Insulator deterioration detection method and related device

Abstract

The invention belongs to the technical field of power equipment detection, and particularly discloses an insulator degradation detection method and related device.The insulator degradation detection method comprises the steps that an insulator infrared image is obtained; inputting the insulator infrared image into a pre-trained degradation detection model, and outputting a detection result; wherein the degradation detection model is obtained by using a training data set to train an improved Faster R-CNN model, the training data set comprises a degraded insulator infrared image and a corresponding degradation label, the improved Faster R-CNN model introduces a feature pyramid network, and a ResNet101 backbone network is used to replace a VGG16 feature extraction network. According to the invention, the detection efficiency and accuracy of the insulator in a complex scene can be improved.

Description

Technical Field

[0001] The invention belongs to the technical field of power equipment detection, and in particular relates to an insulator degradation detection method and a related device. Background Art

[0002] In the power system, the reliability of insulator performance is directly related to the safe and stable operation of the entire power grid. In the long-term operation of insulators, they may be affected by multiple factors such as environmental factors (such as temperature, humidity, pollution, etc.), mechanical stress, and natural aging of materials, resulting in gradual deterioration of insulation performance. This degradation will not only reduce the working efficiency of insulators, but may also cause power grid safety accidents, causing serious impacts on people's production and life. Therefore, timely and accurate status monitoring and fault diagnosis of insulators are of great significance to ensure the safe and stable operation of the power system.

[0003] Traditional insulator detection methods mainly include voltage distribution method, pulse corona current method and electric field measurement method. These methods can evaluate the performance status of insulators to a certain extent, but they all have some obvious shortcomings. For example, although the voltage distribution method is easy to operate, its measurement results are easily affected by external factors such as environmental humidity, resulting in reduced accuracy; the pulse corona current method requires direct contact with the insulator for detection, which may cause additional damage to the insulator, and the detection efficiency is relatively low; the electric field measurement method requires professional measurement equipment and complex operation process, which is not conducive to promotion in practical applications.

[0004] In recent years, infrared imaging, as an emerging non-contact detection technology, has received extensive attention in the field of insulator condition monitoring. This method can intuitively reflect the working status and performance changes of insulators by detecting the temperature distribution on the surface of the insulator. The infrared imaging method has the advantages of non-contact, high reliability and large-range temperature measurement, and is particularly suitable for real-time monitoring and fault diagnosis of power equipment. The analysis and processing of insulator infrared images require high technical requirements. With the continuous development of artificial intelligence technology, especially the rise of deep learning technology, it provides new ideas for insulator condition monitoring. Among them, the Faster R-CNN model, as an advanced target detection algorithm, has been well applied in the fields of image recognition and target detection. The model effectively improves the speed and accuracy of target detection by introducing the region proposal network (RPN). Although the Faster R-CNN model performs well in target detection, its application in the field of power equipment is still relatively limited. Especially when processing large-scale insulator infrared images, how to make full use of the advantages of the Faster R-CNN model and improve data processing efficiency and accuracy is still an urgent problem to be solved. In addition, the model also has certain deficiencies in extracting deep semantic information, especially when the contrast and signal-to-noise ratio of the insulator infrared image are low, it is often difficult to accurately identify the target object, resulting in poor detection effect. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides an insulator degradation detection method and related devices, which can improve the detection efficiency and accuracy of insulators in complex scenarios.

[0006] In order to solve the above technical problems, the present invention is implemented by the following technical solutions:

[0007] According to a first aspect of the present invention, there is provided a method for detecting insulator degradation, comprising:

[0008] Acquire infrared images of insulators;

[0009] The infrared image of the insulator is input into a pre-trained degradation detection model, and a detection result is output; wherein the degradation detection model is obtained by training an improved Faster R-CNN model using training data, the training data includes infrared images of degraded insulators and corresponding degradation labels, and the improved Faster R-CNN model introduces a feature pyramid network and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

[0010] In a possible implementation manner of the first aspect, the improved Faster R-CNN model also includes an improved anchor frame, and the size and proportion of the improved anchor frame are optimized based on the infrared thermal characteristics of the degraded insulator using a K-means clustering algorithm.

