Method and device for determining insulation state of cable and electronic equipment

Abstract

The invention discloses a method and device for determining the insulation state of a cable and electronic equipment. The method comprises the following steps: acquiring a cable image corresponding to a target cable; determining a regional image according to the cable image; segmenting the regional image to obtain a plurality of first sub-images; inputting the plurality of first sub-images into splicing modules corresponding to a plurality of target layers of the target model to obtain a plurality of second sub-images; inputting the plurality of second sub-images into feature recognition modules corresponding to the plurality of target layers respectively to obtain image feature data corresponding to the plurality of second sub-images; and determining a target insulation state corresponding to the target cable according to the multiple pieces of image feature data. According to the method and the device, the technical problems of difficulty in accurately determining the characteristics of the cable and low accuracy of the determined insulation state of the cable due to large data volume of the to-be-detected cable in related technologies are solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and device for determining the insulation state of a cable, and electronic equipment. Background Art

[0002] Determining the insulation status of cables can prevent the risks of leakage, short circuits, and fire, and is crucial for ensuring the safe and stable operation of power systems. Currently, high-voltage cable status detection methods that integrate multi-source heterogeneous data are primarily used to detect cable insulation status. However, when the amount of cable data corresponding to the cable to be tested is very large, these methods struggle to determine cable characteristics from the data, resulting in low accuracy in the determined cable insulation status.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a method, device, and electronic device for determining the insulation status of a cable, so as to at least solve the technical problem in related technologies that, due to the large amount of data on the cable to be detected, it is difficult to accurately determine the characteristics of the cable, resulting in low accuracy of the determined cable insulation status.

[0005] According to one aspect of an embodiment of the present invention, a method for determining the insulation status of a cable is provided, comprising: acquiring a cable image corresponding to a target cable; determining a regional image based on the cable image, wherein the feature intensity of the insulation status feature corresponding to the regional image is greater than a predetermined intensity threshold; segmenting the regional image to obtain a plurality of first sub-images; inputting the plurality of first sub-images into stitching modules corresponding to a plurality of target layers of a target model respectively to obtain a plurality of second sub-images, wherein the corresponding stitching modules are provided with corresponding predetermined stitching sizes, and the target model is obtained through training with sample data; inputting the plurality of second sub-images into feature recognition modules corresponding to the plurality of target layers respectively to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature recognition modules are provided with corresponding feature recognition sizes, and the corresponding feature recognition modules recognize the image feature data corresponding to the corresponding sub-images based on the corresponding feature recognition sizes; and determining the target insulation status corresponding to the target cable based on the plurality of image feature data.

[0006] Optionally, the target insulation state corresponding to the target cable is determined based on multiple image feature data, including: obtaining cable performance parameters corresponding to the target cable; determining timing feature data corresponding to the target cable based on the cable performance parameters; determining fusion features corresponding to the target cable based on the timing feature data and the multiple image feature data; and determining the insulation state corresponding to the target cable based on the fusion features.

[0007] Optionally, determining a regional image based on the cable image includes: calling an image processing model, wherein the image processing model is provided with target parameters, and the target parameters are determined based on initial parameters corresponding to an initial image model and a plurality of sample images; inputting the cable image into the image processing model to obtain target position data corresponding to a target box, wherein the target box is used to determine a region in the cable image where the characteristic intensity of the insulation state feature is greater than a predetermined intensity threshold; and determining the regional image based on the target position data.

[0008] Optionally, before calling the image processing model, it also includes: determining an error function with the goal of minimizing the error value, wherein the error function is used to determine the error value between the predicted position data and the sample position data, the predicted position data is the position data of the target box determined based on the initial image model and the sample image, and the sample position data is the actual position data of the target box corresponding to the sample image; based on the error function, updating the initial parameters corresponding to the initial image model to obtain updated parameters; when the error value corresponding to the updated parameters is less than the error threshold, determining the updated parameters as the target parameters; and determining the image processing model based on the target parameters and the initial image model.

[0009] Optionally, before calling the image processing model, it also includes: acquiring multiple original images, wherein the first number corresponding to the multiple original images is less than a predetermined number threshold, and the cables corresponding to the multiple original images respectively contain corresponding defect areas; based on the multiple original images, determining the defect images corresponding to the multiple defect areas and the defect labels corresponding to the multiple defect areas respectively; based on the multiple defect images and the multiple defect labels, determining the multiple sample images, wherein the second number corresponding to the multiple sample images is greater than the predetermined number threshold.

[0010] Optionally, multiple sample images are determined based on multiple defect images and multiple defect labels, including: determining multiple initial images based on the multiple defect images and the multiple defect labels; determining evaluation values ​​corresponding to the multiple initial images respectively; and determining the corresponding initial image as a sample image when the corresponding evaluation value is greater than a scoring threshold.

[0011] Optionally, determining the insulation status corresponding to the target cable based on multiple image feature data includes: determining status scores corresponding to multiple insulation classification states based on the multiple image feature data; and determining the target insulation status from the multiple insulation classification states based on the multiple status scores.

[0012] According to one aspect of an embodiment of the present invention, a device for determining the insulation status of a cable is provided, comprising: an acquisition module for acquiring a cable image corresponding to a target cable; a first determination module for determining a regional image based on the cable image, wherein the feature intensity of the insulation status feature corresponding to the regional image is greater than a predetermined intensity threshold; a segmentation module for segmenting the regional image to obtain a plurality of first sub-images; a splicing module for inputting the plurality of first sub-images into splicing modules corresponding to a plurality of target layers of a target model respectively to obtain a plurality of second sub-images, wherein the corresponding splicing modules are provided with corresponding predetermined splicing sizes, and the target model is obtained by training with sample data; an identification module for inputting the plurality of second sub-images into feature identification modules corresponding to the plurality of target layers respectively to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature identification modules are provided with corresponding feature identification sizes, and the corresponding feature identification sizes are used to identify the image feature data corresponding to the corresponding sub-images in units of corresponding pixel sizes; and a second determination module for determining the target insulation status corresponding to the target cable based on the plurality of image feature data.

[0013] According to one aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement any of the above methods for determining the insulation status of a cable.

[0014] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above methods for determining the insulation status of a cable.

