Power transmission tower environment inversion monitoring method and device, terminal equipment and storage medium

Abstract

The invention discloses a power transmission tower environment inversion monitoring method and device, terminal equipment and a storage medium, and relates to the technical field of power systems, and the method comprises the steps: obtaining multi-source data of a target region of a power transmission tower; building mask information of the target area is extracted based on the optical remote sensing image and the dual-polarization remote sensing image, and the building mask information in the dual-polarization remote sensing image and the optical remote sensing image is removed to obtain a to-be-input remote sensing image; inputting a remote sensing image to be input and meteorological data into a pre-constructed convolution model as input data; and performing inversion monitoring according to the input data through the convolution model, and outputting a plurality of meteorological parameter predicted values. According to the method, two different types of remote sensing data are combined to extract the building mask information of the power transmission tower in a complex environment, so that the precision error caused by taking the building mask information as a misjudgment area can be effectively avoided, and the environment inversion monitoring effect can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, device, terminal equipment and storage medium for inversion monitoring of a transmission tower environment. Background Art

[0002] With the global energy transition and increasing demand for renewable energy, wind and hydropower technologies have garnered widespread attention as clean, sustainable forms of energy. To further achieve the goals of energy conservation, emission reduction, and carbon reduction, my country is adopting renewable energy technologies such as photovoltaics and wind power, which offer significant environmental advantages. With the continuous expansion of the power system and the annual growth of grid load, the operating environment of transmission lines faces increasingly complex challenges. High-voltage transmission lines and their tower sections, in particular, are significantly impacted by both the natural environment and human factors. To ensure the stability and safety of the power system and promptly identify potential risks and hidden dangers, operating environment monitoring technology for transmission line tower sections has gradually become a research focus in the power sector.

[0003] Existing inversion monitoring methods for transmission tower environments are usually based on single optical image data. However, in complex surface environments, weather and visibility changes greatly restrict the effectiveness of optical remote sensing technology. Using single optical image data, it is impossible to obtain effective information about the transmission tower area, resulting in poor inversion monitoring of the transmission tower environment. Summary of the Invention

[0004] The present invention provides a method, apparatus, terminal device and storage medium for inversion monitoring of a transmission tower environment, which can solve the technical problem in the prior art that single optical image data cannot obtain effective information about the transmission tower area, resulting in poor inversion monitoring effect of the transmission tower environment.

[0005] The present invention provides a method for inversion monitoring of a transmission tower environment, comprising:

[0006] Acquiring multi-source data of a target area of ​​a transmission tower, wherein the multi-source data includes optical remote sensing images, dual-polarization remote sensing images, and meteorological data;

[0007] extracting building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and removing the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain a remote sensing image to be input;

[0008] Inputting the remote sensing image to be input and the meteorological data as input data into a pre-built convolutional model;

[0009] The convolution model is used to perform inversion monitoring based on the input data and output a plurality of meteorological parameter prediction values.

[0010] Furthermore, the extracting of building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image includes:

[0011] Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization;

[0012] Selecting a band based on the optical remote sensing image to determine band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information;

[0013] Building mask information of the target area is determined according to the normalized backscatter coefficient and the band spectrum information.

[0014] Furthermore, determining the building mask information of the target area according to the normalized backscatter coefficient and the band spectrum information includes:

[0015] A polarization building index is determined according to the normalized backscatter coefficient, a spectral building index is determined according to the band spectrum information, and weighted accumulation is performed based on the polarization building index and the spectral building index to obtain building mask information of the target area.

[0016] Furthermore, the step of inputting the remote sensing image to be input and the meteorological data as input data into a pre-built convolutional model includes:

[0017] performing outlier detection on the meteorological data and removing outliers in the meteorological data;

[0018] Adjusting the meteorological data after removing outliers to spatial grid data with the same resolution as the remote sensing image, and determining the spatial grid data as multi-dimensional meteorological data;

[0019] The multi-dimensional meteorological data and the remote sensing image to be input are used as input data.

