Power transmission line sag detection method and system based on lightweight neural network model

Abstract

The invention relates to the technical field of sag detection, and provides a power transmission line sag detection method and system based on a lightweight neural network model, and the method comprises the steps: obtaining a power transmission line image through the aerial photographing of an unmanned plane, and carrying out the preprocessing of the image, and obtaining a sag image; extracting multi-scale features of the sag image through an improved lightweight neural network model, and constructing a multi-scale feature pyramid through a multi-scale feature map of the sag image to output a multi-scale thermodynamic diagram; performing feature fusion on the multi-scale thermodynamic diagram through a dynamic weight fusion module to obtain a fused feature thermodynamic diagram; and positioning sag key points of the fusion feature thermodynamic diagram through a thermodynamic diagram regression method, calculating a sag physical distance according to the sag key points, and correcting the sag physical distance in real time through a temperature compensation algorithm to obtain an actual sag distance. According to the invention, the core bottleneck of sag measurement in electric power inspection is systematically solved, and high-precision, high-efficiency and high-robustness full-automatic detection is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of sag detection, and in particular to a method and system for detecting the sag of a transmission line based on a lightweight neural network model. Background Art

[0002] Sag detection monitors the sag of the conductors of a transmission line, directly correlates with the safety margin and load capacity of the transmission line. Accurate sag detection can effectively prevent the risk of short - circuit or fire caused by discharge to the ground or crossing objects due to excessive sag, and the risk of wire breakage caused by excessive sag under extreme conditions such as strong wind and icing. Real - time sag data can guide the power grid to dynamically adjust the transmission capacity, avoid power loss caused by overheating of the conductors. Combining with meteorological data, abnormal changes in sag can warn of the impact of environmental disasters such as icing and high temperature on the line.

[0003] Transmission lines are often deployed in complex environments such as mountains, forests, rain and fog. Factors such as low light, rain, snow and vegetation occlusion lead to a decline in image quality, and traditional algorithms are prone to failure. When shooting with an aerial drone at high altitude, the width of the conductor only occupies a few pixels in the image, and conventional models are difficult to capture high - frequency details, resulting in a high miss - detection rate. The sag of the conductor fluctuates with temperature, affecting the measurement accuracy.

[0004] Existing sag detections mostly rely on manual measurement or traditional image processing, with low efficiency and being easily interfered by the environment. Deep - learning object - detection models (such as Faster R - CNN) have a large amount of computation and are difficult to be deployed on drone edge devices. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the related technologies. For this purpose, the present invention provides a method and system for detecting the sag of a transmission line based on a lightweight neural network model. Through technologies such as lightweight model optimization, multi - scale feature fusion, and temperature compensation algorithms, the core bottleneck of sag measurement in power inspection is systematically solved, the computing power of the model is significantly reduced, and high - precision, high - efficiency, and high - robustness full - automatic detection is achieved.

[0006] The present invention provides a method for detecting the sag of a transmission line based on a lightweight neural network model, including: S1: Obtain a transmission - line image through aerial drone photography, and pre - process the transmission - line image to obtain a sag image; S2: Extract multi - scale features of the sag image through an improved MobileNetV3 lightweight neural network model to obtain a multi - scale feature map of the sag image; S3: Construct a multi - scale feature pyramid through the multi - scale feature map of the sag image, and predict the multi - scale feature map of the sag image through the multi - scale feature pyramid to obtain a multi - scale heat map; S4: Feature fusion of the multi-scale heat maps is performed through a dynamic weight fusion module to obtain a fused feature heat map; S5: The sag key points of the fused feature heat map are located by the heat map regression method, the physical distance of the sag is calculated based on the sag key points, and the physical distance of the sag is corrected in real time through a temperature compensation algorithm to obtain the actual sag distance.

[0007] According to a transmission line sag detection method based on a lightweight neural network model provided by the present invention, the preprocessing further includes insulator detection, segment transformation, and wire characteristic enhancement. The insulator detection is to locate the coordinates of the top and bottom of the UAV aerial image through a lightweight HRNet model to obtain an insulator detection image. The segment transformation is to divide the insulator detection image into N segment images, calculate the transformation matrix independently for each segment image, and correct each segment of the wire area into a horizontal straight line through the transformation matrix. The wire characteristic enhancement is to enhance the wire characteristics in a low-contrast environment through the CLAHE algorithm.

