A power transmission line icing treatment method, device, equipment, medium and product

Abstract

The application discloses a power transmission line icing treatment method, device, equipment, medium and product. The method comprises the following steps: if an icing detection request for a power transmission line is detected, a target icing image corresponding to the power transmission line and a pre-trained target detection model are determined; according to the target icing image, a multi-level segmentation module and a lightweight classification module in the target detection model are used to identify and process the target icing image, so that a corresponding icing type is obtained; according to the icing type, a corresponding measure for the icing of the power transmission line is determined, so that the icing on the power transmission line is treated according to the corresponding measure for the icing. The technical scheme of the application can identify and process the icing image by using the target detection model, so that the icing condition of the power transmission line is remotely analyzed, and the efficiency of the icing treatment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of big data, and in particular to a power transmission line icing treatment method, device, equipment, medium and product. BACKGROUND

[0002] Icing is a complex integrated physical phenomenon affected by many natural factors. In the process of power transmission, icing on the surface of the power transmission line is prone to induce various safety accidents. In the existing technical environment of the power grid, manual inspection is still the most important inspection method for power transmission line inspection. However, when the power transmission line encounters high-hazard weather such as blizzards and freezing rain, the actual operating environment of the power transmission line is even worse, and it is difficult for humans or vehicles to directly reach the scene, and even if they are on the scene, there is a high risk, making it more difficult to repair the work.

[0003] Therefore, how to use the target detection model to recognize and process the icing image, so as to remotely analyze the icing condition of the power transmission line and improve the efficiency of icing treatment, is a problem to be solved at present. SUMMARY

[0004] The present application provides a power transmission line icing treatment method, device, equipment, medium and product, which uses a target detection model to recognize and process an icing image, so as to remotely analyze the icing condition of the power transmission line and improve the efficiency of icing treatment.

[0005] According to an aspect of the present application, a power transmission line icing treatment method is provided, comprising:

[0006] If an icing detection request for the power transmission line is detected, a target icing image corresponding to the power transmission line and a pre-trained target detection model are determined;

[0007] According to the target icing image, a multi-level segmentation module and a lightweight classification module in the target detection model are used to recognize and process the target icing image to obtain a corresponding icing type;

[0008] According to the icing type, a corresponding measure for the icing on the power transmission line is determined, so that the icing on the power transmission line is treated according to the measure.

[0009] According to another aspect of the present application, a power transmission line icing treatment device is provided, comprising:

[0010] A determination module is configured to determine a target icing image corresponding to the power transmission line and a pre-trained target detection model if an icing detection request for the power transmission line is detected;

[0011] An identification module is configured to use a multi-level segmentation module and a lightweight classification module in the target detection model to recognize and process the target icing image according to the target icing image to obtain a corresponding icing type.

[0012] The processing module is configured to determine an icing response measure for the power transmission line according to the icing type, and to process the icing on the power transmission line according to the icing response measure.

[0013] According to another aspect of the present application, there is provided an electronic device, comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein

[0016] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the power transmission line icing processing method according to any one of the embodiments of the present application.

[0017] According to another aspect of the present application, there is provided a computer readable storage medium storing computer instructions for enabling a processor to perform the power transmission line icing processing method according to any one of the embodiments of the present application when executed by the processor.

[0018] According to another aspect of the present application, there is also provided a computer program product comprising a computer program, the computer program being configured to perform the power transmission line icing processing method according to any one of the embodiments of the present application when executed by a processor.

[0019] The technical solution of the embodiments of the present application, if an icing detection request for the power transmission line is detected, determines a target icing image corresponding to the power transmission line and a pre-trained target detection model; according to the target icing image, uses a multi-level segmentation module and a lightweight classification module in the target detection model to perform identification processing on the target icing image to obtain a corresponding icing type; according to the icing type, determines an icing response measure for the power transmission line, and processes the icing on the power transmission line according to the icing response measure. By using the target detection model to perform identification processing on the icing image, the icing condition of the power transmission line can be remotely analyzed, and the efficiency of icing processing can be effectively improved.

[0020] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to make the technical solution in the embodiments of the present application clearer, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some of the embodiments of the present application, and all other drawings obtained by those skilled in the art without any creative effort should be within the protection scope of the present application.

[0022] Figure 1A is a flow chart of a power transmission line icing treatment method provided by Embodiment One of the present application;

[0023] Figure 1B is a structural schematic diagram of a multi-level segmentation module provided by Embodiment One of the present application;

[0024] Figure 1C is a structural schematic diagram of a CP module provided by Embodiment One of the present application;

[0025] Figure 1D is a structural schematic diagram of a lightweight classification module provided by Embodiment One of the present application;

[0026] Figure 1E is a structural schematic diagram of a DS module provided by Embodiment One of the present application;

[0027] Figure 2 is a flow chart of a power transmission line icing treatment method provided by Embodiment Two of the present application;

[0028] Figure 3 is a structural block diagram of a power transmission line icing treatment device provided by Embodiment Three of the present application;

[0029] Figure 4 is a structural schematic diagram of an electronic device provided by Embodiment Four of the present application. DETAILED DESCRIPTION

[0030] In order to make the technical solution in the embodiments of the present application clearer, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some of the embodiments of the present application, and all other drawings obtained by those skilled in the art without any creative effort should be within the protection scope of the present application.

[0031] It should be noted that the terms "first", "second", "target", "candidate", "alternative" and the like in the description, claims, and drawings of the application are intended to distinguish similar objects, not necessarily describing a particular chronological or sequential order. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The acquisition, storage, use, processing and other data in the technical solutions of the present application comply with the relevant provisions of national laws and regulations.

