Conducting wire ice coating body measuring method

Abstract

The invention relates to the technical field of power transmission, and provides a wire icing body measurement method, which comprises the following steps of: obtaining wire icing body images including a color image, an infrared image and a blue light image; performing edge identification on the lead ice coating body image by adopting an HED algorithm to obtain edge probability graphs corresponding to the color image, the infrared image and the blue light image; fusing the marginal probability graphs of the color image, the infrared image and the blue light image based on the dynamic weight to obtain a fused marginal probability graph; and on the basis of the fused marginal probability graph, obtaining lead ice coating body measurement data. According to the method, the multi-source image is combined with an image processing algorithm, so that the measurement efficiency is improved, and personal errors and personnel safety risks in a traditional method are also avoided.

Description

Technical Field

[0001] The present invention relates to the field of power transmission technology, and in particular to a method for measuring ice coating on a conductor. Background Art

[0002] During the design, operation, and maintenance of power transmission lines, it is crucial to understand the size of ice covering the conductors. Currently, there are two common methods for measuring ice size in the industry: one is to measure the weight of the ice, and the other is to measure the geometric dimensions and density of the ice covering. In the past, manual measurement was generally used to measure the geometric dimensions of conductor ice coverings, with surveyors using rulers or calipers to measure the major and minor diameters of the ice covering. Manual measurement typically requires contact with the object being measured, which is not only inefficient and susceptible to human error, but also can pose safety risks to surveyors in certain situations during ice measurement due to slippery road conditions. In recent years, China has developed a number of devices that use image capture for measurement. These devices primarily use cameras to capture images, then process and calculate the images to obtain the size of the ice. This type of equipment replaces manual observation, improves observation efficiency, and avoids the safety risks of manual observation.

[0003] However, such equipment has obvious shortcomings, resulting in large errors, which are mainly reflected in the following two aspects: 1. Only suitable for measuring ice with a circular cross section. This type of equipment usually uses a single camera to capture images. The camera is generally located at the same height as the conductor or higher. The images it captures are single-view images. It is only suitable for measuring ice with a circular cross section and can measure the diameter of the conductor ice (not the major and minor diameters). Figure 1 shown.

[0004] However, when the ice-covered body is non-circular, it is easy to cause large errors. The cross-section of the ice-covered body formed in nature is often irregular, such as Figure 2a 、 Figure 2b shown.

[0005] As can be seen, in this case, due to the perspective, the ice diameter obtained is essentially the short diameter of the ice, and the long diameter of the ice cannot be obtained. The obtained ice diameter in this case will lead to large errors. Due to the irregularity of ice, only by simultaneously obtaining the long and short diameters of the ice can the ice size be calculated relatively accurately.

[0006] Second, the identified edge lines are broken, discontinuous, or excessive, including some non-edge lines. This can be caused by a variety of factors, including irregular ice bodies, complex textures, noise interference, and improper parameter settings. These issues can lead to significant errors in the measurement results. Therefore, further optimization algorithms are needed to reduce these errors and improve accuracy. Summary of the Invention

[0007] In view of the above problems, the present invention aims to provide a method for measuring ice coating on a conductor.

[0008] In a first aspect, the present invention provides a method for measuring ice coating on a conductor, comprising: Acquire images of ice covering the conductor, including color images, infrared images and blue light images; The HED algorithm is used to perform edge recognition on the wire ice-covered body image to obtain edge probability maps corresponding to the color image, infrared image and blue light image; The edge probability maps of the color image, the infrared image and the blue light image are fused based on the dynamic weight to obtain a fused edge probability map; Based on the fused edge probability map, wire ice-covered body measurement data is obtained.

[0009] In some embodiments, two groups of cameras are used to capture images of ice covering the wires, and each group of cameras includes a color camera, an infrared camera, and a blue light camera; wherein, one group of cameras captures images of ice covering the wires horizontally, and the other group of cameras captures images of ice covering the wires vertically; each group of cameras is equipped with one or more fill lights, and the fill lights are connected to light intensity sensors; the light intensity sensor is used to detect the light intensity of the current environment of the ice covering the wires, and when the detected light intensity is lower than a light intensity threshold, the fill light is turned on.

[0010] In some embodiments, the HED algorithm is an improved HED algorithm; The improved HED algorithm refers to the HED algorithm that adopts a layer-by-layer cumulative evaluation strategy to perform fusion calculation on the side output layer.