[0011] In a possible implementation of the first aspect, the optimization based on the infrared thermal characteristics of the deteriorated insulator using a K-means clustering algorithm includes:

[0012] (1) Extract the coordinates of the bounding boxes of all degraded insulators from the training dataset and convert the coordinates into the width and height of the bounding boxes;

[0013] (2) Initialize k anchor boxes and use a random function to randomly select k values ​​from all bounding boxes as initial values, where k is not less than 2;

[0014] (3) simulate the distance error between the bounding box and the anchor box;

[0015] (4) Select the anchor box with the smallest distance error and classify it to obtain the corresponding bounding box;

[0016] (5) Determine the position of each bounding box in the anchor box and determine the size of the anchor box based on the width, height, and midpoint of the bounding box;

[0017] (6) Repeat steps (3), (4) and (5) until the type of the bounding box completely matches the type of the anchor box and finally determine the anchor box ratio.

[0018] In a possible implementation manner of the first aspect, the ResNet101 backbone network is used to deeply extract feature maps of insulator infrared images.

[0019] In a possible implementation of the first aspect, the feature pyramid network is used to extract multi-scale information of the feature map, the multi-scale information includes low-level high-resolution features and upper-level semantic information, and the feature pyramid network is also used to perform feature fusion on the low-level high-resolution features and the upper-level semantic information.

[0020] In a possible implementation manner of the first aspect, after acquiring the infrared image of the insulator, the method further includes:

[0021] The insulator infrared image is preprocessed, and the preprocessing includes image enhancement and normalization.

[0022] In a possible implementation manner of the first aspect, after outputting the detection result, the method further includes:

[0023] When the test results show that the degree of insulator degradation exceeds the set threshold, an early warning will be issued.

[0024] According to a second aspect of the present invention, there is provided an insulator degradation detection device, comprising:

[0025] An acquisition module, used for acquiring infrared images of insulators;

[0026] A detection module is used to input the infrared image of the insulator into a pre-trained degradation detection model and output a detection result; wherein the degradation detection model is obtained by training an improved Faster R-CNN model using training data, the training data includes infrared images of degraded insulators and corresponding degradation labels, and the improved Faster R-CNN model introduces a feature pyramid network and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

[0027] In a possible implementation manner of the second aspect, the improved Faster R-CNN model also includes an improved anchor frame, and the size and proportion of the improved anchor frame are optimized based on the infrared thermal characteristics of the degraded insulator using a K-means clustering algorithm.

[0028] In a possible implementation manner of the second aspect, the optimization based on the infrared thermal characteristics of the deteriorated insulator using a K-means clustering algorithm includes:

[0029] (1) Extract the coordinates of the bounding boxes of all degraded insulators from the training dataset and convert the coordinates into the width and height of the bounding boxes;

[0030] (2) Initialize k anchor boxes and use a random function to randomly select k values ​​from all bounding boxes as initial values, where k is not less than 2;

[0031] (3) simulate the distance error between the bounding box and the anchor box;

[0032] (4) Select the anchor box with the smallest distance error and classify it to obtain the corresponding bounding box;

[0033] (5) Determine the position of each bounding box in the anchor box and determine the size of the anchor box based on the width, height, and midpoint of the bounding box;

[0034] (6) Repeat steps (3), (4) and (5) until the type of the bounding box completely matches the type of the anchor box and finally determine the anchor box ratio.

[0035] In a possible implementation manner of the second aspect, the ResNet101 backbone network is used to deeply extract feature maps of insulator infrared images.

[0036] In a possible implementation of the second aspect, the feature pyramid network is used to extract multi-scale information of the feature map, the multi-scale information includes low-level high-resolution features and upper-level semantic information, and the feature pyramid network is also used to perform feature fusion on the low-level high-resolution features and the upper-level semantic information.