[0032] Embodiment one

[0033] Figure 1A is a flowchart of a power line icing treatment method provided by an embodiment of the application; Figure 1B is a structural schematic diagram of a multi-level segmentation module provided by the embodiment one of the application; Figure 1C is a structural schematic diagram of a CP module provided by the embodiment one of the application; Figure 1D is a structural schematic diagram of a lightweight classification module provided by the embodiment one of the application; Figure 1E is a structural schematic diagram of a DS module provided by the embodiment one of the application. The embodiment can be applicable to the case that the icing image of the power line is recognized based on the target detection model to obtain the icing type, thereby helping to take corresponding strategies for subsequent processing. The method can be executed by a power line icing treatment device, which can be realized in the form of hardware and / or software, and can be configured in an electronic device, such as an electronic device with power line icing treatment function, and executed by a total control system in a power distribution network, such as Figure 1A as shown, the power line icing treatment method comprises:

[0034] S101, if the icing detection request of the power line is detected, the target icing image corresponding to the power line and the pre-trained target detection model are determined.

[0035] The icing detection request is a request for detecting icing conditions on the power transmission line. The target icing image is an image obtained by image collection of the power transmission line. The target icing image can represent icing conditions of the power transmission line. The target detection model is a pre-trained model for segmenting and classifying the target icing image. The target detection model can include a multi-level segmentation module and a lightweight classification module. The multi-level segmentation module includes a feature extraction layer, a convolution layer, and an up-sampling layer. The lightweight classification module can include a first classification convolution layer, an information extraction layer, a second classification convolution layer, an average pooling layer, and a full connection layer. The target detection model can be, for example, a neural network model.

[0036] Optionally, the general control system of the power distribution network can detect the ambient temperature of the power transmission line in the power grid in real time. If the ambient temperature is less than a preset icing temperature threshold, it is determined that icing detection of the power transmission line is required. At this time, it is considered that the icing detection request of the power transmission line is detected.

[0037] Optionally, if the icing detection request of the power transmission line is detected, the general control system of the power distribution network can use the image collection device corresponding to the power transmission line to collect the target icing image corresponding to the power transmission line.

[0038] Optionally, before determining the pre-trained target detection model, the method further includes: constructing a training sample set according to historical icing images and corresponding icing types; and training an initial detection model based on a cross-entropy loss function and a negative log-likelihood loss function to obtain the target detection model according to the training sample set.

[0039] The historical icing image is an image that can represent icing conditions of the power transmission line obtained by image collection of the outdoor power transmission line in a historical time period. The icing type can be, for example, normal no icing, rime type, snow type, glaze type, and mixed type. The mixed type refers to an icing type containing at least two of the rime type, the snow type, and the glaze type. The training sample set contains historical icing images of different icing types, which can be, for example, normal no icing power transmission line images, rime type icing power transmission line images, snow type icing power transmission line images, glaze type icing power transmission line images, and mixed type icing power transmission line images. The initial detection model can be an initialized neural network model. The target detection model is a model obtained by iterative updating parameters of the initial detection model based on the training sample set and the loss function. The initial detection model can include a multi-level segmentation module and a lightweight classification module. The multi-level segmentation module includes a feature extraction layer, a convolution layer, and an up-sampling layer. The lightweight classification module can include a first classification convolution layer, an information extraction layer, a second classification convolution layer, an average pooling layer, and a full connection layer.

[0040] Optionally, a pad filling method can be used to fill the specified value around each historical icing image, so as to unify the image size, and the Labelme labeling software is used to label the historical icing image according to the icing type corresponding to the historical icing image, so as to construct a training sample set. Wherein, pad is a method for modifying the image size in the data set preprocessing process of the neural network model, which does not change the original data form and does not lose the original information of the image. The Labelme labeling software is a commonly used tool for labeling image information.

[0041] Optionally, the training sample set can be divided into a training set, a validation set and a test set based on a ratio of 6:2:2, to train and verify the initial detection model. For example, a batch of training sample set data can be input into the initial detection model for training, and the training batch size (i.e. batch size) batch_size=16, the training batch epoch=200, the initial learning rate is 0.001, the multi-level segmentation module MUnet uses the cross-entropy loss function to calculate the segmentation loss value seg_loss, and the lightweight classification module MEnet uses the negative log-likelihood loss function to calculate the classification loss value cls_loss, and the total loss loss calculation formula is as follows:

[0042] loss=0.5×seg_loss+0.5×cls_loss

[0043] When the training of all batch data of the training sample set is completed, the training sample set can be input into the initial detection model in batches to obtain the corresponding batch loss value batch_loss. During training and verification, the initial detection model will automatically learn and adjust parameters according to the loss and batch_loss each time. When the loss and batch_loss values tend to converge and the difference between them is very small, the initial detection model training is completed, and the final target detection model is obtained.

[0044] S102、According to the target icing image, the multi-level segmentation module and the lightweight classification module in the target detection model are used to identify and process the target icing image, so as to obtain the corresponding icing type.

[0045] The multilevel segmentation module (MUnet) refers to a module for image segmentation in the target detection model, and the lightweight classification module (Mini Efficientnet) refers to a module for image classification in the target detection model. The multilevel segmentation module can include at least one feature extraction layer (which can be abbreviated as CP, which includes a convolution unit conv and a PSwish activation function processing unit). The lightweight classification module can include at least one information extraction layer (Depth and Spatial feature extraction module, DS).