[0015] In an embodiment of the present invention, a cable image corresponding to a target cable is obtained; a regional image is determined based on the cable image, wherein the characteristic intensity of the insulation state characteristic corresponding to the regional image is greater than a predetermined intensity threshold; the regional image is segmented to obtain a plurality of first sub-images; the plurality of first sub-images are input into stitching modules corresponding to a plurality of target layers of a target model, respectively, to obtain a plurality of second sub-images, wherein the corresponding stitching modules are provided with corresponding predetermined stitching sizes, and the target model is obtained through training with sample data; the plurality of second sub-images are input into feature recognition modules corresponding to the plurality of target layers, respectively, to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature recognition modules are provided with corresponding feature recognition sizes, and the corresponding features The recognition module identifies the image feature data corresponding to the corresponding sub-image based on the corresponding feature recognition size; based on multiple image feature data, the target insulation state corresponding to the target cable is determined by inputting multiple second sub-images into the feature recognition modules corresponding to the multiple target layers respectively, thereby obtaining image feature data corresponding to the multiple second sub-images, thereby achieving the purpose of determining the target insulation state corresponding to the target cable based on multiple image feature data, thereby realizing the technical effect of comprehensively considering the local and global image feature data of the target cable to improve the accuracy of the determined insulation state of the target cable, and thus solving the technical problem in the related art that due to the large amount of data of the cable to be detected, it is difficult to accurately determine the characteristics of the cable, and the accuracy of the determined insulation state of the cable is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 is a flow chart of a method for determining the insulation status of a cable according to an embodiment of the present invention;

[0018] Figure 2 This is a system block diagram of a high-voltage cable insulation status online detection system provided by an optional embodiment of the present invention;

[0019] Figure 3 is a flowchart of parallel processing of an image preprocessing unit provided by an optional embodiment of the present invention;

[0020] Figure 4 is a flow chart of serial processing of an image preprocessing unit provided by an optional embodiment of the present invention;

[0021] Figure 5 4 is a structural block diagram of a device for determining the insulation status of a cable according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] Example 1

[0025] According to an embodiment of the present invention, an embodiment of a method for determining the insulation status of a cable is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] Figure 1 Flowchart of a method for determining the insulation status of a cable according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0027] Step S102: Acquire a cable image corresponding to the target cable.

[0028] In step S102 provided in the present application, a cable image corresponding to the target cable is acquired.

[0029] Among them, the target cable is involved, and the target cable refers to the cable whose insulation status needs to be determined.

[0030] Among them, the cable image is involved, and the cable image refers to the image corresponding to the target cable.

[0031] In this step, it is necessary to collect a cable image of the target cable. The cable image contains detailed information about the target cable, such as the integrity of the insulation layer, the presence of cracks, partial discharge marks, corrosion conditions, etc. It is an important data source for detecting the insulation status of the target cable.

[0032] Step S104: determining a region image based on the cable image, wherein the feature intensity of the insulation state feature corresponding to the region image is greater than a predetermined intensity threshold.

[0033] In step S104 provided in the present application, a regional image is determined.

[0034] Here, a regional image is involved, and the regional image refers to an image in which the characteristic intensity of the insulation state feature in the cable image is greater than a predetermined intensity threshold.

[0035] Among them, insulation status features are involved. Insulation status features refer to specific visual features or attributes in cable images that can be used to evaluate the insulation status of the target cable, such as the halo of partial discharge, the edge contrast of cracks, and the texture changes of corrosion areas.

[0036] Among them, characteristic strength is involved, which refers to the degree to which the characteristic value of a characteristic item of a specific insulation state is higher than its average characteristic value.

[0037] Here, a predetermined intensity threshold is involved, which refers to a pre-set numerical standard used to screen out area images with significant insulation status characteristics in the cable image.

[0038] In this step, the cable image is first preprocessed, such as resizing, denoising, and contrast enhancement, to ensure feature clarity and recognizability. Next, insulation features, such as cracks and partial discharge, are detected. The intensity of these insulation features is calculated, and a predetermined intensity threshold is set to filter out regions with high intensity. Finally, regions within the cable image where the intensity exceeds the predetermined threshold are identified to form a regional image.

[0039] Through this step, the regional image is determined, and the areas with low characteristic intensity of the insulation state characteristics are effectively filtered out, reducing the subsequent data processing volume, and the regional images with high characteristic intensity are screened out, avoiding blind analysis of the entire image range, and improving the pertinence and efficiency of detection.

[0040] Step S106 , segmenting the regional image to obtain a plurality of first sub-images.

[0041] In step S106 provided in the present application, a plurality of first sub-images are obtained.

[0042] The first sub-image is referred to as an image obtained after segmenting the region image. The plurality of first sub-images respectively represent a portion of the region in the region image.

[0043] In this step, the image of the already located critical cable area (i.e., the regional image) is further subdivided into multiple smaller sub-images related to specific insulation characteristics using image segmentation technology, resulting in multiple first sub-images. This step allows the image characteristics of different local areas to be determined later, improving the accuracy of defect detection and more accurately identifying and analyzing different defect types in the cable insulation layer, such as partial discharge, cracks, and corrosion.

[0044] In step S108, the multiple first sub-images are input into the stitching modules corresponding to the multiple target layers of the target model to obtain multiple second sub-images, wherein the corresponding stitching modules are provided with corresponding predetermined stitching sizes, and the target model is obtained through training with sample data.

[0045] In step S108 provided in the present application, a plurality of second sub-images are obtained.

[0046] Among them, a target model is involved. The target model refers to a model obtained by training based on sample data and used to determine image feature data corresponding to different feature recognition sizes based on multiple first sub-images.

[0047] Among them, the target layer is involved. The target layer refers to the model network layer in the target model, including the splicing module and the feature recognition module.

[0048] Herein, a splicing module is involved, which refers to a module for splicing multiple input first sub-images according to corresponding predetermined splicing sizes to restore the contextual relationship between the multiple first sub-images.

[0049] This involves the second sub-image, which is the image obtained after being processed by the target model splicing module. The second sub-image contains deeper physical or geometric features, which can help more accurately identify or evaluate the insulation condition of the target cable.

[0050] Here, a predetermined splicing size is involved, and the predetermined splicing size refers to the size of a second sub-image spliced ​​from a plurality of first sub-images, which is predetermined by the splicing module.

[0051] In this step, after the regional image is cut into multiple first sub-images, each of these first sub-images is input into multiple target layers of the target model. The target layer stitching module then processes the sub-images to generate multiple second sub-images of varying predetermined stitching sizes. By inputting the sub-images into different layers of the model, the stitching module reassembles the processed feature maps according to the preset stitching sizes, restoring the spatial relationships between local regions and simultaneously capturing multiple levels of detail in the cable image, enhancing the model's ability to identify defects.

[0052] In step S110, the plurality of second sub-images are respectively input into the feature recognition modules corresponding to the plurality of target layers to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature recognition module is provided with a corresponding feature recognition size, and the corresponding feature recognition module recognizes the image feature data corresponding to the corresponding sub-image according to the corresponding feature recognition size.

[0053] In step S110 provided in the present application, image feature data corresponding to the plurality of second sub-images is obtained.