[0037] In a possible implementation manner of the second aspect, after acquiring the infrared image of the insulator, the method further includes:

[0038] The insulator infrared image is preprocessed, and the preprocessing includes image enhancement and normalization.

[0039] According to a third aspect of the present invention, there is provided a device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the insulator degradation detection method when executing the computer program.

[0040] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the insulator degradation detection method is implemented.

[0041] Compared with the prior art, the present invention has at least the following beneficial effects:

[0042] The present invention provides an insulator degradation detection method, which enhances the model's detection capability of insulator degradation features at different scales by introducing a feature pyramid network. The feature pyramid network can integrate feature information from low to high layers, so that the model can more accurately capture small targets and detail features in the insulator infrared image, thereby improving the detection accuracy. At the same time, the feature pyramid network reduces the processing of redundant information by integrating features at different levels, and also indirectly improves the detection speed of the model. The ResNet101 backbone network is used to replace the traditional VGG16 feature extraction network. The ResNet101 network has a deeper hierarchical structure and stronger feature extraction capability, can retain more information features, and further improves the detection accuracy. In addition, compared with the VGG16 feature extraction network, the ResNet101 backbone network has less computational effort and higher computational efficiency, so that the model can obtain detection results faster when processing large-scale insulator infrared image data, thereby improving the detection efficiency. The present invention enhances the model's ability to extract deep semantic information by combining the feature pyramid network and the ResNet101 backbone network, so that the model can maintain a high detection accuracy even in complex scenes, thereby enhancing the robustness of the detection. The multi-scale feature extraction capability of the feature pyramid network enables the model to adapt to the infrared images of insulators under different environmental conditions, thereby improving the adaptability of the detection. In summary, the present invention can improve the detection efficiency and accuracy of insulators in complex scenes.

[0043] Furthermore, the size and proportion of the improved anchor frame are optimized based on the infrared thermal characteristics of degraded insulators using the K-means clustering algorithm, which can better adapt to small targets and insulators with different aspect ratios.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the specific implementation modes of the present invention, the drawings required for use in the description of the specific implementation modes will be briefly introduced below. Obviously, the drawings described below are some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 This is a flow chart of a method for detecting insulator degradation according to an embodiment of the present invention;

[0047] Figure 2 This is a flow chart of model training in the embodiment;

[0048] Figure 3This is a schematic diagram of data enhancement preprocessing in the embodiment;

[0049] Figure 4 A typical sample of the training set of the embodiment of the present invention;

[0050] Figure 5 Detection results under different confidence thresholds in the embodiment of the present invention;

[0051] Figure 6 Part of the test results of the deteriorated insulator at position 2 of the embodiment of the present invention;

[0052] Figure 7 The confidence score test result of the deteriorated insulator at position 2 of the embodiment of the present invention;

[0053] Figure 8 Part of the test results of the deteriorated insulator at position 5 of the embodiment of the present invention;

[0054] Fig. 9 The confidence score test result of the deteriorated insulator at position 5 of the embodiment of the present invention;

[0055] Fig.10 Part of the test results of the deteriorated insulator at position 7 of the embodiment of the present invention;

[0056] Fig.11 This is the confidence score test result of the deteriorated insulator at position No. 7 of Example 1 of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0058] like Figure 1 As shown, an insulator degradation detection method provided by an embodiment of the present invention aims to realize efficient and accurate detection of an insulator infrared image, and specifically comprises the following steps:

[0059] Step 1: Obtain the infrared image of the insulator.

[0060] Specifically, an infrared thermal imager is used to take non-contact photos of insulators in the power system to obtain temperature distribution images on their surfaces. It should be understood that the selection of an infrared thermal imager should ensure that it has high resolution and sensitivity so as to capture subtle temperature differences on the surface of the insulator. When shooting, ensure that environmental conditions (such as light, temperature, etc.) are relatively stable to reduce the impact of external factors on image quality. Exemplarily, the infrared thermal imager is a Fluke Ti55 infrared imager.