[0046] Optionally, according to the target icing image, the multilevel segmentation module and the lightweight classification module in the target detection model are used to perform identification processing on the target icing image to obtain the corresponding icing type, including: performing self-splicing processing on the target icing image (denoted as M1) to obtain a first spliced icing image (denoted as M2); using the multilevel segmentation module in the target detection model to perform segmentation processing on the first spliced icing image to obtain a first segmented icing image (denoted as M3), and performing splicing processing on the first segmented icing image and the target icing image to obtain a second spliced icing image (denoted as M4); using the multilevel segmentation module in the target detection model to perform segmentation processing on the second spliced icing image to obtain a second segmented icing image (denoted as M5), and using the lightweight classification module in the target detection model to perform classification processing on the second segmented icing image to obtain the corresponding icing type.

[0047] For example, the size of the target icing image (denoted as M1) can be HxWxC, where H and W are the height and width of the image, and C is the number of channels of the image. The target icing image can be subjected to a concat splicing operation with itself, i.e., self-splicing processing, to obtain a feature map M2 with a size of HxWx2C, which is the first spliced icing image.

[0048] It should be noted that the concat splicing operation is to merge the channel numbers of two feature maps, and the size of the obtained feature map does not change, but the number of channels increases.

[0049] Exemplarily, after obtaining the first spliced icing image M2, M2 can be input to the MUnet for segmentation to obtain a feature map M3 extracting the transmission line embodiment and the line icing profile and position information. M3 is spliced with M1 to obtain a feature map M4, and M4 is input to the MUnet for segmentation to obtain a second segmented icing image M5. The pixel points in M5 are divided into three categories, i.e., transmission line, line icing and background, to realize in-image target segmentation. Finally, M5 is input to the MEnet for line icing classification to output normal no icing, rime type, snow type, fog type or mixed type of icing.

[0050] Exemplarily, after M3 is concatenated with M1, a feature map M4 with a size of HxWx2C can be obtained. M4 is input to the segmentation module MUnet for processing to obtain a second segmented icing image M5 with a size of HxWxC, and the image is divided into three mutually disjoint regions of transmission line, line icing and background.

[0051] Optionally, a multi-level segmentation module in a target detection model is used to segment the first spliced icing image to obtain a first segmented icing image, including: determining that the multi-level segmentation module in the target detection model includes a feature extraction layer, a convolution layer and an up-sampling layer; using each feature extraction layer, convolution layer and up-sampling layer to hierarchically extract, convolve and up-sample the first spliced icing image to segment the position information of the transmission line region, the icing region and the background region in the first spliced icing image, and mark in the first spliced icing image according to the position to obtain the first segmented icing image.

[0052] The number of feature extraction layers (CP) is at least two; the number of convolution layers (conv) is at least two; and the number of up-sampling layers (Upsample) is at least two.

[0053] Optionally, the first segmentation icing image is obtained by: using a first feature extraction layer, a second feature extraction layer, a third feature extraction layer and a fourth feature extraction layer to respectively extract and process the first spliced icing image M2 to obtain a first feature, a second feature, a third feature and a fourth feature (denoted as U1, U2, U3 and U4); performing up-sampling processing on the fourth feature, and performing addition processing and convolution processing on the fourth feature after the up-sampling processing and the third feature to obtain a fifth feature (denoted as U5); performing up-sampling processing on the fifth feature, and performing addition processing and convolution processing on the fifth feature after the up-sampling processing and the second feature to obtain a sixth feature (denoted as U6); performing up-sampling processing on the sixth feature, and performing addition processing on the sixth feature after the up-sampling processing and the first feature to obtain a seventh feature (denoted as U7); and sequentially performing convolution processing, up-sampling processing and twice convolution processing on the seventh feature to obtain position information of a power transmission line region, an icing region and a background region in the first spliced icing image, and marking in the first spliced icing image according to the positions to obtain the first segmentation icing image M3.

[0054] wherein the addition processing refers to an Add operation, specifically adding the values of pixel points in the feature map one by one. The first feature extraction layer, the second feature extraction layer, the third feature extraction layer and the fourth feature extraction layer all essentially are feature extraction layers containing a convolution unit conv and a PSwish activation function processing unit, which can be denoted as a CP module.

[0055] For example, Figure 1B The feature map M2 with a size of HxWx2C is input to the first feature extraction layer (CP module) to obtain a feature map U1 with a size of H / 2xW / 2x64, that is, the first feature, and the feature map U1 is sequentially input to three CP modules to obtain a feature map U2 with a size of H / 4xW / 4x128, a feature map U3 with a size of H / 8xW / 8x256 and a feature map U4 with a size of H / 16xW / 16x512, that is, the first feature, the second feature, the third feature and the fourth feature.

[0056] Further, the U4 can be input into an up-sampling layer Upsample for up-sampling, and the obtained result is subjected to an Add operation with the U3, and then passes through two conv layers with a kernel size of 3*3, a step of 1, and padding of 1 to obtain a feature map U5 with a size of H / 8*W / 8*256. The U5 is input into the up-sampling layer Upsample for up-sampling, and the obtained result is subjected to an Add operation with the feature map U2, and then passes through two conv layers with a kernel size of 3*3, a step of 1, and padding of 1 to obtain a feature map U6 with a size of H / 4*W / 4*128. The U6 is input into the up-sampling layer Upsample for up-sampling, and the obtained result is subjected to an Add operation with the U1 to obtain a feature map U7 with a size of H / 2*W / 2*64.

