically identify and highlight key features in filtered images that are related to power-related behaviors. Power-related behaviors can include the tools, safety equipment, or actions of personnel involved in power services.

[0047] It should be noted that raw images collected at power service sites can include various features such as the status of power equipment, personnel operating behavior, and environmental information. Existing technologies typically use detection models to extract and analyze the feature information of the entire image to determine whether violations exist during power service operations. This results in a significant waste of computational resources on background features unrelated to the core task. Furthermore, interference from irrelevant features can mask weak but crucial signals for determining violations, leading to problems of both high computational cost and low detection accuracy. For example, when determining whether a person is "not wearing a safety helmet," background features such as the sky, trees, or buildings are irrelevant to the judgment and may even interfere.

[0048] In order to reduce the computational load of image feature extraction and analysis, and to improve the detection accuracy of violations in power services, this embodiment can use attention weights to focus on features related to power behavior in the image and suppress irrelevant feature information, thereby reducing computational costs and improving detection accuracy.

[0049] Attention weights can be pre-set according to the safety inspection requirements of power services (for example, for the inspection requirement of wearing a safety helmet in power services, the top area of ​​the image can be set to high attention weight, while the middle and bottom areas of the image can be set to low attention weight; for the inspection requirement of insulating gloves in power services, the middle area of ​​the image can be set to high attention weight, while the top and bottom areas of the image can be set to low attention weight).

[0050] Building upon this, enhanced features can be feature maps weighted by attention weights, which can focus more on visual features related to violations in power services. It's understandable that multiple safety inspection needs in power services can coexist. For example, the safety management process of power services may simultaneously require detecting whether relevant personnel are wearing safety helmets and whether they are in high-voltage danger zones. In this case, enhanced features can include two feature maps: one focusing on helmet wearing and the other on personnel in high-voltage danger zones.

[0051] For example, after obtaining the filtered image, the attention weight of the filtered image can be determined by the detection requirements of power services, and the attention weight can be weighted with the feature information in the filtered image to enhance the features related to power behavior in the filtered image.

[0052] S103. Perform feature fusion on the enhanced features to obtain fused features.

[0053] Among them, the fusion feature can be a feature map or feature vector obtained by uniformly fusing feature maps of different scales and dimensions in the enhanced features.

[0054] It is understandable that, taking the safety management process of power services as an example, which simultaneously requires detecting whether relevant personnel are wearing safety helmets and whether they are in high-voltage danger zones, the feature maps included in the enhanced features, which focus on helmet wearing and personnel in high-voltage danger zones, are at different scales ("safety helmet" is a local small-scale feature map, while "personnel in the area" is a large-scale feature map in a macroscopic scene). To integrate multi-scale feature information and more comprehensively describe the visual content of the power service scenario, this embodiment can perform feature fusion to obtain fused features.

[0055] For example, the feature maps in the enhanced features can be input into a preset feature fusion model to stitch feature maps of different scales together to obtain fused features.

[0056] S104. Based on the fusion features, perform service security detection on the power service, and when the detection result indicates that there is a security risk, report the risk information.

[0057] For example, compliance analysis of power service behavior can be performed by fusing features to identify whether there are any violations or security risks. When violations or security risks are detected, information such as the type, time, location, and severity of the violation can be encapsulated and transmitted to the cloud to enable timely warnings and responses.

[0058] One specific method for compliance analysis of electricity service behavior is as follows: After determining the fusion features, input these features into a pre-trained classifier, which then outputs the probability values ​​for various types of violations. If the probability of a certain type of violation exceeds a preset threshold, it is determined to pose a security risk. It should be noted that the detection process can be fully automated, requiring no manual intervention, and supports real-time processing, meeting the millisecond-level response requirements of electricity service scenarios.

[0059] It should be noted that a classifier typically consists of one or more fully connected layers and a softmax output layer. The fully connected layers are responsible for mapping the fused features to the violation category space, while the softmax function normalizes the output to a probability distribution. Understandably, this embodiment provides not only the detection result but also its confidence level, offering richer information for subsequent risk assessment and decision-making. For example, when the confidence level is low, the detection result can be labeled for manual review, thereby improving the reliability of security management.

[0060] The technical solution of this embodiment can process the original image of the received power service to obtain a corresponding filtered image, then determine the attention weight of the filtered image, and based on the attention weight, determine the enhanced features related to power behavior in the filtered image. Finally, the enhanced features are fused to obtain fused features, and the power service is used for service security detection based on the fused features. When the obtained detection result indicates that there is a security risk, the risk information is reported. By focusing only on the enhanced features related to power behavior, the solution can solve the problems of low detection accuracy and high security management cost of existing solutions for power services, thereby improving the detection accuracy of power services and reducing the computational cost.