[0011] In some embodiments, the step of performing fusion calculation on the side output layer using a layer-by-layer cumulative evaluation strategy includes: When obtaining the i After the side output layer, i >1, immediately fuse the generated side output layer 1 to the side output layer i The result is then binarized to form edges, and the edge density of the current fusion image is calculated according to the following formula: :

[0012] in, N is the number of valid edge pixels in the fusion image, W × H is the total number of pixels of the fusion image; Then, the fusion calculation is terminated early according to the following judgment conditions: i =3, if , directly stop the fusion calculation of side output layer 4 and side output layer 5; when It is necessary to perform the fusion calculation of the side output layer 4. After completing the fusion calculation of the side output layer 4, if Stop the fusion calculation of the side output layer 5, otherwise the fusion calculation of the side output layer 5 is also required; when It is determined that the fusion calculation of the side output layer 4 and the side output layer 5 needs to be performed.

[0013] In some embodiments, fusing the edge probability maps of the color image, the infrared image, and the blue light image based on the dynamic weight to obtain the fused edge probability map includes: Calculating the haze index F ; Based on the haze index F and light intensity, calculate the basic weights of color images, infrared images and blue light images; Adjust the basic weight of each image according to the edge density of each image; Calculate dynamic weights based on the adjusted basic weights of each image; The edge probability maps of the color image, the infrared image, and the blue light image are weightedly calculated with the corresponding dynamic weights to obtain a fused edge probability map.

[0014] In some embodiments, the calculated haze index F ,include: Normalize the single channel value V of the image to the range [0,1]; Divide the image into N × N blocks, calculate the average brightness of each block; select the front block with the highest brightness K blocks, and take the average brightness as the atmospheric light estimation A ; Based on atmospheric light estimation A , find the global brightness deviation GBD :

[0015] in, W is the width of the image, H is the height of the image; Calculate the local contrast attenuation factor LAC: Divide the image into M × M Block, for each block, calculate the contrast according to the formula C i :

[0016] Then calculate the contrast mean of all blocks , then the local contrast attenuation factor LAC for:

[0017] Based on global brightness bias GBD and the local contrast attenuation factor LAC , calculate the haze index F for:

[0018] in, α It is an empirical coefficient that balances the contribution of brightness and contrast.

[0019] In some embodiments, the haze index F and light intensity, calculate the basic weights of color images, infrared images, and blue light images, including: The base weight W of the color image c_base for:

[0020] The basic weight W of the infrared image i_base for:

[0021] The base weight W of the blue light image b_base for:

[0022] in:

[0023] Where, L is the normalized light intensity, L n The light intensity currently obtained by the light intensity sensor.

[0024] In some embodiments, adjusting the base weight of each image according to the edge density of each image includes:

[0025]

[0026]

[0027] in, Represents the basic weight of the adjusted color image, represents the edge density of the color image, Represents the adjusted infrared image basic weight, represents the edge density of infrared image, Represents the adjusted blue image base weight, Indicates the edge density of the blue image.

[0028] In some embodiments, the calculating of the dynamic weight based on the adjusted basic weights of each image includes:

[0029]

[0030]

[0031] Where, W c is the dynamic weight of the color image, W i is the dynamic weight of the infrared image, W b is the dynamic weight of the blue image.

[0032] In some embodiments, it is necessary to exclude the interference of the edge probability map of the wire-covered ice body on edge recognition from the edge probability map output by the HED neural network. Specifically: for the edge probability map output by the HED neural network, the position of each point in the edge probability map is recorded in the form of coordinates, thereby recording the probability and corresponding position of each point on the edge line; then, high-probability points are selected and the corresponding fitting straight lines are obtained by the least squares method, so that there is a fitting straight line on each side of the wire; similarly, for the edge of the wire-covered ice body, the least squares method can also be used to obtain the fitting straight lines of the two edge lines; when it is detected that the angle between the fitting straight line corresponding to the edge of the wire-covered ice body and the edge line of the wire ice exceeds a preset angle threshold, the edge lines corresponding to the two fitting straight lines are directly excluded.

[0033] In a second aspect, the present invention provides an electronic device, comprising: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the above method by executing the instructions stored in the memory.

[0034] In a third aspect, the present invention provides a computer-readable storage medium for storing instructions, which implement the above method when the instructions are executed.

[0035] In a fourth aspect, the present invention provides a computer program product, which, when called by a computer, enables the computer to execute the above method.

[0036] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. Compared with the existing single-view solution, the present invention can not only accurately measure the ice covering of circular-section conductors, but also accurately measure the ice covering of irregular-section conductors, thus avoiding the large errors caused by view angle obstruction in the existing solution.