[0057] Finally, the U7 can be input into a conv layer with a kernel size of 3*3, a step of 1, and padding of 1, and then passes through the Upsample for up-sampling, and finally passes through a conv layer with a kernel size of 1*1 and a step of 1 to finally obtain a feature map M3 with a size of H*W*C, which extracts the contour and position information of the power transmission line and the line icing. In the M3, only the pixel point values within the contour of the power transmission line and the line icing are retained, and the pixel point values outside the contour are all 0, so that the segmentation processing of the first spliced icing image is realized.

[0058] Optionally, the position information of the retained pixel point values in the M3 can be determined as the position information of the power transmission line region and the icing region. Further, according to the common pixel range of the power transmission line region and the icing region, the region of the retained pixel point values can be divided into the power transmission line region and the icing region to respectively obtain the position information of the power transmission line region and the position information of the icing region, and the position information of the pixel point with a value of 0 in the M3 can be determined as the position information of the background region.

[0059] For example, referring to Figure 1C , the feature extraction layer contained in the multi-level segmentation module in the target detection model, that is, the CP module can contain three convolution units. Specifically, after the feature map M2 is input into the CP module, it can pass through two conv layers with a kernel size of 3*3, a step of 1, and padding of 1 (i.e., convolution units) in turn to obtain a feature map P1 with a size of H*W*2C. After the P1 and the corresponding elements of the M1 are added by an Add operation, a feature map P2 with a size of H*W*2C is obtained. After the P2 passes through a conv layer with a kernel size of 3*3, a step of 2, padding of 1, and a number of convolution kernels of 64 and a PSwish activation function, a feature map U1 with a size of H / 2*W / 2*64 can be output.

[0060] For example, the calculation formula of the PSwish activation function can be as follows:

[0061]

[0062] In the formula, x is the value of a pixel point in the input feature map, and a is a parameter adjusted through training learning. The calculation formula of sigmoid(x) is as follows:

[0063]

[0064] In the formula, x is the value of a pixel point in the input feature map, and e is an infinite non-cyclic decimal.

[0065] For example, if the size of M1 is 1024x1024x3, M1 can be spliced with itself to obtain a feature map M2 with a size of 1024x1024x6. M2 is input into MUnet, and after being processed by the CP module, a feature map U1 with a size of 512x512x64 is obtained. U1 is processed by the CP module to obtain a feature map U2 with a size of 256x256x128. U2 is processed by the CP module to obtain a feature map U3 with a size of 128x128x256. U3 is processed by the CP module to obtain a feature map U4 with a size of 64x64x512. U4 is input into the upsampling layer Upsample for upsampling, and the obtained result is subjected to an Add operation with U3, and then two conv layers with a kernel size of 3x3, a step of 1, and padding of 1 are performed to obtain a feature map U5 with a size of 128x128x256. U5 is input into the upsampling layer Upsample for upsampling, and the obtained result is subjected to an Add operation with the feature map U2, and then two conv layers with a kernel size of 3x3, a step of 1, and padding of 1 are performed to obtain a feature map U6 with a size of 256x256x128. U6 is input into the upsampling layer Upsample for upsampling, and the obtained result is subjected to an Add operation with U1 to obtain a feature map U7 with a size of 512x512x64.

[0066] For example, U7 is input into a conv layer with a kernel size of 3x3, a step of 1, and padding of 1, and then is subjected to upsampling by Upsample, and finally is subjected to a conv layer with a kernel size of 1x1 and a step of 1 to obtain a feature map M3 with a size of 1024x1024x3, which extracts the contour and position information of the power transmission line and line icing. In M3, only the pixel point values within the contour of the power transmission line and line icing are retained, and the remaining pixel point values outside the contour are all 0.

[0067] For example, M3 and M1 are spliced by concat to obtain a feature map M4 with a size of 1024x1024x6, and M4 is input into the segmentation module MUnet for processing to obtain a second segmented icing image M5 with a size of 1024x1024x3, and the image is segmented into three mutually disjoint regions of power transmission line, line icing, and background.

[0068] Optionally, a lightweight classification module in the object detection model is used to classify the second segmented icing image to obtain the corresponding icing type. This includes: determining the first classification convolutional layer, information extraction layer, second classification convolutional layer, average pooling layer, and fully connected layer included in the lightweight classification module of the object detection model; using the first classification convolutional layer to perform convolution processing on the second segmented icing image M5 to obtain a convolutional image E1, and using each information extraction layer to sequentially extract the convolutional image to obtain an extracted image; using the second classification convolutional layer, average pooling layer, and fully connected layer to classify the extracted image to obtain the corresponding icing type.

[0069] The information extraction (DS) layer consists of at least two layers. The first classification convolutional layer can be a conv layer with a kernel size of 3×3, a stride of 2, padding of 1, and 32 kernels. The second classification convolutional layer can be a conv layer with a kernel size of 1×1 and a stride of 1.

[0070] For example, see Figure 1D If the size of the second segmented icing image M5 is H×W×C, then after inputting the second segmented icing image M5 into the lightweight classification module, it will first pass through the first classification convolutional layer to obtain a feature map E1 of size H / 2×W / 2×32, i.e., the convolutional image. It can then sequentially pass through the same information extraction layers, such as three identical information extraction layers. The first information extraction layer processes the convolutional image E1 to obtain feature map E2, the second information extraction layer processes feature map E2 to obtain feature map E3, and the third feature extraction layer processes feature map E3 to obtain feature map E4, which is the extracted image. The size of feature map E3 is H / 8×W / 8×128, and the size of feature map E4 is H / 16×W / 16×256.