[0020] Furthermore, the convolution model is used to perform inversion monitoring based on the input data and output multiple meteorological parameter prediction values, including:

[0021] splicing the remote sensing image to be input in the input data and the multi-dimensional meteorological data in the channel dimension through the convolution model to obtain a spliced ​​image;

[0022] Using multiple convolution kernels to extract multi-scale features in the spliced ​​image to obtain a shared feature map vector;

[0023] Performing a local pooling operation along the rows or columns of the shared feature map vector to obtain a pooled feature map;

[0024] The graph vector obtained by flattening the pooled feature graph is input into a fully connected layer, and multiple meteorological parameter prediction values ​​are output through the fully connected layer.

[0025] Furthermore, after outputting the forecast values ​​of multiple meteorological parameters, the following is also included:

[0026] The degree of similarity between the predicted value and the observed value of the meteorological parameter is determined, and when the degree of similarity is higher than a preset threshold, the model inversion result is determined to be good.

[0027] The present invention provides a transmission tower environment inversion monitoring device, comprising:

[0028] A multi-source data acquisition module is used to acquire multi-source data of the target area of ​​the transmission tower, wherein the multi-source data includes optical remote sensing images, dual-polarization remote sensing images and meteorological data;

[0029] a module for acquiring a remote sensing image to be input, configured to extract building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and remove the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain a remote sensing image to be input;

[0030] A data input module, configured to input the remote sensing image to be input and the meteorological data as input data into a pre-built convolution model;

[0031] The inversion monitoring module performs inversion monitoring based on the input data through the convolution model and outputs a plurality of meteorological parameter prediction values.

[0032] Furthermore, the extracting of building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image includes:

[0033] Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization;

[0034] Selecting a band based on the optical remote sensing image to determine band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information;

[0035] Building mask information of the target area is determined according to the normalized backscatter coefficient and the band spectrum information.

[0036] The present invention provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned transmission tower environment inversion monitoring method is implemented.

[0037] The present invention provides a computer-readable storage medium, comprising: a stored computer program, wherein when the computer program is run, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned transmission tower environment inversion monitoring method.

[0038] The following beneficial effects are achieved by implementing the present invention:

[0039] The present invention combines optical remote sensing images and dual-polarization remote sensing images to extract building mask information of the target area, removes the building mask information in the dual-polarization remote sensing images and the optical remote sensing images, and obtains a remote sensing image to be input. By combining two different types of remote sensing data, the building mask information of the transmission tower in a complex environment is extracted, which can effectively avoid the accuracy error caused by the building mask information being used as a misjudgment area, thereby effectively improving the environmental inversion monitoring effect. The processed remote sensing image and meteorological data are input into a pre-built convolution model, and the model is used for inversion monitoring, outputting comprehensive multiple meteorological parameter prediction values, and further improving the effect of transmission tower environmental inversion monitoring.

[0040] Furthermore, the present invention adjusts the meteorological data to the same resolution as the remote sensing image, which can ensure the spatial consistency of the meteorological data and the remote sensing data, thereby facilitating fusion and analysis with the remote sensing image data, and helping the model to more accurately understand and analyze the ground feature in the remote sensing image, thereby improving the overall performance and predictive ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 1 is a flow chart of a method for inversion monitoring of a transmission tower environment provided by one embodiment of the present invention;

[0043] Figure 2 1 is a schematic diagram of the structure of a convolution model provided by one embodiment of the present invention;

[0044] Figure 3 A multi-task branching structure provided by an embodiment of the present invention;

[0045] Figure 4 1 is a flow chart of a method for inversion monitoring of a transmission tower environment provided by one embodiment of the present invention;

[0046] Figure 5 It is a structural diagram of a transmission tower environment inversion monitoring device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0049] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0050] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0051] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0052] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0053] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0054] See also Figure 1 In order to solve the technical problem in the prior art that single optical image data cannot obtain effective information about the transmission tower area, resulting in poor inversion monitoring effect of the transmission tower environment, the present invention provides a transmission tower environment inversion monitoring method, comprising:

[0055] S1. Acquire multi-source data of the target area of ​​the transmission tower, including optical remote sensing images, dual-polarization remote sensing images, and meteorological data;

[0056] In the embodiment of the present invention, the target area may be an area centered on the transmission tower and having a radius of a preset value.