[0037] 2. This invention leverages the HED algorithm's multi-scale edge detection and global and local information fusion capabilities to simultaneously capture edge information at different scales. It also effectively combines global and local image information, resulting in superior performance when processing complex backgrounds and multi-scale edges, significantly improving edge detection accuracy and robustness. Compared to traditional edge detection algorithms, which often suffer from edge line breakage, discontinuity, and misidentification, the HED algorithm more accurately captures edge information, thereby improving measurement accuracy.

[0038] 3. The present invention improves the HED algorithm and adopts an iterative optimization mechanism based on edge density feedback, which reduces the computational complexity of the HED algorithm, improves efficiency while ensuring accuracy, and takes into account both performance and precision.

[0039] 4. This invention constructs a mathematical model to dynamically adjust the weights of three image types (color, infrared, and blue light). This model uses light intensity, haze index, and edge density as independent variables to achieve optimal ice edge recognition under different conditions. Features of this model include: an optimized normalized light intensity method that balances the effects of varying light intensity and ensures a uniform distribution within the normalized range of varying light intensities; strong adaptability under diverse environmental conditions; an edge density feedback mechanism that automatically prioritizes image sources with clearer edges; smooth, non-aggressive weighting changes across different conditions; and the preservation of the unique advantages of different image sources under specific conditions.

[0040] 5. The present invention eliminates edge recognition interference caused by ice covering non-conducting wires such as guy wires based on on-site conditions, thereby improving the accuracy of edge recognition.

[0041] 6. The present invention can measure all elements of conductor ice coating (major diameter, minor diameter, density, and ice weight), fully meeting various needs of power ice coating observation and having comprehensive advantages over single-element observation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a method for measuring ice coating on a conductor proposed in an embodiment of the present invention.

[0043] Figure 2a This is a front view schematic diagram of the camera installation in an embodiment of the present invention.

[0044] Figure 2b This is a side view of the camera installation in an embodiment of the present invention.

[0045] Figure 3 Schematic diagram of the relative positions of the camera and the wire ice covering in an embodiment of the present invention.

[0046] Figure 4 Schematic diagram of the HED neural network structure.

[0047] Figure 5 This is a schematic structural diagram of an electronic device proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0049] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0050] This embodiment of the present invention proposes a method for measuring conductor ice cover. By using two cameras at different positions and viewing angles to capture ice on a conductor, combined with advanced image processing algorithms, the long and short diameters of the conductor ice can be accurately and automatically determined. This method not only improves measurement efficiency but also avoids the human error and safety risks associated with traditional methods.

[0051] like Figure 1 As shown, the method for measuring the ice-covered conductor includes the following steps: S100, acquiring an image of the ice covering the conductor, including a color image, an infrared image, and a blue light image; S200, using a Holistically-Nested Edge Detection (HED) algorithm to perform edge recognition on the wire ice-covered body image to obtain edge probability maps corresponding to the color image, the infrared image, and the blue light image; S300, fusing the edge probability maps of the color image, the infrared image, and the blue light image based on the dynamic weight to obtain a fused edge probability map; S400: Obtaining conductor ice-covered body measurement data based on the fused edge probability map.

[0052] In step S100, two groups of cameras are used to capture images of ice covering the conductors. Each group of cameras includes a color camera, an infrared camera, and a blue light camera (a camera with a blue light filter). One group of cameras captures images of ice covering the conductors horizontally, while the other group of cameras captures images of ice covering the conductors vertically. Figure 2a 、 Figure 2b As shown, in specific implementation, a vertical pole can be installed near the middle of the conductor, a horizontal beam can be installed on the vertical pole, and a set of cameras can be installed at appropriate positions of the vertical pole and the horizontal beam to complete the horizontal and vertical shooting of the conductor ice-covered body image and ensure the consistency of the shooting angle and distance. The relative position of the camera and the conductor ice-covered body is as shown in FIG. Figure 3 As shown, this separate measurement and capture method overcomes the large errors caused by the limitations of a single viewing angle. Furthermore, to improve image processing and detection, the captured image of the ice-covered conductor must ensure that the conductor is located in the center of the image and parallel to the long side of the image.

[0053] In some embodiments, each camera group is equipped with one or more fill lights (e.g., blue LED fill lights), and the fill lights are connected to a light intensity sensor. The light intensity sensor is used to detect the light intensity of the current environment of the wire ice-covered body. When the detected light intensity is lower than the light intensity threshold, it indicates that the light intensity is low, the environment is dim, or the fog is thick. The fill lights are turned on to illuminate the wire ice-covered body in a moderately small range to enhance the effect of edge recognition.