[0008] According to a transmission line sag detection method based on a lightweight neural network model provided by the present invention, the improved MobileNetV3 lightweight neural network model further includes a shallow feature reuse module, a dynamic depthwise separable convolution, and an adaptive spatial attention module. The shallow feature reuse module performs synchronous downsampling through a convolutional path and an average pooling path. The dynamic depthwise separable convolution automatically adjusts the depth of the convolutional kernel according to the complexity of the input image. An adaptive spatial attention module is inserted before the residual connection of each neck network in the MobileNetV3 model.

[0009] According to a transmission line sag detection method based on a lightweight neural network model provided by the present invention, multiple key layer features are obtained through the improved MobileNetV3 lightweight neural network model, multiple feature layers are constructed based on the multiple key layer features, and a multi-scale feature pyramid is obtained. Each key layer feature has a different resolution, and each feature layer performs detection and localization of targets at different scales according to the resolution.

[0010] According to a transmission line sag detection method based on a lightweight neural network model provided by the present invention, the sag key points include the suspension points and the lowest points of the transmission line conductors.

[0011] A method for detecting the sag of a transmission line based on a lightweight neural network model provided by the present invention further includes that the dynamic weight fusion module dynamically adjusts the weights of multi-scale heatmap features according to the height of the unmanned aerial vehicle, and performs feature fusion on the multi-scale heatmaps according to the weights of the multi-scale heatmap features to obtain a fused feature heatmap.

[0012] A method for detecting the sag of a transmission line based on a lightweight neural network model provided by the present invention further includes that locating the sag key points of the fused feature heatmap by the heatmap regression method includes: S51: Obtain an optimized heatmap by removing adjacent response peaks on the fused feature heatmap; S52: Find the position with the maximum response value on the optimized heatmap to obtain the coordinates of the sag key points on the heatmap; S53: Repeat step S52 to obtain the coordinates of multiple sag key points on the heatmap; S54: Map the coordinates of the sag key points on the heatmap back to the coordinate space of the unmanned aerial vehicle aerial photography image through proportional scaling to obtain the positions of the sag key points.

[0013] A method for detecting the sag of a transmission line based on a lightweight neural network model provided by the present invention further includes that the calculation expression of the physical sag distance is: Wherein, is the physical sag distance, is the control coefficient, is the vertical pixel distance of the sag key points in the fused feature heatmap, is the horizontal tension of the wire, is the field of view angle of the unmanned aerial vehicle camera, is the pixel height of the tower pole, is the tangent function.

[0014] A method for detecting the sag of a transmission line based on a lightweight neural network model provided by the present invention further includes that the calculation expression of the actual sag distance is: Wherein, is the actual sag distance, is the physical sag distance, is the wire temperature expansion coefficient, is the difference between the current temperature and the reference temperature.

[0015] The present invention also provides a transmission line sag detection system based on a lightweight neural network model for executing the method for detecting the sag of a transmission line based on a lightweight neural network model described in any one of the above, including: Sag image acquisition module, which acquires transmission line images through UAV aerial photography and preprocesses the transmission line images to obtain sag images; Feature extraction module, which extracts multi-scale features of the sag image through an improved MobileNetV3 lightweight neural network model to obtain a multi-scale feature map of the sag image; Multi-scale heat map acquisition module, which constructs a multi-scale feature pyramid through the multi-scale feature map of the sag image and predicts the multi-scale feature map of the sag image through the multi-scale feature pyramid to obtain a multi-scale heat map; Fusion module, which performs feature fusion on the multi-scale heat map through a dynamic weight fusion module to obtain a fused feature heat map; Sag distance acquisition module, which locates the sag key points of the fused feature heat map through the heat map regression method, calculates the physical sag distance according to the sag key points, and corrects the physical sag distance in real time through a temperature compensation algorithm to obtain the actual sag distance.