[0054] Among them, a feature recognition module is involved, and the feature recognition module refers to a module in the target layer of the target model for identifying image features of corresponding feature recognition sizes.

[0055] This involves image feature data, which refers to quantitative or qualitative information extracted from the second sub-image that can characterize the cable insulation condition. For example, feature vectors of contrast, brightness, color, texture, edge strength, morphological features, partial discharge, cracks, corrosion, etc.

[0056] In this step, multiple second sub-images are fed into their corresponding feature recognition modules. Each feature recognition module is assigned a specific feature recognition size to ensure it can effectively capture the size of the target feature. Based on the assigned feature recognition size, the feature recognition module identifies and extracts image feature data associated with the second sub-image.

[0057] Through this step, the most suitable feature recognition module is selected for each sub-image, which can simultaneously capture multiple scale features in the cable image. Based on the content and feature type of the sub-image, more refined and targeted recognition can be performed, from tiny surface cracks to macroscopic structural damage, thereby improving the comprehensiveness and accuracy of defect detection.

[0058] Step S112 , determining a target insulation state corresponding to the target cable according to the plurality of image feature data.

[0059] In step S112 provided in the present application, a target insulation state corresponding to the target cable is determined.

[0060] Among them, the target insulation state is involved, and the target insulation state refers to the actual condition of the insulation layer of the target cable.

[0061] Through the above steps S102-S112, a cable image corresponding to the target cable can be obtained; based on the cable image, a regional image is determined, wherein the characteristic intensity of the insulation state feature corresponding to the regional image is greater than a predetermined intensity threshold; the regional image is segmented to obtain a plurality of first sub-images; the plurality of first sub-images are input into the splicing modules corresponding to the plurality of target layers of the target model respectively to obtain a plurality of second sub-images, wherein the corresponding splicing modules are provided with corresponding predetermined splicing sizes, and the target model is obtained through training with sample data; the plurality of second sub-images are input into the feature recognition modules corresponding to the plurality of target layers respectively to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature recognition modules are provided with corresponding feature recognition sizes, and the corresponding The feature recognition module identifies the image feature data corresponding to the corresponding sub-image based on the corresponding feature recognition size; the target insulation state corresponding to the target cable is determined based on multiple image feature data. By inputting multiple second sub-images into the feature recognition modules corresponding to the multiple target layers respectively, the image feature data corresponding to the multiple second sub-images are obtained, thereby achieving the purpose of determining the target insulation state corresponding to the target cable based on multiple image feature data, thereby realizing the technical effect of comprehensively considering the local and global image feature data of the target cable to improve the accuracy of the determined insulation state of the target cable, and thus solving the technical problem in the related art that due to the large amount of data of the cable to be detected, it is difficult to accurately determine the characteristics of the cable, and the accuracy of the determined insulation state of the cable is low.

[0062] As an optional embodiment, the target insulation state corresponding to the target cable is determined based on multiple image feature data, including: obtaining cable performance parameters corresponding to the target cable; determining the timing feature data corresponding to the target cable based on the cable performance parameters; determining the fusion features corresponding to the target cable based on the timing feature data and multiple image feature data; and determining the insulation state corresponding to the target cable based on the fusion features.

[0063] In this embodiment, specific steps of determining a target insulation state corresponding to a target cable based on a plurality of image feature data are described.

[0064] Among them, cable performance parameters are involved. Cable performance parameters refer to parameters that describe the operating status and physical characteristics of the target cable, such as the temperature, current, voltage, tension, humidity, etc. of the target cable.

[0065] This involves time-series characteristic data, which refers to the historical record of the target cable's performance parameters changing over time. This data can reveal the dynamic trends of the target cable's performance parameters, providing a time-dimensional analytical basis for evaluating the target cable's insulation condition.

[0066] Among them, fusion features are involved. Fusion features refer to the feature representation obtained by comprehensive processing of image feature data and time series feature data. Fusion features can simultaneously reflect the visual condition and operating status of the cable, providing more comprehensive status assessment information.

[0067] In this step, after determining image feature data relevant to the target cable insulation status assessment based on multiple second sub-images, the target cable's real-time or historical operating data, such as temperature, current, and voltage, is first collected to determine cable performance parameters. These parameters describe the target cable's operating environment and physical state. Subsequently, a time series analysis is performed on the cable performance parameters to identify patterns and trends in their temporal variations, generating time series feature data. This step facilitates understanding the target cable's state changes from a temporal perspective. Next, the image feature data is fused with the time series feature data to generate fused features. This is typically achieved by designing specific models or algorithms, such as a Transformer model with a self-attention mechanism, or by employing methods such as feature splicing and weighted averaging, to ensure that the fused features simultaneously reflect both the cable's visual and operational status information. Finally, based on the fused features, a comprehensive assessment of the target cable's insulation status is performed, and a judgment result for the target cable's insulation status is output, resulting in the target insulation status.

[0068] Through this step, the multiple image feature data determined from the regional image are combined with the time series feature data determined by the cable performance parameters to obtain a fusion feature. This not only takes into account visual defects, but also considers the impact of the operating environment and physical conditions on insulation performance, focusing on the immediate status and long-term change trends of the cable, and can provide a more comprehensive and accurate cable insulation status assessment.

[0069] As an optional embodiment, determining a regional image based on a cable image includes: calling an image processing model, wherein the image processing model is provided with target parameters, and the target parameters are determined based on initial parameters corresponding to an initial image model and a plurality of sample images; inputting the cable image into the image processing model to obtain target position data corresponding to a target box, wherein the target box is used to determine an area in the cable image where the characteristic intensity of the insulation state feature is greater than a predetermined intensity threshold; and determining a regional image based on the target position data.

[0070] In this embodiment, the specific steps of determining the area image based on the cable image are described.

[0071] Among them, an image processing model is involved. The image processing model refers to a machine learning or deep learning model used to determine the area in the input cable image where the feature intensity is greater than a predetermined intensity threshold, such as a convolutional neural network (CNN), a Transformer model, etc.

[0072] Among them, target parameters are involved. Target parameters refer to model parameters in the image processing model. The target parameters are obtained through initial parameter training and optimization of the initial image model, so that the image processing model can more accurately identify specific areas in the cable image.

[0073] Among them, the initial image model is involved. The initial image model refers to the initial state of the image processing model before training begins. It needs to be further optimized to adapt to the specific needs of target cable insulation status detection.

[0074] Among them, the initial parameters are involved, and the initial parameters refer to the model parameters of the initial image model.

[0075] Among them, sample images are involved. Sample images refer to the image set used to train the initial image model. The sample images include the appearance of the cable under different conditions, including images of various defective states of the cable.

[0076] Among them, a target frame is involved, and the target frame refers to a position frame used to frame an area in a regional image whose feature intensity is greater than a predetermined intensity threshold.