[0057] In an embodiment of the present invention, the optical remote sensing image can be a multi-band optical remote sensing image, the dual-polarization remote sensing image can be a dual-polarization SAR remote sensing image, and the meteorological data can be global high-resolution atmospheric reanalysis data; wherein, the dual-polarization remote sensing image includes VV polarization remote sensing image and VH polarization remote sensing image, and the meteorological data includes wind speed data, temperature data and humidity data.

[0058] In this embodiment of the present invention, to ensure temporal and spatial matching of pixel size data between dual-polarization remote sensing imagery and optical remote sensing imagery, the optical satellite and SRA satellite transit times are aligned, with an interval of no more than one week. The meteorological data in this embodiment of the present invention is hourly data from the same date as the SRA remote sensing imagery.

[0059] S2. extracting building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and removing the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain a remote sensing image to be input;

[0060] S3, inputting the remote sensing image and meteorological data to be input as input data into the pre-built convolutional model;

[0061] S4. Perform inversion monitoring based on the input data through the convolution model and output multiple meteorological parameter prediction values.

[0062] The embodiment of the present invention combines optical remote sensing images and dual-polarization remote sensing images to extract building mask information of the target area, removes the building mask information in the dual-polarization remote sensing images and the optical remote sensing images, and obtains the remote sensing image to be input. By combining two different types of remote sensing data, the building mask information of the transmission tower in a complex environment is extracted, which can effectively avoid the accuracy error caused by the building mask information as the misjudgment area, thereby effectively improving the environmental inversion monitoring effect; and the processed remote sensing image and meteorological data are input into a pre-built convolution model, and the model is used for inversion monitoring, and a comprehensive plurality of meteorological parameter prediction values ​​are output, thereby further improving the effect of the transmission tower environmental inversion monitoring.

[0063] In one embodiment, step S2, extracting building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, includes:

[0064] S21. Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization;

[0065] In the embodiment of the present invention, the dual-polarized remote sensing image can be converted into a backscattering coefficient σ by performing data preprocessing such as overheat noise removal, orbit correction, noise filtering, radiation calibration, geocoding and terrain correction on the dual-polarized remote sensing image. 0 (dB); At the same time, in order to reduce the impact of the incident angle information on the backscattered signal, the embodiment of the present invention adopts a scene-based linear regression model to perform normalization processing on it, and the calculation formula is:

[0066] σ=σ 0 -α(θ m -θ n )

[0067] Where, σ is the normalized backscattering coefficient, σ 0 is the pre-processed backscattering coefficient, α is the proportional coefficient, θ m is the incident angle information of the second scene image, θ n is the incident angle information for the first image. Incident angle normalization adjusts the backscatter signal within a single scene but does not account for backscatter distribution across different polarization orbits. Due to the varying influence of near-surface object parameters on the incident angle effect, the image still contains backscatter variations unrelated to environmental factors, such as interference from buildings and other electromagnetic noise.

[0068] In the embodiment of the present invention, the above-mentioned processing is performed on different polarization remote sensing data respectively to obtain the VV (Vertical Transmit, Vertical Receive radar transmits vertically polarized electromagnetic waves and receives vertically polarized echo signals) polarization normalized backscatter coefficient σ VV and VH (Vertical Transmit, Horizontal Receive, the radar transmits vertically polarized electromagnetic waves and receives horizontally polarized echo signals) polarization normalized backscatter coefficient σ VH ,

[0069] S22. Selecting a band based on the optical remote sensing image to determine band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information;

[0070] In an embodiment of the present invention, optical remote sensing images can be subjected to radiation conversion, atmospheric correction, and band calculation to obtain multi-band data. By band selection, different band spectral information can be determined, including infrared band spectral information and mid-infrared band spectral information.

[0071] S23. Determine building mask information of the target area according to the normalized backscattering coefficient and the band spectrum information.

[0072] In the embodiment of the present invention, by analyzing the normalized backscatter coefficient and the band spectrum information, the spectral characteristics of the building structure in the target area can be identified, thereby generating building mask information and realizing accurate identification and positioning of the building.

[0073] In one embodiment, step S23, determining building mask information of the target area according to the normalized backscatter coefficient and the band spectrum information, includes:

[0074] The polarization building index is determined according to the normalized backscattering coefficient, and the spectral building index is determined according to the band spectrum information. The building mask information of the target area is obtained by weighted accumulation based on the polarization building index and the spectral building index.