[0016] One or more of the above technical solutions in the embodiments of the present invention have at least one of the following technical effects: Through technologies such as lightweight model optimization, multi-scale feature fusion, and temperature compensation algorithm, the core bottleneck of sag measurement in power line inspection is systematically solved, the computing power of the model is significantly reduced, and high-precision, high-efficiency, and high-robustness full-automatic detection is achieved.

[0017] The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a schematic flowchart of a transmission line sag detection method based on a lightweight neural network model provided by the present invention.

[0020] Figure 2 It is a schematic structural diagram of a transmission line sag detection system based on a lightweight neural network model provided by the present invention.

[0021] Reference Signs: 101. Sag image acquisition module; 102. Feature extraction module; 103. Multi-scale heat map acquisition module; 104. Fusion module; 105. Sag distance acquisition module. Detailed implementation manners

[0022] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0023] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without conflict, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0024] The following will be combined with Figures 1 to 2 to describe a transmission line sag detection method and system based on a lightweight neural network model of the present invention.

[0025] As Figure 1 shown, a transmission line sag detection method based on a lightweight neural network model includes: S1: Obtain a transmission line image through aerial photography by an unmanned aerial vehicle (UAV), and preprocess the transmission line image to obtain a sag image; In some specific embodiments of the present invention, a transmission line image is obtained by mounting a visible light camera on a UAV.

[0026] The transmission line includes high-voltage cables, substation transmission lines, photovoltaic power station transmission lines, and railway catenary transmission lines.

[0027] The UAV GPS synchronously records the longitude, latitude and altitude of each frame of image, corrects the pitch / yaw angle through a combined inertial navigation system, triggers an alarm when the depression angle > 30°, and dynamically adjusts the shutter speed according to the light intensity, so as to ensure that the wire area is not overexposed.

[0028] The preprocessing includes insulator detection, segment transformation, and wire feature enhancement. Insulator detection is to locate the coordinates of the top and bottom of the UAV aerial image through a lightweight HRNet model to obtain an insulator detection image. Segment transformation is to divide the insulator detection image into N segment images, independently calculate the transformation matrix for each segment image, and correct each segment of the wire area into a horizontal straight line through the transformation matrix. Wire feature enhancement is to enhance the wire features in a low-contrast environment through the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm. Low contrast includes rain, fog, and light and shadow interference.

[0029] In some specific embodiments of the present invention, N = 3, corresponding to the wire sag segment, the left tower segment, and the right tower segment.

[0030] S2: Extract multi-scale features of the sag image through an improved MobileNetV3 lightweight neural network model to obtain a multi-scale feature map of the sag image. The improved MobileNetV3 lightweight neural network model includes a shallow feature reuse module, a dynamic depthwise separable convolution, and an adaptive spatial attention module. The shallow feature reuse module performs synchronous downsampling through a convolutional path and an average pooling path. The convolutional path extracts low-level semantic features of the sag image through a 3×3 convolution and activates them through the Hardswish activation function. The low-level semantic features include wire edges and insulator textures. The Hardswish activation function has a relatively simple calculation process, which can accelerate the training and inference speed of the model and can automatically adjust the activation degree according to different input data.

[0031] The pooling path downsamples the sag image with the same parameters as the convolutional path through 3×3 average pooling. Average pooling retains more overall illumination information than max pooling, avoids direct skipping of pooling resulting in size mismatches, reduces the size of the feature map, reduces the computational amount, and retains the overall features within the region. Synchronous downsampling of the convolutional path and the average pooling path can comprehensively utilize the abstract features extracted by the convolutional operation and the original information retained by the pooling operation, thereby improving the performance and feature expression ability of the model.

[0032] The dynamic depthwise separable convolution automatically adjusts the depth of the convolution kernel according to the complexity of the input image. Dynamic depthwise separable convolution can automatically increase the convolution depth in the area where wires are dense; dynamic depthwise separable convolution includes an inverted residual structure, and the inverted residual structure consists of three convolutional layers: the first ( ) convolutional layer is used for dimensionality increase, the middle depthwise convolutional layer is used for feature extraction, and the last ( ) convolutional layer is used for dimensionality reduction. In this structure, the depthwise convolution is the core component. In sag detection, the depthwise convolution can perform local feature extraction on the images of transmission lines, such as features like the edges and textures of the lines.