[0054] In step S100, to improve edge recognition accuracy, the captured wire ice-covered images require image preprocessing. Specifically, the images captured by each camera group are first aligned by removing pixels along the edges of the images according to a preset threshold. This ensures accuracy in subsequent processing. The images are then subjected to bilateral filtering and median filtering to reduce noise and enhance image clarity.

[0055] In step S200, the HED algorithm is as follows: The HED algorithm is an edge detection method based on deep learning. Its main advantage is that it can simultaneously process the global and local information of the image, thus performing well in complex background and multi-scale edge detection.

[0056] The core concept of the HED algorithm is to capture multi-scale edge information in images using multi-scale convolutional neural networks (CNNs). In its standard implementation, the HED algorithm constructs a multi-scale edge feature extraction network using multiple convolutional and pooling layers, each corresponding to a feature map at a different scale. These feature maps are fused through side output layers to generate a global edge detection result. The side output layer integrates edge information at different scales through a weighted fusion method to obtain a global edge detection result. The weights of the side output layer are optimized through deep neural network training to improve edge detection accuracy. The HED algorithm is trained end-to-end, utilizing a large number of previously annotated wire ice edge images for supervised learning to optimize network parameters and enhance edge detection accuracy.

[0057] The HED algorithm uses the VGG16 architecture and has been optimized. Its network structure features five convolutional layers, three fully connected layers, and a softmax output layer. Max-pooling is used between layers, and all hidden layer activation units use the ReLU function. Details are as follows: Convolution-convolution-pooling-convolution-convolution-pooling-convolution-convolution-pooling-convolution-convolution-convolution-pooling-convolution-convolution-convolution-pooling-convolution-convolution-pooling-convolution-convolution-pooling-fully connected-fully connected-fully connected; In the optimization based on the above structure, the side output layer is connected to the last convolution layer in each level, namely conv1_2, conv2_2, conv3_3, conv4_3, and conv5_3. The size of the perception interval is the same as that of each of these convolutional layers and the corresponding side output layer. At the same time, the last stage of the VGG16 architecture is subtracted, including the fifth pooling layer and all fully connected layers. Finally, the five side output layers are passed through the fusion layer to obtain the final result. In principle, the five side outputs drawn out after the convolution layer and the results of the five side outputs after the fusion layer are trained simultaneously, and six loss functions are calculated. Through multiple side outputs, multi-scale learning is performed, and features are extracted for intermediate details. For each side output, upsampling is performed through the bilinear interpolation algorithm, restored to the size of the original image, and the result is output through the fusion layer. The network structure of the HED neural network thus obtained is as follows Figure 4 shown.

[0058] According to the network structure of the HED neural network designed above, the HED neural network performs side output and fusion output, so the corresponding loss value calculation also comes from these two parts. For the five side outputs, similar to image segmentation, binary cross entropy (BCE) is considered for calculation. The total number of edge pixels and non-edge pixels in the data set are counted separately. For pixels predicted to be edge, their loss value is multiplied by the ratio of one non-edge pixel to the total pixels. Similarly, for pixels predicted to be non-edge, the loss value is also multiplied by the ratio of one edge pixel to the total pixels (weight β ) to achieve class balance adjustment. That is:

[0059] For the fused output, the cross entropy is used directly to calculate the loss value. That is:

[0060] Finally, the loss values ​​of the five side outputs are multiplied by a weight , and the fused output loss value are added to get the final loss value. That is:

[0061]

[0062] By optimizing the minimum value of the above objective loss function, the required network model can be trained.

[0063] The HED neural network is trained using a historical collection of images of conductor ice (the amount of data collected depends on the needs, typically tens of thousands). Images of varying ice types, sizes, weather, and environmental conditions are selected. In this example, 4,000 images are selected, and edge annotation is performed on each. Data augmentation techniques are then used to process these images, including noise reduction, blurring, and rotation, expanding the number of images to over 9,000. Some representative images are also grayscaled to simulate infrared images, and specially processed to blue to simulate blue light images. These images are then used to train the HED neural network, significantly enhancing the network's edge recognition capabilities under various conditions. The trained HED neural network is then used to perform edge recognition on the conductor ice images captured in step S100, thereby obtaining an edge probability map of the actual conductor ice on site.

[0064] The above are the steps used in the standard HED algorithm. In the present invention, in order to reduce the amount of calculation, the standard HED algorithm is improved to improve the calculation efficiency while taking into account the accuracy.