[0061] Step 2: Input the infrared image of the insulator into a pre-trained degradation detection model and output the detection result. The degradation detection model is obtained by training an improved Faster R-CNN model using training data, wherein the training data includes infrared images of degraded insulators and corresponding degradation labels. The improved Faster R-CNN model introduces a Feature Pyramid Network (FPN) and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

[0062] Specifically, in order to train the improved Faster R-CNN model, a training dataset containing a large number of infrared images of degraded insulators and corresponding degradation labels needs to be constructed. It should be noted that these images should cover insulators of different degrees of degradation, different environmental conditions, and different models to ensure the generalization ability of the model. The degradation labels are obtained by manual annotation, including information such as the location and degree of insulator degradation. It should also be understood that in order to enhance the robustness of the model, a certain number of normal insulator infrared images can also be added to the training dataset as negative samples.

[0063] It should be noted that the original Faster R-CNN model usually uses the VGG16 feature extraction network as the backbone network. However, the VGG16 feature extraction network is relatively weak in feature extraction capability and has a large amount of computation. Therefore, the present invention replaces the backbone network with ResNet101. It should be understood that ResNet101 effectively alleviates the gradient vanishing problem in deep neural networks by introducing residual connections and utilizing the identity mapping structure of ResNet, enabling the model to extract features at a deeper level while maintaining a low computational complexity.

[0064] Specifically, ResNet101 is a deep convolutional neural network with powerful feature extraction capabilities. In the present invention, the ResNet101 backbone network is used to deeply extract feature maps of insulator infrared images, which contain rich image information and details.

[0065] It should also be noted that the traditional Faster R-CNN model has difficulties in detecting small targets because it performs detection on high-level feature maps, which often lack detailed information. In order to overcome this shortcoming, the present invention introduces a feature pyramid network (FPN) in the Faster R-CNN model. FPN enables the model to perform target detection at multiple scales simultaneously by fusing feature maps of different scales, thereby improving the detection capability of small targets. In other words, the feature pyramid network is used to extract multi-scale information of feature maps, and the multi-scale information includes low-level high-resolution features and upper-level semantic information. The feature pyramid network is also used to perform feature fusion of low-level high-resolution features and upper-level semantic information.

[0066] Specifically, FPN extracts feature maps from different levels of the ResNet101 backbone network. These feature maps contain information of different scales from low to high layers. Among them, the low-level feature maps have higher resolution and can capture more detailed information; while the high-level feature maps have stronger semantic information and can identify more complex image content. FPN also fuses low-level high-resolution features and upper-level semantic information through upsampling and lateral connection. Through fusion, the feature maps of each scale contain rich information, which not only retains the detailed information, but also enhances the semantic information, improves the expressiveness of the features, and greatly improves the model's ability to detect targets of different scales.

[0067] The constructed training data set is input into the improved Faster R-CNN model for model training. It should be noted that during the training process, the back propagation algorithm and the gradient descent algorithm are used to optimize the parameters of the model so that the model can accurately identify the degraded areas in the infrared image of the insulator. Preferably, in order to prevent the model from overfitting, data enhancement (such as rotation, scaling, flipping, etc.) and regularization can also be used.

[0068] After the improved Faster R-CNN model is trained, a degradation detection model capable of detecting insulator degradation is obtained. The infrared image of the insulator to be detected is input into the trained degradation detection model, and the model will output the detection result. It should be understood that the detection result includes information such as the location, degree, and confidence of the insulator degradation. This information can be used for real-time monitoring and fault diagnosis of the insulator status. According to the detection results, the degradation of the insulator can be evaluated and corresponding maintenance measures can be taken.

[0069] The present invention provides an insulator degradation detection method, which improves the Faster R-CNN model by introducing a feature pyramid network and using ResNet101 as a backbone network, thereby achieving efficient and accurate detection of insulator infrared images.

[0070] As a more preferred implementation, the improved Faster R-CNN model further includes an improved anchor frame, the size and proportion of which are optimized using a K-means clustering algorithm based on the infrared thermal characteristics of the degraded insulator.