[0033] An adaptive spatial attention module is inserted before the residual connection of each neck network in the MobileNetV3 model.

[0034] By inserting an adaptive spatial attention module before the residual connection of each neck network, the MobileNetV3 model can pay more attention to the key features of the transmission line, such as the edges and contours of the line, thus improving the performance in the real-time sag detection task of the transmission line.

[0035] Inserting an adaptive spatial attention module before the residual connection of each neck network enables the model to focus on the key feature regions with the help of the attention mechanism and suppress useless information; while the dynamic depthwise separable convolution adjusts the convolution kernel parameters dynamically according to the input through the dynamic depthwise separable convolution, better adapting to different features and enhancing the feature extraction ability of the model.

[0036] S3: Construct a multi-scale feature pyramid from the multi-scale feature map of the sag image, and predict the multi-scale feature map of the sag image through the multi-scale feature pyramid to obtain a multi-scale heat map; Obtain multiple key layer features through the improved MobileNetV3 lightweight neural network model, construct multiple feature layers according to the multiple key layer features to obtain a multi-scale feature pyramid; Each key layer feature has a different resolution, and each feature layer performs detection and localization of targets at different scales according to the resolution.

[0037] S4: Perform feature fusion on the multi-scale heat map through the dynamic weight fusion module to obtain a fused feature heat map; Features of different layers have different characteristics. Shallow features usually contain more specific detailed information, such as low-level features like the edges and textures of the transmission line; while deep features focus more on abstract semantic information and can capture the overall structure of the transmission line and high-level features related to sag. The dynamic weight fusion module integrates these different levels of features according to weights to better perform sag detection; The dynamic weight fusion module dynamically adjusts the feature weights of the multi-scale heat map according to the UAV altitude and the requirements of the task. The calculation expression of the multi-scale heat map feature weights is: Among them, is the thermal map feature weight of the th scale, is the slope factor, is the UAV altitude, is the th scale of the preset height reference value of the feature, At the same time, ensure that the sum of all weights is 1, and the calculation expression is: Among them, is the number of feature scales.

[0038] S5: Locate the sag key points of the fused feature thermal map through the thermal map regression method, calculate the physical sag distance according to the sag key points, and correct the physical sag distance in real time through the temperature compensation algorithm to obtain the actual sag distance.

[0039] The sag key points include the suspension points and the lowest points of the transmission line conductors.

[0040] Locating the sag key points of the fused feature thermal map through the thermal map regression method includes: S51: Obtain the optimized thermal map by removing adjacent response peaks on the fused feature thermal map; S52: Find the position with the maximum response value on the optimized thermal map to obtain the coordinates of the sag key points on the thermal map; S53: Repeat step S52 to obtain the coordinates of multiple sag key points on the thermal map; S54: Map the coordinates of the sag key points on the thermal map back to the coordinate space of the UAV aerial image through proportional scaling to obtain the positions of the sag key points.

[0041] The calculation expression of the physical sag distance is: Among them, is the physical sag distance, is the control coefficient, is the vertical pixel distance of the sag key points in the fused feature thermal map, is the horizontal tension of the conductor, is the field of view angle of the UAV camera, is the pixel height of the tower pole, is the tangent function.

[0042] The selection and innovation of this formula are derived from the catenary sag calculation formula in the "High Voltage Transmission Line Design Manual for Electric Power Engineering". Convert the image information into physical quantities. Considering that the sag length and tension are coupled, and this relationship is related to environmental factors such as temperature, icing, wind speed, etc., and the wire's own properties such as wire elastic modulus and cross-sectional area, through the mixed operation of various parameters in the line manual, a control coefficient is introduced after simplification. , The unit of , is not a fixed value. Due to environmental changes, it can be obtained by calibrating with a drone on the line path with a known sag.