[0061] Optionally, based on the above embodiments, the present invention also provides an optional embodiment that optimizes the process of obtaining fused features from the fused enhanced features in the above embodiments. The step of fusing the enhanced features to obtain fused features may include:

[0062] Determine the resolution of the enhanced feature;

[0063] Based on the resolution of the enhanced features, the enhanced features are fused across scales to obtain the fused features.

[0064] It should be noted that the feature maps involved in the enhancement features typically have multiple different resolutions. High-resolution feature maps have rich spatial information, while low-resolution feature maps have stronger detail information. Directly stitching together or using features of different resolutions can lead to information misalignment and difficulties in fusion. Therefore, this optional embodiment can integrate and complement the information of enhancement features with different resolutions, comprehensively utilizing feature maps of different resolutions to generate a fused feature representation that combines localization accuracy and classification confidence.

[0065] For example, the resolution of the augmented features can be determined based on the number of pixels contained in the feature map in the height and width directions, and then augmented features at different scales can be adjusted to a uniform intermediate resolution. For instance, for low-resolution features, upsampling operations can be used to increase their spatial size. For high-resolution features, strided convolution or pooling operations can be used for appropriate downsampling.

[0066] Subsequently, these aligned features can be aggregated by feature concatenation or element-wise addition. The aggregated features can then be recalibrated and dimensionality reduced by convolution operations to eliminate conflicts between features of different scales and unify the number of channels, thereby forming a fusion feature with complete information and a unified structure.

[0067] The advantage of this setup is that it can effectively improve the detection of small targets (such as unworn gloves) and enhance the overall perception of violations at different scales.

[0068] Example 2

[0069] Figure 2 This is a flowchart of a power service safety management method provided in Embodiment 2 of the present invention. This embodiment can further optimize step S102 of Embodiment 1, which involves determining the attention weights of the filtered image and, based on the attention weights, determining the enhancement features of the filtered image. For example... Figure 2 As shown, the method includes:

[0070] S201. Process the received original image of the power service to obtain the corresponding filtered image.

[0071] S202. Extract the multi-scale features of the filtered image.

[0072] Among them, multi-scale features can be multiple feature representations extracted from the same filtered image, with different spatial resolutions and different levels of information.

[0073] For example, an improved MobileNetV3 can be used as the backbone network to extract multi-scale features from filtered images to balance detection accuracy and computational efficiency. MobileNetV3 is a depthwise separable convolution, which decomposes standard convolution into depthwise convolution and pointwise convolution, significantly reducing computational cost and the number of parameters, making it suitable for resource-constrained power edge applications.

[0074] Extracted multi-scale features F S Taking three scale feature maps F1, F2, and F3 as an example, F S ={F1,F2,F3}, where F1∈R H1×W1×C1 F2∈R H2×W2×C2 F3∈R H3×W3×C3 H, W, and C can be the pixel height, width, and number of color channels of the filtered image, respectively, while H1=H / 8, W2=W / 8, H2=H / 16, W2=W / 16, H3=H / 32, W3=W / 32, and C1, C2, and C3 can be the number of feature channels.

[0075] Multiscale features F S The extraction process can be:

[0076] ;

[0077] Wherein, MobileNetV3(,;s) can represent the s-th feature output branch of the backbone network.

[0078] S203. The multi-scale features are pooled and convolutionally processed to obtain the attention weights.

[0079] It should be noted that in practical applications, if attention weights are pre-set according to the safety detection requirements of power services, then these attention weights will be a fixed set of values ​​when the model is trained and deployed. However, power service scenarios are complex and varied; violations can occur anywhere in an image, and their visual features (such as safety helmets and insulated gloves) are not static. For example, when a person crouches down, the safety helmet may appear in the center of the image. Therefore, using pre-set static weights cannot adaptively focus on different key points in different images, which will severely limit the model's detection accuracy and generalization ability.

[0080] To address the aforementioned issues, this embodiment can also dynamically generate appropriate attention weights based on the multi-scale feature maps of the filtered image, thus meeting the complex and ever-changing needs of power service scenarios. For each filtered image, the most suitable weights can be calculated in real time based on its actual content, enabling it to flexibly respond to various uncertainties and changes in the power field.