[0075] In an embodiment of the present invention, the expression of the polarization building index is as follows:

[0076] PBI=ln(10*σ VV *σ VH )

[0077] Where PBI is the polarization building index, σ VV is the VV polarization normalized backscattering coefficient, σ VH is the VH polarization normalized backscattering coefficient.

[0078] The expression of the spectral building index is as follows:

[0079]

[0080] Among them, NDBI is the spectral building index, R MIR is the near-infrared spectral information, R NIR It is the mid-infrared spectral information.

[0081] In this embodiment of the present invention, built-up areas typically have higher NDBI values, while other ground features, such as soil and vegetation, have lower or near-zero NDBI values. Based on the calculated building spectral index distribution, this embodiment of the present invention can set an NDBI threshold of 0.75 or above to identify built-up areas, and an NDBI threshold between 0 and 0.75 to identify invalid areas.

[0082] The expression of building mask information is as follows:

[0083] Mask = α Mask PBI +β·Mask NDBI

[0084] In the embodiment of the present invention, Mask is building mask information. PBI For polarized building features, Mask NDBI is the optical building feature, α and β represent the corresponding weight coefficients, and α + β = 1. In the embodiment of the present invention, the weight distribution ratio of α can be set to 0.4 and β to 0.6, so as to better realize the complementary advantages of the two and utilize their respective information to improve the accuracy of building filtering.

[0085] The polarimetric building index in this embodiment of the present invention can effectively detect the geometric features of building structures (such as metal and regular planes), while spectral information can capture the spectral reflectance characteristics of building materials (such as the unique spectral lines of concrete and asphalt). Based on the weighted accumulation of the polarimetric building index and the spectral building index, building mask information of the target area is obtained. This can overcome the limitations of a single data source and significantly reduce false detections (such as misidentifying bare land as a building) and missed detections (such as ignoring glass curtain wall buildings), thereby improving the recognition accuracy of building masks.

[0086] In one embodiment, step S3, inputting the remote sensing image and meteorological data to be input as input data into a pre-built convolutional model, includes:

[0087] S31, performing outlier detection on meteorological data and removing outliers in the meteorological data;

[0088] In the embodiment of the present invention, before performing outlier detection on meteorological data, normalization processing is also included:

[0089] Assuming that the meteorological parameters in the meteorological data are X(t,x,y), where t represents time, and x and y represent spatial coordinates, the normalization process is performed using the following formula:

[0090]

[0091] Among them, X norm (t, x, y) are normalized meteorological parameters, μ X and σ X are the mean and standard deviation of the meteorological parameter X in the entire time and space range. Through normalization, the scale differences of meteorological data can be eliminated, providing a unified standard for subsequent analysis.

[0092] The embodiment of the present invention can use the standard deviation method to detect outliers in the wind speed data, temperature data, and humidity data at each time and space point, including:

[0093] Assume that the standard deviations of wind speed, temperature and humidity are σ wind , σ temp , σ wet Outlier detection is performed using the following methods:

[0094] |X(t,x,y)-μ X |>k·σ X

[0095] Here, k is a set threshold factor, usually k=2, which is used to determine whether there are outliers.

[0096] S32, adjusting the meteorological data after removing outliers to spatial grid data with the same resolution as the remote sensing image, and determining the spatial grid data as multi-dimensional meteorological data;

[0097] In an embodiment of the present invention, bilinear interpolation can be used to adjust meteorological data to spatial grid data with the same resolution as the remote sensing image, thereby achieving spatial consistency between image pixels of different resolutions, providing more detailed information for high-resolution images, reducing certain noise impacts, ensuring the data quality of subsequent model input parameters, and avoiding training errors caused by inconsistent resolution or data distortion, thereby improving the accuracy of the description of environmental features around transmission towers.