[0043] According to the flight attitude sensor data of the drone or based on the geometric relationship between the wire and objects such as towers in the image, determine the vertical direction angle during shooting, and then calculate ; Consult design documents, operation and maintenance records and other materials of the transmission line, or measure the horizontal tension of the wire in real time through a tension sensor; Obtain information such as the image resolution set when the drone camera takes pictures to directly get the image height; Obtain the current temperature from the meteorological monitoring equipment in the environment where the transmission line is located, and determine the wire temperature expansion coefficient according to the relevant material characteristics of the wire; In a specific embodiment of the present invention, the wire model is LGJ-185 / 35, the span is 150 meters, and the vertical pixel distance of the sag key point in the fusion feature heat map is , and the field of view angle of the drone camera during shooting is known through the attitude sensor carried by the drone , then .

[0044] Consult the transmission line design document to know the wire horizontal tension , the vertical pixel distance of the sag key point in the fusion feature heat map generated by the lightweight model for the image taken by the drone camera is , , the pixel height of the tower , the tower height , , , obtain the current temperature as 30 °C through the meteorological sensor, the reference temperature is 20 °C, and the temperature change is 10 °C.

[0045] Thus, the physical distance of the sag is obtained .

[0046] Eliminate seasonal errors through real-time temperature correction.

[0047] The calculation expression for the actual sag distance is: Among them, is the actual sag distance, is the physical sag distance, is the temperature expansion coefficient of the wire, is the difference between the current temperature and the reference temperature.

[0048] Thus, the actual sag distance is: Through real-time temperature correction, the accuracy error of the sag distance is less than 5%.

[0049] The present invention applies the vision geometry of unmanned aerial vehicles and the theoretical basis of catenary mechanics. It not only considers the influence of temperature changes on the wire, but also as other factors such as the environment, wire aging, and corrosion change, the thermal expansion coefficient is also changing. By introducing the control coefficient , to cope with the influence brought by these changes, the present invention is the reasonable application and optimization of existing standards in specific scenarios; the vertical pixel distance of the sag key point in the fused feature heat map couples the tension of the transmission line in the overhead line in the original theory with the sag and image pixels. By continuously fitting and adjusting the relevant parameters of the adverse conditions such as corrosiveness and multi-water mist in the chemical industrial area, the measurement accuracy of the wire sag is greatly improved.

[0050] And the control coefficient has calibratability and scene adaptability, and is not a fixed value.

[0051] As Figure 2 shown, a transmission line sag detection system based on a lightweight neural network model and a transmission line sag detection method based on a lightweight neural network model include: The sag image acquisition module 101 obtains the transmission line image through aerial photography by an unmanned aerial vehicle, and preprocesses the transmission line image to obtain a sag image; The feature extraction module 102 extracts multi-scale features of the sag image through an improved MobileNetV3 lightweight neural network model to obtain a multi-scale feature map of the sag image; The multi-scale heat map acquisition module 103 constructs a multi-scale feature pyramid through the multi-scale feature map of the sag image, and predicts the multi-scale feature map of the sag image through the multi-scale feature pyramid to obtain a multi-scale heat map; The fusion module 104 performs feature fusion on the multi-scale heat map through a dynamic weight fusion module to obtain a fused feature heat map; The sag distance acquisition module 105 locates the sag key point of the fused feature heat map through the heat map regression method, calculates the physical sag distance according to the sag key point, and corrects the physical sag distance in real time through a temperature compensation algorithm to obtain the actual sag distance.

[0052] Through the collaborative work of the above-mentioned modules, by means of technologies such as lightweight model optimization, multi-scale feature fusion, and temperature compensation algorithms, the core bottleneck of sag measurement in power inspection is systematically solved, the computing power of the model is significantly reduced, and full-automatic detection with high precision, high efficiency, and high robustness is achieved.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the sag of a transmission line based on a lightweight neural network model, characterized in that, Including: S1: Obtain the transmission line image through UAV aerial photography, and preprocess the transmission line image to obtain the sag image; S2: Extract the multi-scale features of the sag image through the improved MobileNetV3 lightweight neural network model to obtain the multi-scale feature map of the sag image; S3: Construct a multi-scale feature pyramid through the multi-scale feature map of the sag image, and predict the multi-scale feature map of the sag image through the multi-scale feature pyramid to obtain the multi-scale heat map; S4: Perform feature fusion on the multi-scale heat map through the dynamic weight fusion module to obtain the fused feature heat map; S5: Locate the sag key points of the fused feature heat map through the heat map regression method, calculate the physical sag distance according to the sag key points, and correct the physical sag distance in real time through the temperature compensation algorithm to obtain the actual sag distance.