[0071] For example, the extracted image E4 can be input into a second classification convolutional layer, an average pooling layer, and a fully connected layer to obtain the corresponding icing type. The second classification convolutional layer and the pooling layer (i.e., the average pooling layer) perform global average pooling on the feature map output by the last DS layer, compressing the feature map of each channel into a single value, thereby obtaining a global feature vector. This step helps reduce the number of model parameters and prevent overfitting, outputting a global feature vector. The fully connected layer maps the global feature vector obtained from the second classification convolutional layer and the pooling layer to the final classification result. This is a typical classification layer, responsible for mapping high-level features to class probabilities, outputting the probability value of each class, and finally, based on the output probability values ​​of each class, the one with the highest probability value is determined as the final icing type.

[0072] For example, if the size of the second segmented icing image M5 is 1024x1024x3, the process of inputting it into the lightweight classification module MEnet for classification can be as follows: a first classification convolutional layer with a convolution kernel size of 3x3, a step size of 2, and padding of 1 is used to obtain a convolutional image E1 with a size of 512x512x32. After E1 passes through the DS module, a feature map E2 with a size of 256x256x64 is obtained. After E2 passes through the DS module, a feature map E3 with a size of 128x128x128 is obtained. After E3 passes through the DS module, a feature map E4 with a size of 64x64x256 is obtained, i.e., an extracted image. After the extracted image E4 passes through a conv layer with a convolution kernel size of 1x1 and a step size of 1, an average pooling layer Pooling, and a fully connected layer FC, the icing type of the image can be finally output.

[0073] For example, referring to Figure 1E The information extraction layer in the lightweight classification module of the target detection model, i.e., the DS module, can include a normal convolutional layer conv, a depthwise convolutional layer (Depthwise convolution, DWconv), a max pooling layer (MaxPool), and a spatial and channel reconstruction convolutional layer (Spatial and Channel Reconstruction Convolution, SCconv). Specifically, after inputting the DS module, the convolutional image E1 can first pass through a normal convolutional layer conv with a convolution kernel size of 1x1 and a step size of 1, and the channel number of E1 is increased by n times (through experimental testing, setting n=6 has the best effect, which enriches the channel feature information without increasing the calculation amount too much), to obtain a feature map D1 with a size of H / 2xW / 2x192.

[0074] Further, D1 can be evenly split into two groups according to the channel number, to obtain a feature map D2 with a size of H / 2xW / 2x96 and a feature map D3 with a size of H / 2xW / 2x96. D2 passes through a normal convolutional layer conv with a convolution kernel size of 1x1 and a step size of 1 to increase the channel number, then passes through a depthwise convolutional layer DWconv with a convolution kernel size of 3x3 and a step size of 2 for feature extraction, and finally passes through a normal convolutional layer conv with a convolution kernel size of 1x1 and a step size of 1 to reduce the channel number, to obtain a feature map D4 with a size of H / 4xW / 4x96.

[0075] Further, the D3 first passes through a 2*2 maximum pooling layer MaxPool, and then passes through a channel reconstruction convolution SCconv to obtain a feature map D5 with a size of H / 4*W / 4*96. Finally, the D4 and the D5 are subjected to an Add operation, and then pass through a normal convolution conv layer with a convolution kernel size of 1*1, a step size of 1, and a convolution kernel number of 64, to output a feature map E2 with a size of H / 4*W / 4*64.

[0076] S103, according to the icing type, determining the icing coping measure for the power transmission line, so as to process the icing on the power transmission line according to the icing coping measure.

[0077] The icing coping measure refers to a measure of deicing the icing on the power transmission line. The icing coping measure may be, for example, robot deicing, manual deicing, or measures such as increasing the load current or using a short-circuit current to raise the temperature of the conductor.

[0078] Optionally, if the icing type is normal and no icing, the corresponding icing coping measure can be not performing any operation.

[0079] Optionally, according to the icing type, the corresponding relationship between the preset icing type and the icing coping measure can be combined to determine the icing coping measure for the power transmission line, so as to process the icing on the power transmission line according to the icing coping measure.

[0080] The technical scheme of the embodiment of the application, if the icing detection request for the power transmission line is detected, the target icing image corresponding to the power transmission line and the target detection model pre-trained are determined; according to the target icing image, the multi-level segmentation module and the lightweight classification module in the target detection model are used to identify and process the target icing image, so as to obtain the corresponding icing type; according to the icing type, the icing coping measure for the power transmission line is determined, so as to process the icing on the power transmission line according to the icing coping measure. By using the target detection model to identify and process the icing image, the icing condition of the power transmission line can be remotely analyzed, and the efficiency of the icing processing can be effectively improved.

[0081] Embodiment two

[0082] Figure 2 is a flowchart of a power transmission line icing processing method provided by the embodiment two of the application; the embodiment provides an optimized scheme for icing processing based on the above-mentioned embodiments. As shown in the figure, the method comprises the following processes: Figure 2

[0083] ​Optionally, if the icing detection request of the power transmission line is detected, the outdoor power transmission line can be image collected to obtain an outdoor power transmission line image, that is, a target icing image M1, M1 is concatenated with itself to obtain a first concatenated icing image M2, M2 is input into a multi-level segmentation module MUnet for segmentation to obtain a first segmented icing image M3 in which the power transmission line embodiment and the line icing contour and position information are extracted.