[0081] For example, taking the calculation of channel attention weights for a filtered image as an example, this can be achieved by applying multi-scale features F... S Pooling and convolution operations are performed to obtain the channel attention weights W. C The specific formula is as follows:

[0082] ;

[0083] Among them, Conv 1×1 It can be a 1×1 convolution; σ() can be an activation function; GlobalAvgPool(F S )) can be global average pooling.

[0084] It should be noted that this process can compress the two-dimensional feature map of each channel into a scalar using global average pooling, representing the global distribution of features on that channel. Subsequently, 1x1 convolutions are used to first compress the number of channels, significantly reducing the number of parameters and computational cost. Then, the information is restored and integrated, achieving lightweight computation. Finally, the sigmoid function is used for normalization, serving as the importance weight for each channel.

[0085] S204. The multi-scale features are processed using the attention weights to obtain enhanced features related to power behavior in the filtered image.

[0086] For example, after determining the attention weights, element-wise multiplication can be performed between the attention weights and multi-scale features. This amplifies high-weight features in the output and weakens low-weight features, resulting in an enhanced feature map. It's important to note that element-wise multiplication does not change the structure of the feature map; rather, it recalibrates the importance of each feature element.

[0087] Specifically, taking the attention weights of the filtered image as channel attention weights as an example, when determining the channel attention weights W... C Then, the enhancement features can be determined as follows:

[0088] ;

[0089] in, Channel attention-based W C Enhanced features obtained by strengthening key channel features; ⊙ can be element-wise multiplication.

[0090] It should be noted that, to achieve more refined feature extraction, attention weights can include channel attention weights and spatial attention weights. Channel attention focuses more on clustering important features, while spatial attention focuses more on clustering important regions. Combining the two allows for comprehensive and refined feature extraction from both channel and spatial dimensions.

[0091] Accordingly, processing the multi-scale features using the attention weights to obtain enhanced features related to power behavior in the filtered image may include:

[0092] By processing the multi-scale features using the channel attention weights, the key features of the filtered image related to power behavior in the channel dimension are obtained.

[0093] By processing the key features using the spatial attention weights, enhanced features related to power behavior in the channel and spatial dimensions of the filtered image are obtained.

[0094] Channel attention weights can be used to evaluate the importance of each channel in a feature map, determining which feature channels contain more critical information. Spatial attention weights can be used to evaluate the importance of each spatial location in a feature map, determining which regions in the feature map are more critical.

[0095] For example, after determining the channel attention weights based on multi-scale features, the channel attention weights and multi-scale features can be multiplied element-wise to generate key features of the filtered image related to power behavior in the channel dimension. Then, spatial attention weights can be further determined based on the key features, and the key features can be multiplied element-wise with the spatial attention weights to further focus on the local regions in the image most relevant to the violation, thus obtaining enhanced features of the filtered image related to power behavior in both the channel and spatial dimensions.

[0096] Understandably, this embodiment can first filter the feature channels using channel attention, removing a large amount of irrelevant noise. Then, spatial attention can perform finer-grained spatial selection based on this, thereby further reducing the computational burden.

[0097] Specifically, the process of determining channel attention weights based on multi-scale features and generating key features of the filtered image related to power behavior in the channel dimension by performing element-wise multiplication of the channel attention weights and multi-scale features has already been described above, and will not be repeated here. The process of determining spatial attention weights based on key features can be as follows:

[0098] ;

[0099] Among them, W S It could be spatial attention weights; Key features can be obtained by enhancing key channel features based on channel attention WC; Maxpool can be the maximum pooling operation; AvgPool can be the average pooling operation; Cat() is the channel concatenation function; Conv 3×3 It is a 3×3 convolution; σ() can be an activation function.

[0100] It is understandable that the enhancement features of the final filtered image related to power behavior in both the channel and spatial dimensions can be:

[0101] ;

[0102] in, It can provide enhanced features related to electrical behavior in both the channel and spatial dimensions of the filtered image; Key features can be obtained by enhancing key channel features based on channel attention; W S It could be spatial attention weights; W C It could be channel attention weights.

[0103] S205. The enhanced features are fused to obtain fused features.

[0104] S206. Based on the fusion features, perform service security detection on the power service, and when the detection result indicates that there is a security risk, report the risk information.

[0105] The technical solution of this embodiment can extract multi-scale features from the filtered image, perform pooling and convolution processing on the multi-scale features to obtain the attention weights, and finally process the multi-scale features with the attention weights to obtain the enhanced features related to power behavior in the filtered image. The relevant attention weights can be determined in real time by the content of the currently input filtered image, realizing the adaptive generation of attention weights, thereby improving the detection accuracy and robustness of power service safety management in complex and dynamically changing power scenarios.