[0065] Due to the asynchronous computational nature of the side output layers in the HED algorithm, the five side output layers are generated incrementally during the forward propagation of the backbone network. Each side output layer relies solely on the output of its preceding convolutional layer. This chained structure allows for real-time access to intermediate results during inference. Therefore, the improved HED algorithm employs a layer-by-layer cumulative evaluation strategy to perform fused computations on the side output layers. Specifically: When obtaining the i After the side output layer ( i >1), immediately fuse the generated side output layer 1 to the side output layer i Results (the standard HED algorithm is to fuse and calculate the five side output layers at once after all five side output layers are calculated), then binarize the fused graph to form edges, and then calculate the edge density of the current fused graph according to the following formula :

[0066] in, N is the number of valid edge pixels in the fusion image, W × H is the total number of pixels in the fused image.

[0067] Then, the fusion calculation is terminated early according to the following judgment conditions.

[0068] when i =3, if , directly stop the fusion calculation of side output layer 4 and side output layer 5; when It is necessary to perform the fusion calculation of the side output layer 4. After the fusion calculation of the side output layer 4 is completed (i.e. i =4), if Stop the fusion calculation of the side output layer 5, otherwise the fusion calculation of the side output layer 5 is also required; when It is determined that the fusion calculation of the side output layer 4 and the side output layer 5 needs to be performed. It should be noted that the comparison value of the fusion density in the judgment condition is a preferred value in this embodiment and can also be set to other values ​​as needed.

[0069] Based on the above judgment conditions, it is possible to dynamically determine whether to calculate subsequent layers based on the results of the upper-side output layer, thereby reducing the amount of calculation on demand and effectively maintaining accuracy.

[0070] Since the HED algorithm outputs edge probability maps for color, infrared, and blue-light images, the effectiveness of the edge probability maps corresponding to each image varies under different conditions. Therefore, in step S300, the three edge probability maps are automatically weighted based on dynamic weights, thereby automatically maintaining the clearest and most effective edges in the final output. The dynamic weight is a function based on light intensity, haze index, and edge density, and is calculated according to the following steps: S310, calculate haze index F (0< F <1, 0 means no fog, 1 means dense fog) The haze index is an indicator that uses image features to measure the concentration of fog. It is calculated using the following method.

[0071] S311, normalize the single channel value V of the image to the range of [0,1]; S312, divide the image into N × N block (set as needed, such as 16×16 blocks), calculate the average brightness of each block; select the front block with the highest brightness K Blocks (set as needed, such as K =6), and take the mean of their brightness as the atmospheric light estimation A .

[0072] S313, based on atmospheric light estimation A , find the global brightness deviation GBD :

[0073] in, W is the width of the image, H is the height of the image. Global brightness deviation GBD The smaller the value, the more concentrated the brightness (the thicker the fog). The value range is [0,1].

[0074] S314, calculate the local contrast attenuation factor LAC: Divide the image into M × M Block (set as needed, such as 16×16 block), calculate the contrast of each block according to the following formula C i :

[0075] Then calculate the contrast mean of all blocks , then the local contrast attenuation factor LAC for:

[0076] The larger the local contrast attenuation factor LAC is, the thicker the fog is, and its value range is [0,1].

[0077] S315, based on global brightness deviation GBD and the local contrast attenuation factor LAC , calculate the haze index F for:

[0078] Where, α is an empirical coefficient that balances the contribution of brightness and contrast. In this embodiment, the value α =0.69.

[0079] S320, based on haze index F and light intensity, and calculate the basic weights of color images, infrared images, and blue light images.

[0080] The base weight W of the color image c_base for:

[0081] The basic weight W of the infrared image i_base for:

[0082] The base weight W of the blue light image b_base for:

[0083] in, L Indicates the light intensity, normalized to [0,1], where 0 means very dim and 1 means very bright. L The value of can be obtained based on the light intensity sensor. Because the light intensity in mountainous areas is unevenly distributed in the range of dim (typically tens to hundreds) to normal (typically close to a thousand to a few thousand) to bright (typically close to ten thousand to tens of thousands), it is normalized as follows.

[0084]

[0085] Where, L is the normalized light intensity, L n The light intensity currently obtained by the light intensity sensor.

[0086] S330, adjusting the basic weight of each image according to the edge density of each image:

[0087]

[0088]

[0089] in, Represents the basic weight of the adjusted color image, represents the edge density of the color image, Represents the adjusted infrared image basic weight, represents the edge density of infrared image, Represents the adjusted blue image base weight, Indicates the edge density of the blue image.

[0090] S340, calculating the dynamic weight based on the adjusted basic weight of each image:

[0091]

[0092]

[0093] Where, W c is the dynamic weight of the color image, W i is the dynamic weight of the infrared image, W b is the dynamic weight of the blue image.