[0071] It should be noted that in the Faster R-CNN model, the anchor frame is a rectangular frame used to preliminarily locate the target object. Traditional anchor frame design is usually set based on experience or the size and proportion of common objects. The present invention uses the K-means clustering algorithm to perform statistical analysis on the size and proportion of degraded insulators in infrared images, and optimizes the design of the anchor frame accordingly.

[0072] Specifically, first, the bounding boxes of all degraded insulators are extracted from the training dataset. These bounding boxes are obtained through manual annotation and represent the location and size of the degraded area. Then, the K-means clustering algorithm is used to perform cluster analysis on these bounding boxes. The goal of clustering is to find a set of optimal anchor box sizes and proportions that can best cover and match the degraded insulators in the training dataset. The number of clusters (K value) can be determined experimentally to balance the detection accuracy and computational complexity of the model. Finally, based on the clustering results, the anchor box size and proportion in the Faster R-CNN model are adjusted to make it more consistent with the actual characteristics of degraded insulators in infrared images. The optimized anchor box can improve the detection accuracy and recall rate of the model for degraded insulators.

[0073] Based on the improved anchor frame, during the model training process, the improved Faster R-CNN model learns how to more accurately detect the degraded area in the infrared image of the insulator according to the optimized anchor frame. At the same time, the back propagation algorithm and gradient descent algorithm are used to optimize the parameters of the model, so that it can accurately identify degraded insulators in complex image environments.

[0074] In summary, this implementation achieves efficient and accurate detection of insulator infrared images by introducing FPN, replacing the ResNet101 backbone network, and optimizing the anchor frame design. In particular, the optimized anchor frame design further improves the detection accuracy and recall rate of the model for deteriorated insulators.

[0075] More specifically, regarding the infrared thermal characteristics of deteriorated insulators, the K-means clustering algorithm is used for optimization. The specific steps are as follows:

[0076] (1) First, the bounding box coordinates of all degraded insulators are extracted from the training dataset. These coordinates represent the location of the degraded area in the image. Then, the coordinates are converted into the width and height of the bounding box.

[0077] (2) Initialize k anchor boxes and use a random function to randomly select k values ​​from all bounding boxes as initial values, where k is not less than 2.

[0078] That is to say, select an appropriate k value (k is not less than 2), indicating the number of anchor boxes you want to get in the end, and use a random function to randomly select k values ​​from all bounding boxes as the width and height of the initial anchor boxes. These initial anchor boxes will serve as the starting point for clustering.

[0079] (3) Model the distance error between the bounding box and the anchor box.

[0080] That is to say, the distance error between each bounding box and all anchor boxes is calculated, and the distance error reflects the matching degree between the bounding box and the anchor box. Specifically, using the IOU value instead of the Euclidean distance can effectively reduce the error derived from the algorithm.

[0081] (4) Select the anchor box with the smallest distance error and classify it to obtain the corresponding bounding box.

[0082] Specifically, the anchor box with the smallest distance error is selected and assigned to the corresponding bounding box with the highest matching degree between the anchor box and the bounding box. Then, the size of the anchor box is updated according to the width and height of all bounding boxes assigned to the anchor box.

[0083] (5) Determine the position of each bounding box in the anchor box and determine the size of the anchor box based on the width, height, and midpoint of the bounding box.

[0084] That is, for each bounding box, its position in the updated anchor box is determined by comparing the midpoint position of the bounding box with the anchor box. Based on the width, height and midpoint of the bounding box, the size of the anchor box is further fine-tuned to ensure that it can more accurately cover the deteriorated insulator.

[0085] (6) Repeat steps (3), (4) and (5) until the type of the bounding box completely matches the type of the anchor box. Finally, a set of optimal anchor box sizes and ratios are determined. The optimal anchor box can best match the degraded insulators in the training dataset.

[0086] In an achievable manner, after acquiring the infrared image of the insulator, the method further includes: preprocessing the infrared image of the insulator to improve the image quality and expand the sample data set. The preprocessing includes image enhancement and normalization.