[0098] The bilinear interpolation method of the embodiment of the present invention uses the four neighboring pixels of the target pixel and uses a linear weighted average to solve the new pixel value. The mathematical description is as follows:

[0099] Set the original target domain f k The four adjacent points are (x1, y1), (x2, y2), (x3, y3), (x4, y4), and the expression of the bilinear difference model is:

[0100] I′(x′,y′)=I(x1,y1)·(1-ω x )·(1-ω y )+I(x2,y1)·ω x ·(1-ω y )+I(x1,y2)·(1-ω x )·ω y +I(x2,y2)·ω x ·ω y

[0101] Among them, I′(x′, y′) is the pixel value of the point (x′, y′) to be estimated, ω x With ω y It represents the normalized weight of (x′, y′) relative to x1, x2 and y1, y2 in the main image, and the calculation formula is:

[0102]

[0103] S33. Using the multi-dimensional meteorological data and the remote sensing image to be input as input data.

[0104] The embodiment of the present invention adjusts the meteorological data to the same resolution as the remote sensing image, which can ensure the spatial consistency of the meteorological data and the remote sensing data, thereby facilitating the fusion and analysis with the remote sensing image data, and helping the model to more accurately understand and analyze the ground feature in the remote sensing image, thereby improving the overall performance and predictive ability of the model.

[0105] In one embodiment, step S4 performs inversion monitoring based on input data using a convolution model to output multiple meteorological parameter prediction values, including:

[0106] S41, splicing the remote sensing image to be input and the multi-dimensional meteorological data in the input data in the channel dimension through a convolution model to obtain a spliced ​​image;

[0107] In the embodiment of the present invention, please refer to Figure 2 , which is a schematic diagram of the structure of a convolution model provided by an embodiment of the present invention. The method for constructing the convolution model includes:

[0108] The remote sensing data and meteorological information label sample sets are used as training data, and the meteorological forecast data is used as output data for model training. When the model converges, the convolutional model is obtained.

[0109] In an embodiment of the present invention, the convolution model can mine the information features between meteorological data related to the environment based on the input remote sensing image data and meteorological information label sample set, and realize the inversion of information changes such as wind speed data, temperature data and humidity data based on remote sensing images.

[0110] In an embodiment of the present invention, a method for generating a stitched image may include:

[0111] S411, input size is H×W×C image The multi-channel remote sensing image is the above-mentioned remote sensing image to be input, where H and W are the height and width (number of pixels) of the image, and C image is the number of image channels, which may include the calculated band ratio value and the normalized backscattering coefficient value of the optical image;

[0112] In an embodiment of the present invention, the calculated value of the wavelength ratio of the optical image can be calculated using the following formula:

[0113]

[0114] Among them, CB is the band ratio of the optical image, B i To enhance the target band of specific ground features; B j B is the reference band for comparison with the target band; k It is a normalized band, usually a band related to surface material characteristics.

[0115] The number of channels in the image of the present invention includes the calculated value of the band ratio of the optical image. The band ratio of the optical image can effectively reveal the spectral differences between different types of ground objects, especially in the area around the transmission tower, and can further highlight the spectral characteristics of the tower and the surrounding ground objects. Through this band calculation method, the distribution characteristics of the ground objects in the environment around the transmission tower in the area can be enhanced, and their recognizability in the image can be improved.

[0116] S412, input size is H×W×C weather Multi-channel meteorological image, C weather The number of channels of the image includes multi-dimensional meteorological data, including temperature data, humidity data, and wind speed data that are aligned with the image in time and space.

[0117] S413, stitching the multi-channel remote sensing image and the multi-channel meteorological image in the channel dimension to obtain a stitched image X, which is expressed as H×W×(C image +C weather ).

[0118] S42, using multiple convolution kernels to extract multi-scale features in the spliced ​​image to obtain a shared feature map vector;

[0119] In the embodiment of the present invention, the spliced ​​image is input to the convolution kernel, and information of different scales of the spliced ​​image is extracted to obtain a shared feature map vector X conv , the expression is:

[0120] Xconv =Conv2D(X,kernel size=s)

[0121] Among them, s∈{3,5,7,…} represents different convolution kernel sizes, which are used to capture channel features at different scales. By introducing adaptive convolution kernels, the convolution operation is dynamically adjusted so that it can adaptively select convolution kernels of different sizes or shapes according to different areas or features of the input image.