[0084] Optionally, M3 is concatenated with M1 to obtain a second concatenated icing image M4, M4 is input into the multi-level segmentation module MUnet for segmentation to obtain a second segmented icing image M5, the pixel points in M5 are divided into three categories, that is, the power transmission line, the line icing and the background, and the in-image target segmentation is realized. Finally, the second segmented icing image M5 is input into a lightweight classification module MEnet for line icing classification, and the icing classification corresponding to the target icing image and the segmented image are output, so that the target icing image is classified as a normal no-icing power transmission line image, a rime icing power transmission line image, a snow icing power transmission line image, a hoar frost icing power transmission line image or a mixed icing power transmission line image.

[0085] The technical scheme of the present application can effectively and accurately segment the outdoor power transmission line image into a power transmission line, a line icing and a background, and classify the line icing into a normal no-icing power transmission line image, a rime icing power transmission line image, a snow icing power transmission line image, a hoar frost icing power transmission line image or a mixed icing power transmission line image, which helps to determine the icing coping measures. At the same time, the target detection model designed in the present application has fewer operation parameters, higher operation efficiency and more accurate results.

[0086] Embodiment three

[0087] Figure 3 is a structural block diagram of a power transmission line icing processing device provided by an embodiment of the present application; the present embodiment can be applicable to the case that the icing image of the power transmission line is recognized based on a target detection model to obtain the icing type so as to help to take the corresponding strategy for processing subsequently, the power transmission line icing processing device provided by the embodiment of the present application can execute the power transmission line icing processing method provided by any embodiment of the present application, and has the function modules and beneficial effects corresponding to the execution method; the power transmission line icing processing device can be realized in the form of hardware and / or software, and is configured in an electronic device with a power transmission line icing processing function, such as an electronic device with a power transmission line icing processing function, which is executed by a general control system in a power distribution network, as shown in the figure, the power transmission line icing processing device specifically comprises: Figure 3

[0088] ​The determining module 301 is configured to determine a target icing image corresponding to the power transmission line and a pre-trained target detection model if the icing detection request for the power transmission line is detected.

[0089] The identifying module 302 is configured to perform identification processing on the target icing image by using a multi-level segmentation module and a lightweight classification module in the target detection model according to the target icing image, so as to obtain a corresponding icing type.

[0090] The processing module 303 is configured to determine an icing countermeasure for the power transmission line according to the icing type, and perform processing on the icing on the power transmission line according to the icing countermeasure.

[0091] The technical scheme of the embodiment of the application can determine a target icing image corresponding to the power transmission line and a pre-trained target detection model if the icing detection request for the power transmission line is detected, perform identification processing on the target icing image by using a multi-level segmentation module and a lightweight classification module in the target detection model according to the target icing image, so as to obtain a corresponding icing type, and determine an icing countermeasure for the power transmission line according to the icing type, and perform processing on the icing on the power transmission line according to the icing countermeasure. The icing image is identified by using the target detection model, so that the icing condition of the power transmission line can be remotely analyzed, and the efficiency of the icing processing can be effectively improved.

[0092] Further, the identifying module 302 can include:

[0093] The first splicing unit is configured to perform self-splicing processing on the target icing image to obtain a first spliced icing image.

[0094] The second splicing unit is configured to perform segmentation processing on the first spliced icing image by using a multi-level segmentation module in the target detection model to obtain a first segmented icing image, and perform splicing processing on the first segmented icing image and the target icing image to obtain a second spliced icing image.

[0095] The classification unit is configured to perform segmentation processing on the second spliced icing image by using a multi-level segmentation module in the target detection model to obtain a second segmented icing image, and perform classification processing on the second segmented icing image by using a lightweight classification module in the target detection model to obtain a corresponding icing type.

[0096] Further, the second splicing unit can include:

[0097] The determining subunit is configured to determine a feature extraction layer, a convolution layer and an up-sampling layer included in the multi-level segmentation module in the target detection model, wherein the number of the feature extraction layers is at least two.

[0098] The splicing subunit is configured to: adopt the feature extraction layers, the convolution layers, and the up-sampling layers to perform hierarchical extraction processing, convolution processing, and up-sampling processing on the first splicing icing image, so as to segment the position information of the power line region, the icing region, and the background region in the first splicing icing image, and mark in the first splicing icing image according to the position, so as to obtain the first segmented icing image.

[0099] Further, the splicing subunit is specifically configured to:

[0100] adopt the first feature extraction layer, the second feature extraction layer, the third feature extraction layer, and the fourth feature extraction layer to perform extraction processing on the first splicing icing image respectively, so as to obtain the first feature, the second feature, the third feature, and the fourth feature;

[0101] perform up-sampling processing on the fourth feature, and perform addition processing and convolution processing on the fourth feature after the up-sampling processing and the third feature, so as to obtain the fifth feature;

[0102] perform up-sampling processing on the fifth feature, and perform addition processing and convolution processing on the fifth feature after the up-sampling processing and the second feature, so as to obtain the sixth feature;

[0103] perform up-sampling processing on the sixth feature, and perform addition processing on the sixth feature after the up-sampling processing and the first feature, so as to obtain the seventh feature;

[0104] perform convolution processing, up-sampling processing, and twice convolution processing on the seventh feature in sequence, so as to obtain the position information of the power line region, the icing region, and the background region in the first splicing icing image, and mark in the first splicing icing image according to the position, so as to obtain the first segmented icing image.