[0077] Among them, target position data is involved. The target position data refers to the specific position information of the target box in the cable image, usually expressed in the form of coordinates, which is used to guide subsequent processing of cropping or focusing on feature abnormal areas from the cable image.

[0078] In this step, a pre-trained image processing model with target parameters is invoked. The cable image is input into the model, which identifies regions with abnormal feature intensities, generates a target frame, and outputs the target frame's location information, known as the target location data. Based on this target location data, the region with abnormal feature intensities is cropped or extracted from the cable image to form a regional image. These images contain high-intensity insulation characteristics, which serve as a crucial basis for subsequent analysis and judgment.

[0079] As an optional embodiment, before calling the image processing model, it also includes: determining an error function with the goal of minimizing the error value, wherein the error function is used to determine the error value between the predicted position data and the sample position data, the predicted position data is the position data of the target box determined based on the initial image model and the sample image, and the sample position data is the actual position data of the target box corresponding to the sample image; based on the error function, updating the initial parameters corresponding to the initial image model to obtain updated parameters; when the error value corresponding to the updated parameters is less than the error threshold, determining the updated parameters as the target parameters; and determining the image processing model based on the target parameters and the initial image model.

[0080] In this embodiment, specific steps of determining an image processing model are described.

[0081] Among them, the error function is involved. The error function refers to a mathematical function used to quantify the error between the target box position data predicted by the initial image model and the actual position data.

[0082] Among them, the predicted position data is involved, and the predicted position data refers to the target frame position data of the sample image determined by the initial image model.

[0083] Among them, sample position data is involved, and the sample position data refers to the actual target frame position data corresponding to the sample image.

[0084] Among them, the update parameters are involved. The update parameters refer to the model parameters obtained by adjusting the initial parameters according to the error function.

[0085] This involves an error threshold, which is a pre-set value used to determine whether the target box position data determined by the updated image model meets the expected level of accuracy. When the error value is lower than the error threshold, the updated parameters are considered to meet the requirements and are determined as the target parameters.

[0086] In this step, an error function is first determined with the goal of minimizing the error value, measuring the difference between the model-predicted position data and the actual position data in the sample image. Next, the error function is used to update the initial parameters of the initial image model to obtain updated parameters that make the predicted position as close as possible to the actual position, that is, minimize the error value. When the error value corresponding to the updated parameters is lower than the preset error threshold, this set of updated parameters is determined as the target parameters, and based on the updated parameters and the initial image model, an image processing model is determined for subsequent image analysis and processing tasks. Through this step, the error function guides the update of the initial parameters of the initial image model, minimizes the error between the predicted position data and the sample position data, and improves the accuracy and reliability of the model in locating cable insulation features.

[0087] As an optional embodiment, before calling the image processing model, it also includes: obtaining multiple original images, wherein the first number corresponding to the multiple original images is less than a predetermined number threshold, and the cables corresponding to the multiple original images respectively contain corresponding defect areas; based on the multiple original images, determining the defect images corresponding to the multiple defect areas and the defect labels corresponding to the multiple defect areas respectively; based on the multiple defect images and the multiple defect labels, determining multiple sample images, wherein the second number corresponding to the multiple sample images is greater than the predetermined number threshold.

[0088] In this embodiment, specific steps of determining a plurality of sample images are described.

[0089] Here, the original image is involved, and the original image refers to the collected cable image containing the cable defect area.

[0090] Here, a first quantity is involved, and the first quantity refers to the number of images of the original image.

[0091] Here, a predetermined number threshold is involved, and the predetermined number threshold refers to a value used to determine whether the number of determined sample images is sufficient for subsequent image processing model determination.

[0092] Among them, defective areas are involved. Defective areas refer to cable areas with abnormal conditions, such as cable areas with partial discharge, cracks, corrosion and other abnormal conditions.

[0093] Among them, defect labels are involved, and defect labels refer to label information of defect types used to mark defect areas in images.

[0094] Here, the second number is involved, and the second number refers to the number of determined sample images.

[0095] In this step, first, a series of original images containing cable defects are collected. From these original images, specific defect areas, such as the spot of partial discharge or the crack on the cable surface, are identified and located. The defect areas are cut out from the original images to form defect images that focus on the defect details. A defect label, such as "partial discharge" or "crack", is assigned to each defect image to clarify the type of defect in the image. Next, the defect images are processed to generate a larger number of sample images that simulate different conditions. Through this process, the number of sample images (the second number) increases significantly, exceeding the predetermined threshold, which improves the diversity and quantity of the model's training data, helps the model learn more representations of defect features, and enhances its generalization ability.

[0096] As an optional embodiment, multiple sample images are determined based on multiple defect images and multiple defect labels, including: determining multiple initial images based on the multiple defect images and multiple defect labels; determining evaluation values ​​corresponding to the multiple initial images respectively; and determining the corresponding initial image as a sample image when the corresponding evaluation value is greater than a scoring threshold.

[0097] In this embodiment, specific steps of determining a plurality of sample images based on a plurality of defect images and a plurality of defect labels are described.

[0098] Herein, an initial image is involved, and the initial image refers to a cable image containing a defective area generated based on a plurality of defect images and defect labels corresponding to the plurality of defect images.

[0099] Here, an evaluation value is involved, and the evaluation value refers to a quantitative indicator used to evaluate the image quality of the determined initial image.

[0100] This involves a scoring threshold, a pre-set value used to screen defective images that meet quality requirements for use as sample images. Only when the initial image's evaluation value exceeds the scoring threshold will it be included in the sample image set for subsequent image processing model training.

[0101] In this step, multiple initial images are first generated based on the defect labels corresponding to the multiple defect images. Each initial image contains the visual representation of the defect, its location, and its type. A quality assessment is then performed on each initial image, and a corresponding evaluation value is determined. A scoring threshold is set, and high-quality defect images are selected as sample images based on the evaluation value. This step ensures the quality of the image processing model training dataset, ensuring that the sample images only contain images that provide clear and useful information, thereby avoiding the negative impact of low-quality images on image processing model training.

[0102] As an optional embodiment, the insulation status corresponding to the target cable is determined based on multiple image feature data, including: determining the status scores corresponding to multiple insulation classification states based on the multiple image feature data; and determining the target insulation status from the multiple insulation classification states based on the multiple status scores.

[0103] In this embodiment, specific steps of determining the insulation status corresponding to the target cable based on a plurality of image feature data are described.

[0104] Among them, the insulation classification status is involved. The insulation classification status refers to the different status categories of the cable insulation layer, such as normal, slight damage, severe damage, partial discharge, water intrusion, etc., which are used to describe and classify the quality status of the cable insulation layer.