[0094] The basic weights calculated in this way ensure that blue light and infrared images are prioritized in dim, foggy conditions; color images are prioritized in moderately bright, fog-free conditions; and infrared images are prioritized in bright, direct sunlight. Adjusting the basic weights further enhances edge recognition based on edge density. Calculating dynamic weights in this way allows for better adaptation to varying external conditions, ensuring clear and accurate edges.

[0095] S350: Perform weighted calculation on the edge probability maps of the color image, infrared image, and blue light image and the corresponding dynamic weights to obtain a fused edge probability map, which is expressed as:

[0096] Where, is the fused edge probability map, E c is the edge probability map of the color image, E i is the edge probability map of the infrared image, E b is the edge probability map of the blue image.

[0097] Further binarization of the fused edge probability map allows identification of the edges of the ice covering the conductor. This method allows for clear and accurate edge detection under varying lighting intensity, haze, and other environmental conditions, ensuring accurate edge distance measurement. The resulting edge lines are distributed along both sides of the conductor, and the overall orientation of the edge points is roughly parallel to the conductor.

[0098] In some embodiments, for a vertical camera, the image is taken from directly above the wire, so in addition to capturing the ice covering the wire, the ice covering on the poles and beams and on the ground will also be captured. When the lighting conditions are good, the edge lines are obvious, so the edge lines obtained by the HED algorithm are relatively accurate. However, when the lighting conditions are poor, since the wire is relatively thin, the two edge lines of the wire ice covering are relatively close, and at the same time, they intersect with the edge lines of the wire ice covering at a close position, which affects the edge recognition. The edge of the wire ice covering is often identified together with the edge of the wire ice covering. Therefore, it is necessary to exclude the interference of the edge probability map of the wire ice covering on edge recognition from the edge probability map output by the HED neural network.

[0099] Since the edges of the ice covering the wire are also generally distributed in a straight line and are not parallel to the wire, in the image, the edges of the two ice coverings form a "Y" shape, intersecting at a certain point, or not intersecting but at a large angle. Therefore, this distribution feature is used to eliminate it. Specifically: First, for the edge probability map output by the HED neural network, the position of each point in the edge probability map is recorded as coordinates. This records the probability and corresponding position of each point on the edge line. High-probability points are then selected and the corresponding fitting line is obtained using the least squares method. This fitting line primarily corresponds to the high-probability edge line of the conductor ice cover, while low-probability points have little effect on the generation of the fitting line. This results in a fitting line on each side of the conductor, and the fitting lines are generally parallel to the conductor, meaning that the angle between the fitting line and the conductor is relatively small. High-probability points and low-probability points are relative, and probability values ​​are set as needed to distinguish them.

[0100] Similarly, for the edge of the wire-covered ice, the least squares method is used to obtain two fitted lines. Because the angle between the wire-covered ice and the conductor-covered ice is large, if the angle between the fitted line corresponding to the wire-covered ice edge and the conductor-covered ice edge exceeds a preset angle threshold (set as needed), the edge lines corresponding to these two fitted lines are directly excluded, thus eliminating wire interference.

[0101] In step S400, the conductor ice coating measurement data includes the major and minor diameters and density of the conductor ice coating.

[0102] (1) The length and short diameter of the conductor ice coating In this embodiment, the scene is a simple scene with a fixed distance and a fixed viewing angle, and the calculation method and parameters are fixed. Specifically: Binarize the fused edge probability map to identify the edge of the ice-covered wire; Based on the two edges corresponding to the major or minor diameters, the pixel distance between the corresponding pixels of the two edges in the pixel space is calculated. In order to reduce the amount of calculation, the pixel distance is not calculated pixel by pixel, but is calculated every several pixels (for example, 5 pixels), and then the average of all pixel distances is obtained; The actual major or minor diameter of the wire ice is calculated using the pixel distance or the average pixel distance according to the scaling factor between pixel space and actual space, which is determined during the initial camera calibration process.

[0103] (2) Density of ice coating on conductors According to the "Technical Code for Meteorological Survey of Power Engineering Projects" (DL / T 5158-2021), the density of ice covering the conductor can be calculated according to the following formula.

[0104]

[0105] in: is the density of the ice covering the conductor (g / cm 3 ) is the weight of ice covering the conductor (g), which can be obtained by the weighing sensor on the conductor.

[0106] is the length of the ice covering the conductor (m) is the long diameter of the conductor ice cover (mm) is the short diameter of the conductor ice cover (mm) is the wire radius (mm) The density of the conductor ice coating can be obtained according to the above formula, and the ice coating type can be further determined based on the density of the conductor ice coating.