[0105] Further, the classification unit is specifically configured to:

[0106] determine that the target detection model includes a first classification convolution layer, an information extraction layer, a second classification convolution layer, an average pooling layer, and a full connection layer in a lightweight classification module; and the number of the information extraction layers is at least two;

[0107] adopt the first classification convolution layer to perform convolution processing on the second segmented icing image, so as to obtain a convolution image, and adopt the information extraction layers to perform extraction processing on the convolution image in sequence, so as to obtain an extraction image;

[0108] adopt the second classification convolution layer, the average pooling layer, and the full connection layer to perform classification processing on the extraction image, so as to obtain a corresponding icing type.

[0109] Further, the above device is further configured to:

[0110] construct a training sample set according to the historical icing image and the corresponding icing type.

[0111] According to the training sample set, the initial detection model is trained based on a cross-entropy loss function and a negative log-likelihood loss function to obtain the target detection model.

[0112] Embodiment Four

[0113] Figure 4 is a structural schematic diagram of an electronic device provided by Embodiment Four of the present application. Figure 4 A structural schematic diagram of an electronic device 10 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present application described and / or claimed in this document.

[0114] As shown in Figure 4 The electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11, wherein the memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0115] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, a loudspeaker, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0116] The processor 11 can be various general and / or special purpose processing components having processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the power line icing treatment method.

[0117] In some embodiments, the power line icing treatment method can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded onto the RAM 13 and executed by the processor 11, one or more steps of the power line icing treatment method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the power line icing treatment method by any other suitable means, such as by means of firmware.

[0118] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0119] Computer programs used to implement the methods of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program, when executed by the processor of the machine, implements the functions / acts specified in the flowcharts and / or block diagrams. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0120] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer-readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of a machine-readable storage medium will include one or more lines of a program of instructions in a transitory signal, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0121] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0122] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0123] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. Servers can be cloud servers, also known as cloud computing servers or cloud hosts, which are a host product in the cloud computing service system to solve the defects of large management difficulty and weak business scalability in traditional physical hosts and VPS services.

[0124] In an embodiment, the present embodiment further includes a computer program product comprising a computer program which, when executed by a processor, implements the power transmission line icing treatment method of any of the embodiments of the present application.

[0125] The computer program product can be written in any of various programming languages, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0126] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the spirit and scope of the present application. For example, the steps recited in the present application can be performed in parallel, in series, or in a different order, and the present application is not limited in this regard.

[0127] The specific embodiments described above are not intended to limit the scope of the present application. Those skilled in the art will understand that various modifications, combinations, sub-combinations, and alternatives can be made to the specific embodiments without departing from the spirit and principles of the present application. Any further modifications, equivalents, and / or alternatives come within the scope of the present application as recited by the claims.

Claims

1. A method for ice accretion management on a power transmission line, characterized in that, The method comprises the following steps: If an icing detection request for the power transmission line is detected, a target icing image corresponding to the power transmission line and a pre-trained target detection model are determined; According to the target icing image, a multi-level segmentation module and a lightweight classification module in the target detection model are used to perform identification processing on the target icing image to obtain a corresponding icing type; the icing type includes normal no icing, rime type, snow type, glaze type, and mixed type; the mixed type refers to an icing type containing at least two types of rime type, snow type, and glaze type; According to the icing type, an icing countermeasure for the power transmission line is determined to process the icing on the power transmission line according to the icing countermeasure; the icing countermeasure is a robot deicing, manual deicing, increasing load current, or using short-circuit current to increase the temperature of the conductor; According to the target icing image, a multi-level segmentation module and a lightweight classification module in the target detection model are used to perform identification processing on the target icing image to obtain a corresponding icing type, comprising: The target icing image is concatenated with itself, that is, self-splicing processing is performed to obtain a first spliced icing image; the concat splicing operation is to combine the channel numbers of two feature maps, and the size of the obtained feature map remains unchanged, but the channel number increases; The multi-level segmentation module in the target detection model is used to perform segmentation processing on the first spliced icing image to obtain a first segmented icing image, and the first segmented icing image is spliced with the target icing image to obtain a second spliced icing image; The multi-level segmentation module in the target detection model is used to perform segmentation processing on the second spliced icing image to obtain a second segmented icing image, and the lightweight classification module in the target detection model is used to perform classification processing on the second segmented icing image to obtain a corresponding icing type; According to the target icing image, a multi-level segmentation module and a lightweight classification module in the target detection model are used to perform identification processing on the target icing image to obtain a corresponding icing type, comprising: The number of feature extraction layers is at least two; each feature extraction layer, convolution layer, and up-sampling layer is used to perform extraction processing, convolution processing, and up-sampling processing on the first spliced icing image in a hierarchical manner to segment the position information of the power transmission line region, the icing region, and the background region in the first spliced icing image, and mark in the first spliced icing image according to the position information to obtain the first segmented icing image; The position information of the power transmission line region, the icing region and the background region in the first spliced icing image is obtained by segmentation, including: the position information of the region with reserved pixel point values in the first spliced icing image is determined as the position information of the power transmission line region and the icing region; the region with reserved pixel point values is divided into the power transmission line region and the icing region according to the common pixel range of the power transmission line region and the icing region, so as to obtain the position information of the power transmission line region and the position information of the icing region respectively; and the position information of the region with pixel points of 0 in the first spliced icing image is determined as the position information of the background region. The corresponding icing type is obtained by classifying the second segmented icing image by using a lightweight classification module in the target detection model, including: The first classification convolution layer, the information extraction layer, the second classification convolution layer, the average pooling layer and the full connection layer contained in the lightweight classification module in the target detection model are determined; the number of information extraction layers is at least two; the information extraction layer includes an ordinary convolution layer, a depth separable convolution layer, a maximum pooling layer and a channel reconstruction convolution layer; The first classification convolution layer is used to perform convolution processing on the second segmented icing image to obtain a convolution image, and each information extraction layer is used to perform extraction processing on the convolution image in turn to obtain an extraction image; The second classification convolution layer, the average pooling layer and the full connection layer are used to perform classification processing on the extraction image to obtain the corresponding icing type.