[0105] Among them, the status score is involved, which refers to the quantitative evaluation score of the target cable in each insulation classification state.

[0106] In this step, first, multiple image feature data are extracted from the collected cable image. Then, based on all the extracted image feature data, a status score is calculated for each of the multiple insulation classification states. Finally, based on the status score of each insulation classification state, the insulation classification state with the highest score is determined as the final insulation status evaluation of the target cable.

[0107] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0108] High-voltage cables play a vital role in power grids, and their insulation condition is directly related to the safe operation of the grid. The continuous development of sensor technology, electronics technology, signal processing, and network technology has provided a solid technical foundation for online monitoring of high-voltage cable insulation. The application of these technologies makes real-time monitoring of cable insulation status possible, enabling rapid data processing and accurate analysis.

[0109] In related technologies, a high-voltage cable status detection method based on multi-source heterogeneous data fusion is usually used to detect the insulation status of the cable. By collecting the original monitoring images, original partial discharge images, original cable tension data, original temperature data, and original wind speed data of the cable insulator at multiple moments, the pre-trained fast regional convolutional neural network (FasterRCNN network), Alex network classifier (AlexNet classification network) and fully connected classification network are used to perform deep feature extraction and deep feature fusion on the collected data respectively, and the multi-source heterogeneous data fusion high-voltage cable status detection model is trained based on the deep features of the multi-source data fused at the same moment. The high-voltage cable status detection model based on the multi-source heterogeneous data fusion after training is used to make predictions, and the predicted status detection results of the high-voltage cable to be identified are obtained. However, during the use of the above method, when the amount of data is large, the pre-processing work such as cropping and resizing of the monitoring images and partial discharge images may be very time-consuming and complicated, which reduces the detection accuracy.

[0110] In view of this, an optional embodiment of the present invention provides an online detection method for the insulation status of a high-voltage cable, which can solve the shortcomings of the prior art that, when the amount of data is large, the pre-processing work such as cropping and resizing of monitoring images and partial discharge images may be very time-consuming and complicated, thereby reducing the detection accuracy.

[0111] The specific steps of the method for online detection of the insulation status of a high-voltage cable provided in an optional embodiment of the present invention are described in detail below.

[0112] S1. Acquire a cable image corresponding to a target cable.

[0113] Relevant data of high-voltage cables are collected in real time from sensors (such as image sensors, tension sensors, temperature sensors, wind speed sensors, etc.). These data include the appearance image of the cable (same as the above cable image), mechanical parameters (same as the above cable performance parameters), environmental parameters, etc.

[0114] The collected raw data is preliminarily processed, including image cropping, resizing, denoising, data standardization, etc. Machine learning algorithms are used to achieve intelligent image cropping and resizing to improve processing accuracy. At the same time, multi-source data is standardized to ensure data consistency and comparability.

[0115] S2. Determine the area image based on the cable image.

[0116] The image data is cropped to retain only the image of the cable insulator area, and the size is adjusted to ensure the image size is consistent for subsequent processing. A denoising algorithm is applied to remove noise and interference in the image.

[0117] Specifically, it can be divided into the following steps:

[0118] A1. Determine the initial image model.

[0119] Prepare sample image data and collect a large number of data sets containing monitoring images and partial discharge images; annotate the images, mark the areas that need to be cropped (same as the target box mentioned above) (usually the cable body or key parts) and provide resizing suggestions; perform preprocessing operations such as normalization and scaling on the images to adapt to the input requirements of the convolutional neural network model.

[0120] The convolutional neural network model is designed, which is specifically divided into: input layer: accepts preprocessed images as input; convolution layer: uses multiple convolution kernels (filters) to perform convolution operations on the input image to extract image features; activation layer: uses rectified linear units (ReLU) to increase nonlinearity; fully connected layer: expands the feature map into a one-dimensional vector and connects it to several fully connected layers for classification or regression tasks; output layer: outputs the coordinates of the cropped area (such as a bounding box) or size adjustment parameters (same as the above position data) according to the task requirements.

[0121] Among them, the convolution operation can be expressed as:

[0122]

[0123] Among them, Y(x,y) is the feature map after convolution (same as the above region image), X(x+i,y+j) is the input image (same as the above sample image), W is the convolution kernel, and k is the size of the convolution kernel.

[0124] A2. Determine the error function with the goal of minimizing the error value.

[0125] Define a loss function (same as the error function above) to measure the difference between the model prediction (same as the predicted position data above) and the true annotation (same as the sample position data above). For the cropping task, use the bounding box regression loss (such as IoU loss, Smooth L1 loss, etc.); for the resizing task, use the pixel-level error or structural similarity (SSIM) indicator as the loss.

[0126] A3. Based on the error function, the initial parameters corresponding to the initial image model are updated to obtain the image processing model.

[0127] Use gradient descent to update model parameters to minimize the loss function; continuously adjust model parameters through multiple iterative training until the model's performance on the validation set no longer improves significantly; evaluate the model's performance on the test set, including the accuracy of cropping and the rationality of resizing; deploy the trained model and integrate it with the image preprocessing unit.

[0128] A4. Determine the regional image based on the image processing model.

[0129] For real-time surveillance images, the image processing model directly crops and resizes the input images and feeds the results back to the data enhancement and annotation module or the model training and optimization module in real time. For a large number of historical images or offline data, the images are sent to the image processing model in batches for processing to save time and improve efficiency.

[0130] It should be noted that before training the image processing model, it is necessary to determine the sample image. The specific steps are as follows:

[0131] B1. Obtain multiple original images.

[0132] Input the original data set (same as the original image above), and receive an original data set containing a limited number of high-voltage cable defect images.

[0133] B2. Determine defect images corresponding to the plurality of defect regions and defect labels corresponding to the plurality of defect regions.

[0134] The generated image is automatically segmented using the pre-trained Mask R-CNN model and a preliminary pseudo-label is generated. The output of Mask R-CNN is represented as (b, c, m), where b is the bounding box coordinate, c is the classification label, and m is the corresponding segmentation mask. Given an input image I, the model will output a series of (b i ,c i ,m i), where i represents the i-th detected object.

[0135] Consider an image of a high-voltage cable defect, including a partial discharge (PD) defect. The pre-trained Mask R-CNN model processes this image. The model outputs a bounding box with coordinates b = (x1, y1, x2, y2), a classification label c = "partial discharge," and a corresponding segmentation mask m. This segmentation mask is a binary image of the same size as the input image, with the defective area labeled as 1 and the rest of the area labeled as 0.

[0136] Manual correction: Human annotators review the pseudo-labels generated by the pre-trained model and make corrections as needed. Due to the high accuracy of the pre-trained model, the workload of manual correction is relatively small, and usually only needs to focus on areas where the model is uncertain or the segmentation is inaccurate.