[0122] S43, performing a local pooling operation along the rows or columns of the shared feature map vector to obtain a pooled feature map;

[0123] In the embodiment of the present invention, stripe pooling can be used to perform local pooling operations. Stripe pooling can further enhance the model's perception of specific stripe (or linear) features between the convolution layer and the output branch. By performing local pooling operations along the row or column direction of the image, the spatial structure vector of the linear feature is retained, and the pooled feature map X is obtained. stripe-pool , the expression is:

[0124] X stripe-pool = StripePool(X conv )

[0125] The present invention retains these directional features by applying stripe pooling on the feature map after the shared convolution layer, which helps the model better capture the linear or striped structures that may exist in remote sensing images, thereby effectively improving the prediction accuracy of meteorological parameters.

[0126] S44. Input the graph vector obtained by flattening the pooled feature map into the fully connected layer, and output multiple meteorological parameter prediction values ​​through the fully connected layer.

[0127] In this embodiment of the present invention, it is assumed that the vector after the feature map after convolution and pooling is flattened is y f , input to the fully connected layer (FC), the fully connected layer maps the extracted features into the final output vector y f . Then we have:

[0128] y f =FC(X stripe-pool )

[0129] See also Figure 3 In the embodiment of the present invention, multiple output branches can be used to predict each meteorological parameter. Each branch corresponds to a corresponding meteorological parameter channel and has an independent output channel to achieve joint prediction of multiple meteorological parameters. That is, for different meteorological parameters, different fully connected layers are selected for prediction. The expression is:

[0130] y temp =FC1(X stripe-pool )

[0131] y wet =FC2(X stripe-pool )

[0132] y wind =FC3(X stripe-pool )

[0133] Among them, y temp 、y wet 、y wind are independent prediction results, these parameters share the features extracted by the convolution and pooling layers, y temp 、y wet 、y wind They represent the predicted values ​​of wind speed data, temperature data and humidity data in the target area of ​​the transmission tower respectively.

[0134] In an embodiment of the present invention, the prediction error of each meteorological parameter can be weighted and summed to calculate the total loss function to optimize all tasks. The total loss function is expressed as follows:

[0135] L all =λ1·L temp +λ2·L wet +λ3·L wind

[0136] Among them, L all is the total loss function, L temp 、L wet 、L wind are the loss functions for the predicted values ​​of temperature data, humidity data, and wind speed data respectively; λ1, λ2, and λ3 are weighting coefficients used to adjust the importance of each task in the total loss.

[0137] By using multiple convolution kernels to extract multi-scale features from the stitched image, the embodiment of the present invention enables the model to capture feature information at different levels and scales, thereby comprehensively understanding the image content and effectively improving the efficiency and accuracy of feature extraction. The shared feature map vector contains these multi-scale features, providing rich information for subsequent analysis. Furthermore, through local pooling operations, the model can further extract and compress important information in the feature map vector, reducing the data dimension while retaining key features, which is conducive to improving the efficiency of transmission tower environmental inversion monitoring.

[0138] In one embodiment, after step S4, outputting a plurality of meteorological parameter forecast values, the method further includes:

[0139] S5. Determine the degree of similarity between the predicted value and the observed value of the meteorological parameter. When the degree of similarity is higher than a preset threshold, determine that the model inversion result is good.

[0140] The embodiment of the present invention uses the mean absolute error (MAE) and the index of agreement (IA) to measure the difference between the predicted value and the measured value. The calculation formulas of MAE and IA are as follows:

[0141]

[0142] Where N is the number of data samples, P i is the predicted value at sample i (i.e., the predicted value of meteorological parameters), O i is the observation value at sample i, MAE can quantify the size of its average error, P′ i and O′ i They represent the normalized predicted value and observed value respectively; IA characterizes the similarity between the predicted value and the observed value of the meteorological parameter, and the measurement range is between 0 and 1. The closer it is to 1, the higher the correlation and the better the model inversion result. For example, when the IA value is higher than 0.7, the model inversion result is determined to be good.