2. The method of claim 1, wherein, The first segmented icing image is obtained, including: The first feature, the second feature, the third feature and the fourth feature are obtained by using the first feature extraction layer, the second feature extraction layer, the third feature extraction layer and the fourth feature extraction layer to perform extraction processing on the first spliced icing image respectively; The fourth feature is up-sampled, and the up-sampled fourth feature and the third feature are added and convoluted to obtain a fifth feature; The fifth feature is up-sampled, and the up-sampled fifth feature and the second feature are added and convoluted to obtain a sixth feature; The sixth feature is up-sampled, and the up-sampled sixth feature and the first feature are added to obtain a seventh feature; The seventh feature is convoluted, up-sampled and twice convoluted in turn to obtain the position information of the power transmission line region, the icing region and the background region in the first spliced icing image, and the first segmented icing image is labeled in the first spliced icing image according to the position information.

3. The method of claim 1, wherein, Before the pre-trained target detection model is determined, it further includes: A training sample set is constructed according to historical icing images and corresponding icing types; An initial detection model is trained based on a cross-entropy loss function and a negative log-likelihood loss function to obtain the target detection model according to the training sample set.

4. An apparatus for de-icing power lines, characterized in that It includes: The determination module is configured to determine the target icing image corresponding to the power transmission line and the pre-trained target detection model if the icing detection request for the power transmission line is detected. The identification module is configured to perform identification processing on the target icing image by using a multi-level segmentation module and a lightweight classification module in the target detection model according to the target icing image, so as to obtain a corresponding icing type; the icing type includes normal no icing, rime type, snow type, glaze type, and mixed type; the mixed type refers to an icing type containing at least two types of the rime type, the snow type, and the glaze type; The processing module is configured to determine an icing countermeasure for the power transmission line according to the icing type, so as to process the icing on the power transmission line according to the icing countermeasure; The icing countermeasure is a measure of robot deicing, manual deicing, increasing load current, or increasing conductor temperature by short-circuit current; The identification module includes: a first splicing unit configured to perform concat splicing operation on the target icing image and itself, that is, to perform self-splicing processing, so as to obtain a first spliced icing image; the concat splicing operation is to merge the channel numbers of two feature maps, so that the size of the obtained feature map remains unchanged, but the channel number increases; a second splicing unit configured to perform segmentation processing on the first spliced icing image by using the multi-level segmentation module in the target detection model, so as to obtain a first segmented icing image, and perform splicing processing on the first segmented icing image and the target icing image, so as to obtain a second spliced icing image; and a classification unit configured to perform segmentation processing on the second spliced icing image by using the multi-level segmentation module in the target detection model, so as to obtain a second segmented icing image, and perform classification processing on the second segmented icing image by using the lightweight classification module in the target detection model, so as to obtain the corresponding icing type; The second splicing unit includes: a determination subunit configured to determine a feature extraction layer, a convolution layer, and an up-sampling layer included in the multi-level segmentation module in the target detection model; the number of the feature extraction layers is at least two; and a splicing subunit configured to perform extraction processing, convolution processing, and up-sampling processing on the first spliced icing image in layers by using the feature extraction layers, the convolution layer, and the up-sampling layer, so as to segment and obtain position information of a power transmission line region, an icing region, and a background region in the first spliced icing image, and mark in the first spliced icing image according to the positions, so as to obtain the first segmented icing image; The splicing subunit is further configured to: determine the position information of the power transmission line region and the icing region by using the position information of the first spliced icing image in which pixel point values are retained; divide the region in which the pixel point values are retained into the power transmission line region and the icing region according to common pixel ranges of the power transmission line region and the icing region, so as to obtain position information of the power transmission line region and position information of the icing region, respectively; and determine the position information of the background region by using the position information of a region in which pixel points are 0 in the first spliced icing image. The classification unit is specifically configured to determine that the target detection model comprises a first classification convolutional layer, an information extraction layer, a second classification convolutional layer, an average pooling layer, and a full connection layer in a lightweight classification module; the number of the information extraction layer is at least two; the information extraction layer comprises a normal convolutional layer, a depth separable convolutional layer, a maximum pooling layer, and a channel reconstruction convolutional layer; the first classification convolutional layer is used to perform convolution processing on the second segmented icing image to obtain a convolution image, and each information extraction layer is used to perform extraction processing on the convolution image in sequence to obtain an extraction image; the second classification convolutional layer, the average pooling layer, and the full connection layer are used to perform classification processing on the extraction image to obtain a corresponding icing type.

5. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores a computer program executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the power line icing processing method in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for causing the processor to execute the power line icing processing method in any one of claims 1-3 when executed.

7. A computer program product, characterised in that, The computer program product includes a computer program that, when executed by a processor, implements the power line icing processing method according to any one of claims 1-3.

Images