[0137] Active learning strategy: The active learning algorithm evaluates the uncertainty of the model for the sample and prioritizes the recommendation of uncertain samples for manual labeling. The entropy of the classification probability is used to measure the uncertainty. The entropy (H(p)) is calculated as follows:

[0138] H(p)=-∑ i p i log p i

[0139] Among them, p i is the probability of being classified into class i.

[0140] For each sample, the entropy of its classification probability is calculated and sorted according to the size of the entropy. A high entropy indicates that the model is uncertain about the classification result of the image.

[0141] B3. Determine multiple sample images.

[0142] Generative data augmentation uses a style generative adversarial network (StyleGAN3) to generate cable defect images. StyleGAN3 generates synthetic images that are highly similar to real images (the same as the sample images mentioned above) by learning the potential distribution of the original dataset. The generation process of StyleGAN3 is expressed as, where is the generator network, is a vector randomly sampled from the latent space, and the generator will map it to the image space to generate a synthetic image.

[0143] The specific steps of B3 determining multiple sample images are as follows:

[0144] C1. Determine the evaluation values ​​corresponding to the multiple initial images.

[0145] C2. When the corresponding evaluation value is greater than the scoring threshold, the corresponding initial image is determined to be a sample image.

[0146] The generated image will undergo a quality check step, represented by an evaluation function Q(x), where x is the generated image. If the value of Q(x) (the same as the evaluation value mentioned above) is greater than a certain threshold (the same as the scoring threshold mentioned above), the image is considered qualified; otherwise, the image is preprocessed or filtered.

[0147] S3. Segment the regional image to obtain a plurality of first sub-images.

[0148] S4. Input the multiple first sub-images into the stitching modules corresponding to the multiple target layers of the target model to obtain the second sub-images corresponding to the multiple first sub-images.

[0149] S5. Input the plurality of second sub-images into feature recognition modules corresponding to the plurality of target layers respectively, to obtain image feature data corresponding to the plurality of second sub-images.

[0150] In the image branch, the image data of the high-voltage cable is input into the Swin Transformer model. The Swin Transformer model divides the image into multiple non-overlapping small windows, calculates self-attention in each window, gradually merges the windows through a hierarchical structure, expands the receptive field, extracts features from local to global, and obtains a feature map containing image features (the same as the above image feature data).

[0151] S6. Determine a target insulation state corresponding to the target cable based on the plurality of image feature data.

[0152] The parameter branch inputs the time series data of mechanical parameters and environmental parameters into the TFT model. The TFT model captures the dependencies in the time series data through the self-attention mechanism and uses the gating mechanism to control the flow of information, thereby modeling the time series data and obtaining a feature vector containing time series features (the same as the above-mentioned time series feature data).

[0153] Cross-modal interaction: input image features and time series features into the cross-attention mechanism, calculate the attention weight of image features to time series features, and use the attention weight to perform weighted summation on the time series features to obtain the fused features, and obtain the fused cross-modal features (same as the above fused features); insulation status evaluation and judgment: input the fused cross-modal features into the classifier, and the classifier calculates the score (same as the above status score) of each insulation status category (same as the above insulation status classification) according to the input features, and selects the category with the highest score as the final insulation status judgment result, and obtains the insulation status judgment result of the high-voltage cable (same as the above target insulation status).

[0154] An optional embodiment of the present invention further provides an online detection system for the insulation status of a high-voltage cable. Figure 2: is a system block diagram of a high-voltage cable insulation status online detection system provided by an optional embodiment of the present invention, such as Figure 2 As shown, the high-voltage cable insulation status online detection system includes:

[0155] Data acquisition module: responsible for collecting relevant data of high-voltage cables from sensors (such as image sensors, tension sensors, temperature sensors, wind speed sensors, etc.) in real time. These data include the appearance image, mechanical parameters, environmental parameters, etc. of the cables. The data acquisition module includes an image acquisition unit, a mechanical parameter acquisition unit, an environmental parameter acquisition unit and a data acquisition control unit. The image acquisition unit is responsible for using an image sensor (such as a camera) to capture the appearance image of the high-voltage cable, including image information of insulators, cable sheaths and other parts. The mechanical parameter acquisition unit is responsible for using equipment such as tension sensors. The data acquisition control unit is responsible for coordinating the work of each acquisition unit to ensure the synchronization and accuracy of data acquisition. At the same time, the collected data is preliminarily processed and integrated to prepare for subsequent data transmission and processing. The image acquisition unit, mechanical parameter acquisition unit, and environmental parameter acquisition unit transmit the data they have collected to the data acquisition control unit. The data acquisition control unit sends synchronous control signals or instructions to enable each acquisition unit to collect data at the same time or at a predetermined time interval. During the data transmission process, if a certain acquisition unit fails or the data is abnormal (such as exceeding the normal range, signal loss, etc.), the data acquisition control unit will perform exception processing, such as recording error information, sending alarm signals, etc., and promptly notify the system administrator or maintenance personnel for processing.

[0156] Automated preprocessing module: responsible for the preliminary processing of the collected raw data, including image cropping, resizing, denoising, data standardization, etc., using machine learning algorithms to achieve intelligent cropping and resizing of images to improve processing accuracy. At the same time, standardize multi-source data to ensure data consistency and comparability. The automated preprocessing module includes an image preprocessing unit, a data standardization unit, and a machine learning auxiliary unit. The image preprocessing unit is responsible for the preliminary processing of image data, including cropping, resizing, denoising, etc. The image preprocessing unit includes a cropping subunit, a resizing subunit, and a denoising subunit. The cropping subunit is responsible for automatically cropping the image according to preset rules or machine learning algorithms to remove useless or interfering parts. The resizing subunit is responsible for adjusting the image to a uniform size for subsequent processing and analysis. The denoising subunit is responsible for applying a denoising algorithm to reduce noise in the image and improve image quality. The image preprocessing unit receives data from the data acquisition The image data of the module is processed and the results are passed to the data standardization unit. The data standardization unit is responsible for standardizing multi-source data to ensure data consistency and comparability. The data standardization unit includes a numerical standardization sub-unit and a classification data encoding sub-unit. The numerical standardization sub-unit is responsible for normalizing or standardizing numerical data so that they have the same dimension or distribution range. The classification data encoding sub-unit is responsible for encoding classification data, such as converting text labels into numerical labels. The machine learning auxiliary unit is responsible for using machine learning algorithms to optimize image cropping, size adjustment and other processing processes to improve the degree of automation and processing accuracy.