[0087] Among them, the exemplary image enhancement aims to improve the contrast and clarity of the image, making the degraded area more obvious in the image. Figure 3 As shown, the enhancement methods may include logarithmic transformation, rotation, translation, adding salt and pepper noise, and mixed enhancement, so as to expand the insulator infrared sample data set.

[0088] Normalization is to adjust the pixel values ​​of an image to a certain range, usually between 0 and 1 or -1 and 1, which helps to eliminate the brightness differences between different images and makes the model easier to learn and recognize. Normalization can be achieved through linear transformation, logarithmic transformation, power law transformation and other methods.

[0089] Exemplarily, when training a degradation detection model, the insulator infrared images are preprocessed, and the preprocessed insulator infrared images are used to form a sample set, which is divided into a training set and a verification set. The insulator infrared image training set is used to train an improved Faster R-CNN model, and the insulator infrared image verification set is used to verify the training model, and the model performance is evaluated based on the verification results.

[0090] In an achievable manner, after outputting the detection result, the method further includes: issuing an early warning when the detection result shows that the degree of insulator degradation exceeds a set threshold.

[0091] Specifically, after the test results are output, the degradation degree of the insulator is compared with the set threshold. When the test results show that the degradation degree of the insulator exceeds the set threshold, the early warning mechanism is triggered. It should be noted that the early warning method may include but is not limited to sounding an alarm, sending text messages or emails to notify relevant personnel, and displaying a red warning in the monitoring system.

[0092] After the warning is issued, relevant personnel can promptly carry out follow-up processing on the deteriorated insulators, such as on-site inspection, repair or replacement, etc. This embodiment performs timely warning processing after the detection is completed, which helps to improve the safety of the power system and reduce failures and accidents caused by insulator deterioration.

[0093] Example

[0094] In order to verify the effectiveness of the insulator degradation detection method proposed in this invention, this implementation case simulates the operating environment in the actual power system, selects degraded insulators with different resistance values ​​as test samples, uses infrared imaging technology to obtain their heating images, and constructs a data set for training and testing the Faster R-CNN model. Finally, by comparing the detection results of the model with the actual degradation situation, the accuracy and reliability of the method are evaluated.

[0095] Three deteriorated porcelain insulators replaced on site were selected as test samples, with resistance values ​​of 7.5MΩ, 19MΩ and 150MΩ respectively. Fluke Ti55 was selected as the infrared imager, and the emissivity was set to 0.9. Under different pollution levels, the temperature range was 7℃ to 25℃, the relative humidity was 20% to 80%, and the high-voltage power supply applied 65.9kV / 50Hz voltage for 2 hours. Voltage was applied to the insulator string at the high-voltage end, and the insulator was photographed every 0.5 hour with an infrared imager. After the experiment is completed, wait for the insulator to cool completely, replace it with a low-value insulator with different resistance values, and repeat the above steps. A series of infrared heating images of deteriorated insulator strings were recorded, and a total of 401 experimental infrared images were obtained. Deteriorated insulators were manually marked in the infrared images, and 401 corresponding .xml files were generated. According to the standard format of the VOC2007 data sample set, combined with Figure 2 As shown, create Faster R-CNN training and test sets.