[0107] Based on the same technical concept, an embodiment of the present invention further provides an electronic device that can implement the conductor ice coverage measurement method provided in the above embodiment of the present invention. In one embodiment, the electronic device can be a server, or a terminal device or other electronic device. Figure 5 As shown, the electronic device may include: At least one processor, and a memory connected to the at least one processor. The embodiment of the present invention does not limit the specific connection medium between the processor and the memory. Figure 5 The example in this article is that the processor and memory are connected via a bus. Figure 5 The connections between the other components are shown in bold lines, which are only for illustration and not intended to be limiting. The bus can be divided into address bus, data bus, control bus, etc. Figure 5 The processor is represented by a single thick line, but this does not mean that there is only one bus or only one type of bus. Alternatively, the processor can also be called a controller, without any limitation on the name.

[0108] In an embodiment of the present invention, the memory stores instructions that can be executed by at least one processor, and the at least one processor can execute the wire ice coverage measurement method discussed above by executing the instructions stored in the memory. The processor can implement Figure 5 The functions of each module in the device shown.

[0109] Among them, the processor is the control center of the device, which can use various interfaces and lines to connect the various parts of the entire control device, and monitor the device as a whole by running or executing instructions stored in the memory and calling data stored in the memory, the various functions of the device and processing data.

[0110] In an optional design, the processor may include one or more processing units, and the processor may integrate an application processor and a modem processor, wherein the application processor primarily processes the operating system, user interface, and application programs, and the modem processor primarily processes wireless communications. It is understood that the modem processor may not be integrated into the processor. In some embodiments, the processor and memory may be implemented on the same chip, or in some embodiments, they may be implemented on separate chips.

[0111] The processor can be a general-purpose processor, such as a CPU, digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, and can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the wire ice coverage measurement method disclosed in the embodiments of the present invention can be directly executed by a hardware processor or by a combination of hardware and software modules within the processor.

[0112] As a non-volatile computer-readable storage medium, memory can be used to store non-volatile software programs, non-volatile computer executable programs and modules. Memory can include at least one type of storage medium, for example, can include flash memory, hard disk, multimedia card, card-type memory, random access memory (Random Access Memory, RAM), static random access memory (Static Random Access Memory, SRAM), programmable read-only memory (Programmable Read Only Memory, PROM), read-only memory (Read Only Memory, ROM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), magnetic memory, disk, optical disk, etc. Memory is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in the embodiment of the present invention can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or data.

[0113] By programming a processor, the code corresponding to the method for measuring ice coverage on a conductor described in the aforementioned embodiment can be embedded in the chip, enabling the chip to execute the steps of the method described in the aforementioned embodiment when running. Designing and programming a processor is well known to those skilled in the art and will not be further described here.

[0114] Based on the same inventive concept, an embodiment of the present invention further provides a storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer executes the wire ice coverage measurement method discussed above.

[0115] In some optional embodiments, the present invention also provides various aspects of a method for measuring ice coating on a conductor, which can also be implemented in the form of a program product, which includes program code. When the program product is run on an apparatus, the program code is used to enable the control device to execute the steps of a method for measuring ice coating on a conductor according to various exemplary embodiments of the present invention described above in this specification.

[0116] It should be noted that although several units or subunits of the device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of a unit described above can be further divided into multiple units to be embodied. In addition, although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0117] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, 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 magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] 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 a 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 generate a server, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process 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.

[0119] Program code for performing the operations of the present invention may be written using any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0120] Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0121] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0123] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for measuring ice coverage on a conductor, characterized in that: include: Acquire images of ice covering the conductor, including color images, infrared images and blue light images; The HED algorithm is used to perform edge recognition on the wire ice-covered body image to obtain edge probability maps corresponding to the color image, infrared image and blue light image; The edge probability maps of the color image, the infrared image and the blue light image are fused based on the dynamic weight to obtain a fused edge probability map; Based on the fused edge probability map, wire ice-covered body measurement data is obtained.

2. The method for measuring ice coating on a conductor according to claim 1, wherein: Two sets of cameras are used to capture images of ice covering the wires. Each set of cameras includes a color camera, an infrared camera, and a blue light camera. One set of cameras captures images of the ice covering the wires horizontally, while the other set of cameras captures images of the ice covering the wires vertically. Each set of cameras is equipped with one or more fill lights, each of which is connected to a light intensity sensor. The light intensity sensor is used to detect the light intensity of the current environment of the ice covering the wires. When the detected light intensity is lower than a light intensity threshold, the fill light is turned on.