[0157] Figure 3 : is a flowchart of parallel processing of the image preprocessing unit provided by an optional embodiment of the present invention, such as Figure 3 As shown, the data acquisition module transmits the collected data to multiple sub-units of the image pre-processing unit at the same time. Figure 4 Flowchart of serial processing of the image preprocessing unit provided by an optional embodiment of the present invention, such as Figure 4 As shown, the data acquisition module transmits the acquired data to the cropping subunit of the image pre-processing unit.

[0158] Data augmentation and annotation module: responsible for expanding the original data set, improving the robustness of the model, and formulating standardized annotation specifications.

[0159] Model Training and Optimization Module: Responsible for training and optimizing the detection model. The model training and optimization module includes an insulation state sample library construction unit, a data enhancement unit, a model training unit, a model tuning and optimization unit, and a regularization and ensemble learning unit. The insulation state sample library construction unit is responsible for collecting, organizing, and labeling various insulation state samples and building an insulation state sample library. The data enhancement unit is responsible for applying data enhancement technology to process the insulation state sample dataset to increase sample diversity. The model training unit is responsible for training the deep convolutional neural network using the enhanced insulation state sample dataset. The model tuning and optimization unit is responsible for tuning parameters and optimizing hyperparameters of the initially trained detection model. The regularization and ensemble learning unit is responsible for introducing regularization technology and ensemble learning methods to prevent model overfitting. The labeled insulation state sample dataset output by the insulation state sample library construction unit serves as the input of the data enhancement unit. The enhanced insulation state sample dataset output by the data enhancement unit serves as the input of the model training unit. The initially trained detection model and validation dataset output by the model training unit serve as the input of the model tuning and optimization unit. The optimized detection model output by the model tuning and optimization unit serves as the input of the regularization and ensemble learning unit.

[0160] Parallel processing and real-time analysis module: responsible for parallel processing and analysis of large amounts of data. The parallel processing and real-time analysis module includes a data distribution unit, a parallel processing unit, a real-time analysis algorithm unit, a result summary unit, and a result output unit. The data distribution unit is responsible for distributing the large amount of collected data to multiple processing nodes. The parallel processing unit contains multiple processing nodes or processors. Each node is responsible for processing a portion of the data. These nodes can work independently or collaboratively to perform data processing and analysis tasks in parallel. The real-time analysis algorithm unit is responsible for further analyzing the processed data using stream processing algorithms and outputting detection results. The result summary unit is responsible for summarizing and integrating the output results of each parallel processing unit to form the final detection result. The result output unit is responsible for outputting the detection results (such as graphical interfaces, reports, alarm signals, etc.) to users or other parts of the system. The data distribution unit distributes the large amount of collected data to each parallel processing unit. The parallel processing unit sends the processed data to the real-time analysis algorithm unit for further analysis. The real-time analysis algorithm unit sends the analysis results to the result summary unit for summary and integration. The result summary unit sends the final results to the result output unit for output.

[0161] Decision-making and early warning module: Responsible for combining image data with the temporal changes of mechanical and environmental parameters based on real-time analysis results using the Transformer model, capturing cross-modal correlations through the self-attention mechanism, and evaluating and judging the insulation status of high-voltage cables. Once potential safety hazards or abnormal conditions are discovered, the early warning mechanism is immediately triggered to notify relevant personnel for processing. The decision-making and early warning module includes a status evaluation unit, an early warning trigger unit, a notification sending unit, a detection report generation unit and a statistical information generation unit. The status evaluation unit is responsible for receiving the output results of the parallel processing and real-time analysis module, and evaluating and judging the insulation status of the high-voltage cable. The early warning trigger unit is responsible for judging whether to trigger the early warning mechanism based on the evaluation results of the status evaluation unit. The notification sending unit is responsible for sending early warning notifications to relevant personnel when the early warning is triggered. The detection report generation unit is responsible for generating a detailed detection report based on the real-time analysis results and status evaluation results. The report content includes: the insulation status of the cable, detection time, potential problems, recommended measures, etc. The statistical information generation unit is responsible for performing statistical analysis on the detection data and generating statistical information, such as failure rate, trend analysis, etc. The status evaluation unit passes the evaluation results to the early warning trigger unit. When the early warning is triggered, the early warning trigger unit passes the early warning information to the notification sending unit.

[0162] User interaction and remote monitoring module: responsible for providing the user interface and remote monitoring functions, enabling operators to remotely view real-time images and detection data of high-voltage cables. At the same time, it supports users to customize detection parameters and view historical detection records, improving the flexibility and ease of use of the system.

[0163] The data acquisition module passes the original data (such as images, mechanical parameters, environmental parameters, etc.) to the automatic preprocessing module, and the data preprocessed by the automatic preprocessing module is passed to the data enhancement and annotation module. The data set and annotation information enhanced by the data enhancement and annotation module are used for model training. The model training and optimization module passes the trained model to the parallel processing and real-time analysis module for real-time detection. The real-time analysis results generated by the parallel processing and real-time analysis module (such as insulation status assessment, anomaly detection, etc.) are passed to the decision-making and early warning module.

[0164] Monitor abnormal conditions such as image changes in the insulator area, abnormal fluctuations in mechanical parameters, and sudden changes in environmental parameters. Based on real-time analysis results, determine whether the insulation condition of the high-voltage cable is good or abnormal. Combine historical data and experience knowledge to further analyze and judge the abnormal conditions.

[0165] Once a potential safety hazard or abnormal situation is discovered, the early warning mechanism will be triggered immediately, and relevant personnel will be notified through sound and light alarms, SMS notifications, email reminders, etc. to handle the situation, and detailed detection reports and statistical information will be generated for subsequent analysis and management.

[0166] Through the above optional implementation, at least the following beneficial effects can be achieved:

[0167] (1) The automated preprocessing module uses convolutional neural networks to optimize image cropping, resizing, and other processing. This automated processing method greatly reduces manual intervention and improves preprocessing efficiency, especially when the amount of data is large, which can significantly shorten the processing time.

[0168] (2) The parallel processing and real-time analysis module uses parallel computing technology to process multiple images or data sets simultaneously, further accelerating the preprocessing speed. This parallel processing capability is crucial for processing large-scale data sets and ensures the stability and response speed of the system under high load;

[0169] (3) The system not only processes image data but also integrates multi-source data such as tension, temperature, and wind speed. This multi-source data fusion method enables the system to more comprehensively evaluate the insulation status of high-voltage cables and improve the accuracy and reliability of detection.

[0170] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0171] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0172] Example 2

[0173] According to an embodiment of the present invention, a device for implementing the above-mentioned method for determining the insulation status of a cable is also provided. Figure 5 : is a structural block diagram of a device for determining the insulation status of a cable according to an embodiment of the present invention, Figure 5As shown, the device includes: an acquisition module 502, a first determination module 504, a segmentation module 506, a splicing module 508, a recognition module 510 and a second determination module 512. The device will be described in detail below.