2. The method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, wherein The preprocessing includes insulator detection, segment transformation, and wire feature enhancement. The insulator detection is to locate the coordinates of the top and bottom of the UAV aerial photography image through the lightweight HRNet model to obtain the insulator detection image. The segment transformation is to divide the insulator detection image into N segment images, calculate the transformation matrix independently for each segment image, and correct each segment of the wire area into a horizontal straight line through the transformation matrix; The wire feature enhancement is to enhance the wire features in a low-contrast environment through the CLAHE algorithm.

3. A method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, The improved MobileNetV3 lightweight neural network model includes a shallow feature reuse module, dynamic depthwise separable convolution, and an adaptive spatial attention module; The shallow feature reuse module performs synchronous downsampling through the convolutional path and the mean pooling path; The dynamic depthwise separable convolution automatically adjusts the convolution kernel depth according to the complexity of the input image; Insert an adaptive spatial attention module before the residual connection of each neck network in the MobileNetV3 model.

4. A method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, Obtain multiple key layer features through the improved MobileNetV3 lightweight neural network model, construct multiple feature layers according to the multiple key layer features, and obtain the multi-scale feature pyramid; Each key layer feature has a different resolution, and each feature layer performs detection and localization of targets at different scales according to the resolution.

5. The method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, The sag key points include the suspension points and the lowest points of the transmission line conductors.

6. The method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, The dynamic weight fusion module dynamically adjusts the feature weights of the multi-scale heat map according to the UAV height, and performs feature fusion on the multi-scale heat map according to the feature weights of the multi-scale heat map to obtain the fused feature heat map.

7. A method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, Locating the sag key points of the fused feature heat map through the heat map regression method includes: S51: Obtain the optimized heat map by removing adjacent response peaks on the fused feature heat map; S52: Find the position with the largest response value on the optimized heat map to obtain the coordinates of the sag key points on the heat map; S53: Repeat step S52 to obtain the coordinates of multiple sag key points on the heat map; S54: Map the coordinates of the sag key points on the heat map back to the coordinate space of the UAV aerial photography image through scale scaling to obtain the positions of the sag key points.

8. A method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, wherein, The calculation expression of the physical sag distance is: Among them, is the sag physical distance, is the control coefficient, is the vertical pixel distance of the sag key point in the fusion feature heat map, is the horizontal tension of the conductor, is the field of view angle of the UAV camera, is the pixel height of the tower pole, is the tangent function.

9. A method for detecting the sag of a transmission line based on a lightweight neural network model according to claim 1, characterized in that, The calculation expression of the actual sag distance is: Among them, is the actual sag distance, is the physical sag distance, is the wire temperature expansion coefficient, is the difference between the current temperature and the reference temperature.

10. A sag detection system for transmission lines based on a lightweight neural network model, characterized in that, A method for detecting the sag of a transmission line based on a lightweight neural network model as described in any one of claims 1 to 9, comprising: A sag image acquisition module, which acquires a transmission line image through aerial photography by a drone and preprocesses the transmission line image to obtain a sag image; A feature extraction module, which extracts multi-scale features of the sag image through an improved MobileNetV3 lightweight neural network model to obtain a multi-scale feature map of the sag image; A multi-scale heat map acquisition module, which constructs a multi-scale feature pyramid through the multi-scale feature map of the sag image, and predicts the multi-scale feature map of the sag image through the multi-scale feature pyramid to obtain a multi-scale heat map; A fusion module, which performs feature fusion on the multi-scale heat map through a dynamic weight fusion module to obtain a fused feature heat map; A sag distance acquisition module, which locates the sag key points of the fused feature heat map through a heat map regression method, calculates the physical sag distance according to the sag key points, and corrects the physical sag distance in real time through a temperature compensation algorithm to obtain the actual sag distance.

Images