3. The method for measuring ice coating on a conductor according to claim 1, wherein: The HED algorithm is an improved HED algorithm; The improved HED algorithm refers to the HED algorithm that adopts a layer-by-layer cumulative evaluation strategy to perform fusion calculation on the side output layer.

4. The method for measuring ice coating on a conductor according to claim 3, wherein: The layer-by-layer cumulative evaluation strategy is used to perform fusion calculation on the side output layer, including: When obtaining the i After the side output layer, i >1, immediately fuse the generated side output layer 1 to the side output layer i The result is then binarized to form edges, and the edge density of the current fusion image is calculated according to the following formula: : in, N is the number of valid edge pixels in the fusion image, W × H is the total number of pixels of the fusion image; Then, the fusion calculation is terminated early according to the following judgment conditions: i =3, if , directly stop the fusion calculation of side output layer 4 and side output layer 5; when It is necessary to perform the fusion calculation of the side output layer 4. After completing the fusion calculation of the side output layer 4, if Stop the fusion calculation of the side output layer 5, otherwise the fusion calculation of the side output layer 5 is also required; when It is determined that the fusion calculation of the side output layer 4 and the side output layer 5 needs to be performed.

5. The method for measuring ice coating on a conductor according to claim 1, wherein: The step of fusing the edge probability maps of the color image, the infrared image, and the blue light image based on the dynamic weight to obtain a fused edge probability map includes: Calculating the haze index F ; Based on the haze index F and light intensity, calculate the basic weights of color images, infrared images and blue light images; Adjust the basic weight of each image according to the edge density of each image; Calculate dynamic weights based on the adjusted basic weights of each image; The edge probability maps of the color image, the infrared image, and the blue light image are weightedly calculated with the corresponding dynamic weights to obtain a fused edge probability map.

6. The method for measuring ice coating on a conductor according to claim 5, characterized in that: The calculation of haze index F ,include: Normalize the single channel value V of the image to the range [0,1]; Divide the image into N × N blocks, calculate the average brightness of each block; select the front block with the highest brightness K blocks, and take the average brightness as the atmospheric light estimation A ; Based on atmospheric light estimation A , find the global brightness deviation GBD : in, W is the width of the image, H is the height of the image; Calculate the local contrast attenuation factor LAC: Divide the image into M × M Block, for each block, calculate the contrast according to the formula C i : Then calculate the contrast mean of all blocks , then the local contrast attenuation factor LAC for: Based on global brightness bias GBD and the local contrast attenuation factor LAC , calculate the haze index F for: in, α It is an empirical coefficient that balances the contribution of brightness and contrast.

7. The method for measuring ice coating on a conductor according to claim 6, wherein: The haze index F and light intensity, calculate the basic weights of color images, infrared images, and blue light images, including: The base weight W of the color image c_base for: The basic weight W of the infrared image i_base for: The base weight W of the blue light image b_base for: in: Where, L is the normalized light intensity, L n The light intensity currently obtained by the light intensity sensor.

8. The method for measuring ice coating on a conductor according to claim 7, characterized in that: The adjusting the basic weight of each image according to the edge density of each image includes: in, Represents the basic weight of the adjusted color image, represents the edge density of the color image, Represents the adjusted infrared image basic weight, represents the edge density of infrared image, Represents the adjusted blue image base weight, Indicates the edge density of the blue image.

9. The method for measuring ice coating on a conductor according to claim 8, characterized in that: The calculating of the dynamic weight based on the adjusted basic weights of each image includes: Where, W c is the dynamic weight of the color image, W i is the dynamic weight of the infrared image, W b is the dynamic weight of the blue image.

10. The method for measuring ice coating on a conductor according to claim 1, wherein: It is necessary to exclude the interference of the edge probability map of the wire-covered ice body on edge recognition from the edge probability map output by the HED neural network. Specifically: for the edge probability map output by the HED neural network, the position of each point in the edge probability map is recorded in the form of coordinates, thereby recording the probability and corresponding position of each point on the edge line; then, high-probability points are selected and the corresponding fitting straight lines are obtained by the least squares method, so that there is a fitting straight line on each side of the wire; similarly, for the edge of the wire-covered ice body, the least squares method can also be used to obtain the fitting straight lines of two edge lines; when it is detected that the angle between the fitting straight line corresponding to the edge of the wire-covered ice body and the edge line of the wire ice exceeds the preset angle threshold, the edge lines corresponding to these two fitting straight lines are directly excluded.

Images