[0174] An acquisition module 502 is used to acquire a cable image corresponding to a target cable; a first determination module 504 is connected to the acquisition module 502 and is used to determine a regional image based on the cable image, wherein the characteristic intensity of the insulation state feature corresponding to the regional image is greater than a predetermined intensity threshold; a segmentation module 506 is connected to the first determination module 504 and is used to segment the regional image to obtain a plurality of first sub-images; a splicing module 508 is connected to the segmentation module 506 and is used to input the plurality of first sub-images into the splicing modules corresponding to the plurality of target layers of the target model respectively to obtain a plurality of second sub-images, wherein the corresponding splicing modules are provided with a plurality of The target model is obtained by training with sample data according to the predetermined splicing size. The recognition module 510 is connected to the splicing module 508 and is used to input the multiple second sub-images into the feature recognition modules corresponding to the multiple target layers respectively, to obtain image feature data corresponding to the multiple second sub-images, wherein the corresponding feature recognition module is provided with a corresponding feature recognition size, and the corresponding feature recognition size is used to identify the image feature data corresponding to the corresponding sub-image in units of the corresponding pixel size. The second determination module 512 is connected to the recognition module 510 and is used to determine the target insulation state corresponding to the target cable based on the multiple image feature data.

[0175] It should be noted here that the above-mentioned acquisition module 502, first determination module 504, segmentation module 506, splicing module 508, identification module 510 and second determination module 512 correspond to steps S102 to S112 in the method for determining the insulation status of the cable. The examples and application scenarios implemented by the multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0176] Example 3

[0177] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute the instructions to implement any of the above methods for determining the insulation status of a cable.

[0178] Example 4

[0179] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above methods for determining the insulation status of a cable.

[0180] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0181] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0182] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0183] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0184] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0185] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0186] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for determining the insulation state of a cable, characterized in that: include: acquiring a cable image corresponding to a target cable; Determining a region image based on the cable image, wherein a feature intensity of an insulation state feature corresponding to the region image is greater than a predetermined intensity threshold; Segmenting the region image to obtain a plurality of first sub-images; Inputting the plurality of first sub-images into stitching modules corresponding to the plurality of target layers of the target model, respectively, to obtain a plurality of second sub-images, wherein the corresponding stitching modules are provided with corresponding predetermined stitching sizes, and the target model is obtained by training with sample data; Inputting the plurality of second sub-images into feature recognition modules corresponding to the plurality of target layers respectively, to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature recognition modules are provided with corresponding feature recognition sizes, and the corresponding feature recognition modules recognize image feature data corresponding to the corresponding sub-images according to the corresponding feature recognition sizes; A target insulation state corresponding to the target cable is determined based on the plurality of image feature data.

2. The method according to claim 1, characterized in that Determining a target insulation state corresponding to the target cable based on the plurality of image feature data includes: Acquiring cable performance parameters corresponding to the target cable; Determining timing characteristic data corresponding to the target cable based on the cable performance parameters; Determining a fusion feature corresponding to the target cable based on the time series feature data and the plurality of image feature data; The insulation state corresponding to the target cable is determined based on the fusion feature.

3. The method according to claim 1, characterized in that Determining a regional image based on the cable image includes: Retrieving an image processing model, wherein the image processing model is set with target parameters, and the target parameters are determined based on initial parameters corresponding to the initial image model and a plurality of sample images; Inputting the cable image into the image processing model to obtain target position data corresponding to a target frame, wherein the target frame is used to determine an area in the cable image where the characteristic intensity of the insulation state feature is greater than a predetermined intensity threshold; The area image is determined according to the target position data.

4. The method according to claim 3, characterized in that Before calling the image processing model, the method further includes: Determining an error function with the goal of minimizing an error value, wherein the error function is used to determine an error value between predicted position data and sample position data, the predicted position data being position data of a target frame determined based on the initial image model and the sample image, and the sample position data being actual position data of the target frame corresponding to the sample image; updating initial parameters corresponding to the initial image model according to the error function to obtain updated parameters; When the error value corresponding to the updated parameter is less than the error threshold, determining the updated parameter as the target parameter; The image processing model is determined according to the target parameters and the initial image model.

5. The method according to claim 3, characterized in that Before calling the image processing model, the method further includes: Acquire a plurality of original images, wherein a first number corresponding to the plurality of original images is less than a predetermined number threshold, and the cables corresponding to the plurality of original images respectively all contain corresponding defective areas; Determining defect images corresponding to a plurality of defect regions and defect labels corresponding to the plurality of defect regions, respectively, based on the plurality of original images; The plurality of sample images are determined according to the plurality of defect images and the plurality of defect labels, wherein a second quantity corresponding to the plurality of sample images is greater than a predetermined quantity threshold.

6. The method according to claim 5, characterized in that The determining of a plurality of sample images based on a plurality of defect images and a plurality of defect labels includes: determining a plurality of initial images based on the plurality of defect images and the plurality of defect labels; Determining evaluation values ​​corresponding to each of the multiple initial images; When the corresponding evaluation value is greater than the scoring threshold, the corresponding initial image is determined to be a sample image.

7. The method according to any one of claims 1 to 6, characterized in that Determining the insulation state corresponding to the target cable based on the plurality of image feature data includes: Determining status scores corresponding to a plurality of insulation classification states respectively according to the plurality of image feature data; A target insulation state is determined from the plurality of insulation classification states according to the plurality of state scores.

8. A device for determining the insulation status of a cable, characterized in that: include: an acquisition module, used for acquiring a cable image corresponding to a target cable; A first determining module is configured to determine a region image based on the cable image, wherein a feature intensity of an insulation state feature corresponding to the region image is greater than a predetermined intensity threshold; a segmentation module, configured to segment the regional image to obtain a plurality of first sub-images; a stitching module, configured to input the plurality of first sub-images into stitching modules corresponding to the plurality of target layers of the target model, respectively, to obtain a plurality of second sub-images, wherein the corresponding stitching modules are provided with corresponding predetermined stitching sizes, and the target model is obtained by training with sample data; an identification module, configured to input the plurality of second sub-images respectively into feature identification modules corresponding to the plurality of target layers, to obtain image feature data corresponding to the plurality of second sub-images, wherein the corresponding feature identification modules are provided with corresponding feature identification sizes, and the corresponding feature identification sizes are used to identify the image feature data corresponding to the corresponding sub-images in units of corresponding pixel sizes; The second determining module is configured to determine a target insulation state corresponding to the target cable according to a plurality of image feature data.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method for determining the insulation status of a cable according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the method for determining the insulation state of a cable according to any one of claims 1 to 7.

Images