[0106] Based on the above embodiments, the present invention also provides an optional embodiment, which can be further optimized based on the above embodiments, specifically:

[0107] After performing pooling and convolution processing on the multi-scale features to obtain the attention weights, the scene feature vectors related to the power scene in the filtered image are obtained; the attention weights are then adjusted based on the scene feature vectors.

[0108] It should be noted that dynamic changes in the power scenario (such as changes in lighting conditions, weather conditions, and equipment operating status) can cause feature distribution shifts, leading to feature extraction failure. To address dynamic interference in the power scenario, this optional embodiment can further adapt and adjust the attention weights using the scene feature vector, enabling the feature focusing strategy to be adaptively adjusted according to changes in the power scenario.

[0109] For example, in this optional embodiment, a scene feature vector S related to the current power scene can be extracted from the filtered image first, where S∈R 1×K (K can be the scene dimension). The scene feature vector S can typically be extracted from a global scene encoding branch. This branch can perform operations such as global pooling on the features of the backbone network to obtain a vector that represents the global style and environment of the image. For example, S can encode scene information such as "indoor during the day," "outdoor at dusk," or "strong light at night."

[0110] Then, the scene feature vector can be mapped to the adjustment coefficient ΔW of the attention weights using a fully connected layer. It should be noted that, taking the attention weights as an example, which include spatial attention weights and channel attention weights, the adjustment coefficient ΔW of the attention weights can include the channel attention adjustment coefficient ΔW. C Spatial attention adjustment coefficient ΔW S .

[0111] Finally, the attention weights can be adjusted using the adjustment coefficient ΔW. For example, based on the channel attention adjustment coefficient ΔW... C Adjusting the channel attention weights can be done as follows:

[0112] ;

[0113] in, This can be used to adjust the channel attention weights.

[0114] Correspondingly, the adjusted channel attention weights can also be obtained using the same method. .

[0115] It should be noted that after determining the adjusted attention weights, the enhancement features that can be adapted to the changed power scenario can be re-determined based on the adjusted attention weights:

[0116] ;

[0117] in, This can be an enhanced feature after adjustment.

[0118] Example 3

[0119] Figure 3 This is a schematic diagram of the structure of a power service safety management device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device may include:

[0120] Processing module 31 can be used to process the received raw image of the power service to obtain the corresponding filtered image;

[0121] The determining module 32 can be used to determine the attention weights of the filtered image and, based on the attention weights, determine the enhancement features related to power behavior in the filtered image;

[0122] The fusion module 33 can be used to fuse the enhanced features to obtain fused features;

[0123] The detection module 34 can be used to perform service security detection on the power service based on the fusion features, and report risk information when the obtained detection result indicates that there is a security risk.

[0124] The technical solution of this embodiment can process the original image of the received power service to obtain a corresponding filtered image, then determine the attention weight of the filtered image, and based on the attention weight, determine the enhanced features related to power behavior in the filtered image. Finally, the enhanced features are fused to obtain fused features, and the power service is used for service security detection based on the fused features. When the obtained detection result indicates that there is a security risk, the risk information is reported. By focusing only on the enhanced features related to power behavior, the solution can solve the problems of low detection accuracy and high security management cost of existing solutions for power services, thereby improving the detection accuracy of power services and reducing the computational cost.

[0125] Optionally, the processing module 31 can be used to perform standardization processing on the original image to obtain a standardized image;

[0126] The normalized image is denoised using an adaptive Gaussian filter to obtain the filtered image.

[0127] Optionally, the determining module 32 may include: an extraction unit, a weight determining unit, and a feature determining unit.

[0128] The extraction unit can be used to extract multi-scale features of the filtered image;

[0129] The weight determination unit can be used to perform pooling and convolution processing on the multi-scale features to obtain the attention weights;

[0130] The feature determination unit can be used to process the multi-scale features through the attention weights to obtain enhanced features related to the power service in the filtered image.

[0131] Optionally, the attention weights may include channel attention weights and spatial attention weights.

[0132] Accordingly, the feature determination unit can be specifically used to process the multi-scale features through the channel attention weights to obtain the key features of the filtered image related to power behavior in the channel dimension;

[0133] By processing the key features using the spatial attention weights, enhanced features related to power behavior in the channel and spatial dimensions of the filtered image are obtained.

[0134] Optionally, the determining module 32 may further include: a scene adaptation unit;

[0135] The scene adaptation unit can be used to obtain the scene feature vector of the filtered image after performing pooling and convolution processing on the multi-scale features to obtain the attention weights.