[0096] like Figure 4 As shown, typical samples of the training set of this embodiment are shown. The degraded insulators are located at positions 2, 5, and 7, respectively. The infrared heating characteristics of the degraded insulators are that the steel cap is evenly heated, but the porcelain part is usually not heated or heated unevenly. At the same time, for normal insulators, the heated part is mainly concentrated at the connection between the steel cap and the porcelain part, or no heat is generated. Figure 5 The results show the detection effect of the degradation detection model on degraded insulators under different confidence thresholds. It can be seen that when the confidence threshold is set to a lower value (confidence is 0.1), the detection information related to degraded insulators will not be missed, but a large number of false detections will occur; when the confidence threshold is set to a higher value (confidence is 0.7), insulators will not be falsely detected, but they will be missed. Therefore, finding a reasonable confidence threshold is very important to balance the false detection rate and missed detection rate. Figure 6 Part of the test results for the deteriorated insulator at position 2. The deteriorated insulator detected can be accurately located. Figure 6 As can be seen from (a), (b), (c) and (d) in the figure, the present invention is not affected by the complex background experimental environment and can accurately identify the heating characteristics of the steel cap of the low-resistance insulator and the non-heating characteristics of the zero-resistance insulator among the degraded insulators. Figure 7 The confidence score test results of the deteriorated insulator at position 2 are shown. It can be seen that Figure 7 The accuracy and confidence scores of the identification of the deteriorated insulator at position 2 were maintained at a relatively high level. Figure 8 Part of the test results of the deteriorated insulator at position 5. It can be seen that the Figure 8 (a) in the figure corresponds to the degraded insulator, while the remaining test images (b), (c) and (d) successfully achieve accurate identification and positioning of the degraded insulator. Fig. 9The confidence score test results of the deteriorated insulator at position 5 are shown. It can be seen that Fig. 9 The accuracy and confidence scores of the identification of the deteriorated insulator at position 5 remained at a relatively high level. Fig.10 This is part of the test results of the deteriorated insulator at position 7. It can be seen that Fig.10 The deteriorated insulator corresponding to (a) is actually located in the seventh piece, but the deteriorated insulator detected is in the first piece. The remaining test images (b), (c) and (d) achieve accurate identification and positioning of the deteriorated insulator. Fig.11 The confidence score test results of the deteriorated insulator at position 7 are shown. It can be seen that Fig.11 The accuracy and confidence scores of the identification of the deteriorated insulator at position 7 remained at a relatively high level.

[0097] The following Table 1 shows the identification accuracy of degraded insulators at different confidence levels.

[0098] Table 1

[0099]

[0100] The model has different detection effects on degraded insulators at different confidence thresholds. When the confidence threshold is low, the false detection rate is high; when the confidence threshold is high, the missed detection rate is high. The detection results show that the model can accurately locate degraded insulators and is not affected by the complex background experimental environment. In particular, the model shows good recognition ability in the characteristics of steel cap heating of low-resistance insulators and non-heating of zero-resistance insulators. The confidence scores of the model are different for degraded insulators at different locations, but in general, the recognition accuracy and confidence scores of degraded insulators at specific locations remain at a relatively high level. The recognition accuracy of 24 degraded insulator test images at three different locations was evaluated at different confidence levels. The results show that when the confidence threshold is set to 0.5, the false detection rate and missed detection rate are relatively balanced, and the recognition accuracy is optimal.

[0101] This implementation case verifies the effectiveness of the insulator degradation detection method proposed in the present invention by simulating the operating environment in an actual power system. The experimental results show that the method can accurately identify and locate degraded insulators, and exhibits good stability under different experimental conditions. The feature extraction capability and detection performance of the model are improved.

[0102] In another embodiment of the present invention, an insulator degradation detection device is provided, which is used to implement the above-mentioned insulator degradation detection method. The insulator degradation detection device specifically includes an acquisition module and a detection module, and the specific configuration is as follows:

[0103] The acquisition module is used to acquire the infrared image of the insulator.

[0104] A detection module is used to input the infrared image of the insulator into a pre-trained degradation detection model and output a detection result; wherein the degradation detection model is obtained by training an improved Faster R-CNN model using training data, the training data includes infrared images of degraded insulators and corresponding degradation labels, and the improved Faster R-CNN model introduces a feature pyramid network and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

[0105] Specifically, the ResNet101 backbone network is used to deeply extract feature maps of insulator infrared images, which contain rich image information and details. The feature pyramid network extracts feature maps from different levels of the ResNet101 backbone network, which contain information of different scales from low to high levels. Among them, the low-level feature maps have higher resolution and can capture more detailed information; while the high-level feature maps have stronger semantic information and can identify more complex image content. FPN also fuses low-level high-resolution features and upper-level semantic information through upsampling and lateral connection. Through fusion, the feature maps of each scale contain rich information, which not only retains the detailed information, but also enhances the semantic information, improves the expressiveness of the features, and greatly improves the model's ability to detect targets of different scales.