[0136] The attention weights are adjusted based on the scene feature vector.

[0137] Optionally, the fusion module 33 can be specifically used to determine the resolution of the enhanced feature;

[0138] Based on the resolution of the enhanced features, the enhanced features are fused across scales to obtain the fused features.

[0139] The power service safety management device provided in the embodiments of the present invention can execute the power service safety management method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0140] Example 4

[0141] Figure 4 A schematic diagram of an electronic device 40 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0142] like Figure 4 As shown, the electronic device 40 includes at least one processor 41 and a memory, such as a read-only memory (ROM) 42 or a random access memory (RAM) 43, communicatively connected to the at least one processor 41. The memory stores computer programs executable by the at least one processor. The processor 41 can perform various appropriate actions and processes based on the computer program stored in the ROM 42 or loaded from storage unit 48 into the RAM 43. The RAM 43 may also store various programs and data required for the operation of the electronic device 40. The processor 41, ROM 42, and RAM 43 are interconnected via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0143] Multiple components in electronic device 40 are connected to I / O interface 45, including: input unit 46, such as keyboard, mouse, etc.; output unit 47, such as various types of monitors, speakers, etc.; storage unit 48, such as disk, optical disk, etc.; and communication unit 49, such as network card, modem, wireless transceiver, etc. Communication unit 49 allows electronic device 40 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0144] Processor 41 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 41 performs the various methods and processes described above, such as power service security management methods.

[0145] In some embodiments, the power service security management method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 40 via ROM 42 and / or communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the power service security management method described above may be performed. Alternatively, in other embodiments, processor 41 may be configured to perform the power service security management method by any other suitable means (e.g., by means of firmware).

[0146] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0147] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0148] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0149] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0150] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0151] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0152] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0153] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A power service security management method characterized by, The method comprises the following steps: processing a received original image of a power service to obtain a corresponding filtered image; determining an attention weight of the filtered image, and determining an enhanced feature related to a power behavior in the filtered image based on the attention weight; performing feature fusion on the enhanced feature to obtain a fused feature; performing service security detection on the power service based on the fused feature, and reporting risk information when the obtained detection result is that there is a security risk.

2. The method of claim 1, wherein, The processing of the received original image of the power service to obtain the corresponding filtered image comprises the following steps: performing standardization processing on the original image to obtain a standardized image; performing denoising on the standardized image using adaptive Gaussian filtering to obtain the filtered image.

3. The method of claim 1, wherein, The determination of the attention weight of the filtered image and the determination of the enhanced feature of the filtered image based on the attention weight comprise the following steps: extracting a multi-scale feature of the filtered image; performing pooling and convolution processing on the multi-scale feature to obtain the attention weight; processing the multi-scale feature through the attention weight to obtain the enhanced feature related to the power behavior in the filtered image.

4. The method of claim 3, wherein, The attention weight comprises a channel attention weight and a spatial attention weight. The processing of the multi-scale feature through the channel attention weight to obtain the enhanced feature related to the power behavior in the filtered image comprises the following steps: processing the multi-scale feature through the channel attention weight to obtain a key feature related to the power behavior in the filtered image in the channel dimension; processing the key feature through the spatial attention weight to obtain an enhanced feature related to the power behavior in the filtered image in the channel dimension and the spatial dimension.

5. The method of claim 3, wherein, After the pooling and convolution processing on the multi-scale feature to obtain the attention weight, the method further comprises the following steps: obtaining a scene feature vector related to a power scene in the filtered image; adjusting the attention weight according to the scene feature vector.

6. The method of claim 1, wherein, The feature fusion of the enhanced feature to obtain the fused feature comprises the following steps: determining a resolution of the enhanced feature; performing cross-scale feature fusion on the enhanced feature based on the resolution of the enhanced feature to obtain the fused feature.

7. A power service security management apparatus characterized by comprising: The method comprises the following steps: a processing module configured to process a received original image of a power service to obtain a corresponding filtered image; a determination module configured to determine an attention weight of the filtered image, and determine an enhanced feature related to a power behavior in the filtered image based on the attention weight; a fusion module configured to perform feature fusion on the enhanced feature to obtain a fused feature; a detection module configured to perform service security detection on the power service based on the fused feature, and report risk information when the obtained detection result is that there is a security risk.

8. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the power service security management method in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to implement the power service security management method in any one of claims 1-6 when executed.

10. A computer program product, characterised in that, The computer program product comprises a computer program which, when executed by the processor, implements the power service security management method in any one of claims 1-6.

Images