[0106] All relevant contents of each step involved in the embodiment of the aforementioned insulator degradation detection method can be referred to the functional description of the functional module corresponding to the insulator degradation detection device in the embodiment of the present invention, and will not be repeated here. The division of modules in the embodiment of the present invention is schematic, which is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention can be integrated into a processor, or it can exist physically separately, or two or more modules can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of software functional modules.

[0107] In another embodiment of the present invention, a computer device is provided, the computer device comprising a processor and a memory, the memory being used to store a computer program, the computer program comprising program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of an insulator degradation detection method.

[0108] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the insulator degradation detection method in the above embodiment.

[0109] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0110] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0113] In the present invention, the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the described specific features, structures, materials or characteristics may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0114] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for detecting insulator degradation, characterized in that: include: Acquire infrared images of insulators; The infrared image of the insulator is input into a pre-trained degradation detection model, and a detection result is output; wherein the degradation detection model is obtained by training an improved Faster R-CNN model using a training data set, wherein the training data set includes infrared images of degraded insulators and corresponding degradation labels, and the improved Faster R-CNN model introduces a feature pyramid network and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

2. The insulator degradation detection method according to claim 1, characterized in that: The improved FasterR-CNN model also includes an improved anchor frame, the size and proportion of which are optimized based on the infrared thermal characteristics of the degraded insulator using a K-means clustering algorithm.

3. The insulator degradation detection method according to claim 2, characterized in that: The infrared thermal characteristics of the deteriorated insulator are optimized by using a K-means clustering algorithm, including: (1) Extract the coordinates of the bounding boxes of all degraded insulators from the training dataset and convert the coordinates into the width and height of the bounding boxes; (2) Initialize k anchor boxes and use a random function to randomly select k values ​​from all bounding boxes as initial values, where k is not less than 2; (3) simulate the distance error between the bounding box and the anchor box; (4) Select the anchor box with the smallest distance error and classify it to obtain the corresponding bounding box; (5) Determine the position of each bounding box in the anchor box and determine the size of the anchor box based on the width, height, and midpoint of the bounding box; (6) Repeat steps (3), (4) and (5) until the type of the bounding box completely matches the type of the anchor box and finally determine the anchor box ratio.

4. The insulator degradation detection method according to claim 1, characterized in that: The ResNet101 backbone network is used to deeply extract the feature map of the insulator infrared image.

5. The insulator degradation detection method according to claim 4, characterized in that: The feature pyramid network is used to extract multi-scale information of the feature graph, the multi-scale information includes low-level high-resolution features and upper-level semantic information, and the feature pyramid network is also used to perform feature fusion on the low-level high-resolution features and the upper-level semantic information.

6. The insulator degradation detection method according to claim 1, characterized in that: After obtaining the infrared image of the insulator, the method further includes: The insulator infrared image is preprocessed, and the preprocessing includes image enhancement and normalization.

7. The insulator degradation detection method according to claim 1, characterized in that: After outputting the detection result, the method further includes: When the test results show that the degree of insulator degradation exceeds the set threshold, an early warning will be issued.

8. An insulator degradation detection device, characterized in that: include: An acquisition module, used for acquiring infrared images of insulators; A detection module is used to input the infrared image of the insulator into a pre-trained degradation detection model and output a detection result; wherein the degradation detection model is obtained by training an improved Faster R-CNN model using training data, the training data includes infrared images of degraded insulators and corresponding degradation labels, and the improved Faster R-CNN model introduces a feature pyramid network and replaces the VGG16 feature extraction network with a ResNet101 backbone network.

9. A device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the insulator degradation detection method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, an insulator degradation detection method according to any one of claims 1 to 7 is implemented.

Images