[0143] In one embodiment, after performing inversion monitoring based on input data using a convolution model and outputting multiple meteorological parameter prediction values ​​in step S4, the following steps may also be included:

[0144] The predicted value of meteorological parameter y temp 、y wet 、y wind As a risk factor, calculate the comprehensive risk index factor R total :

[0145] R total =ω temp ·y temp +ω wet ·y wet +ω wind ·y wind

[0146] Among them, ω temp 、ω wet 、ω wind are the corresponding risk factor weights for each risk factor, and ω temp +ω wet +ω wind =1, since wind speed can directly affect the safety of transmission tower lines, and humidity and temperature factors are more stable than changes caused by wind speed, the wind speed weight is given as 0.6, the temperature weight is 0.2, and the humidity weight is 0.2.total Threshold division is performed, as shown in Table 1, and areas with green disaster warning levels are selected as reference areas for wind tower site selection.

[0147] Table 1 Comprehensive risk index evaluation table

[0148]

[0149] See also Figure 4 In one embodiment, another flow chart of a method for inverse monitoring of a transmission tower environment is provided. Figure 4 The process includes data preparation, feature extraction, model inversion and risk prediction. In the data preparation step, ERA5 data, Landsat data, and Sentinel-1 data are collected. ERA5 data includes wind speed, humidity, and temperature data, Landsat data includes multi-band optical data, and Sentinel-1 data includes VH polarization data and VV polarization data. In the feature extraction step, the data obtained in the data preparation step are preprocessed, including radiation correction, atmospheric correction, image registration, geometric correction, and geocoding. Building masks are then extracted from the preprocessed data, including the calculation of building polarization index and building spectral index, and threshold segmentation is performed. In the model inversion stage, model inversion is performed based on the remote sensing image data after extracting the building mask, including a multidimensional ecological environment monitoring sample annotation dataset. The dataset is divided into a training set and a test set. The training set is used for model training to construct the MSDAC model, and parameter iteration and model optimization are performed. When the iteration number condition is reached, the test set is input to obtain the accuracy assessment results of the multidimensional ecological environment monitoring results around the transmission tower. In the risk prediction stage, the risk level distribution around the transmission tower is predicted and a transmission tower site selection recommendation is generated.

[0150] The implementation of the present invention has the following beneficial effects:

[0151] The embodiment of the present invention combines optical remote sensing images and dual-polarization remote sensing images to extract building mask information of the target area, removes the building mask information in the dual-polarization remote sensing images and the optical remote sensing images, and obtains the remote sensing image to be input. By combining two different types of remote sensing data, the building mask information of the transmission tower in a complex environment is extracted, which can effectively avoid the accuracy error caused by the building mask information as the misjudgment area, thereby effectively improving the environmental inversion monitoring effect; and the processed remote sensing image and meteorological data are input into a pre-built convolution model, and the model is used for inversion monitoring, and a comprehensive plurality of meteorological parameter prediction values ​​are output, thereby further improving the effect of the transmission tower environmental inversion monitoring.

[0152] Furthermore, the embodiment of the present invention adjusts the meteorological data to the same resolution as the remote sensing image, which can ensure the spatial consistency of the meteorological data and the remote sensing data, thereby facilitating fusion and analysis with the remote sensing image data, and helping the model to more accurately understand and analyze the characteristics of the land objects in the remote sensing image, thereby improving the overall performance and predictive ability of the model.

[0153] See also Figure 5 Based on the same inventive concept as the above embodiment, one embodiment of the present invention provides a transmission tower environment inversion monitoring device, comprising:

[0154] A multi-source data acquisition module 10 is used to acquire multi-source data of a target area of ​​a transmission tower, wherein the multi-source data includes optical remote sensing images, dual-polarization remote sensing images, and meteorological data;

[0155] The module 20 for acquiring the remote sensing image to be input is used to extract the building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and remove the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain the remote sensing image to be input;

[0156] The inversion monitoring module 30 is used to input the remote sensing image and meteorological data to be input as input data into a pre-built convolution model, perform inversion monitoring based on the input data through the convolution model, and output multiple meteorological parameter prediction values.

[0157] In one embodiment, extracting building mask information of a target area based on an optical remote sensing image and a dual-polarization remote sensing image includes:

[0158] Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization;

[0159] Selecting a band based on the optical remote sensing image to determine the band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information;

[0160] The building mask information of the target area is determined based on the normalized backscattering coefficient and band spectral information.

[0161] In one embodiment, determining building mask information of a target area based on normalized backscatter coefficient and band spectrum information includes:

[0162] The polarization building index is determined according to the normalized backscattering coefficient, and the spectral building index is determined according to the band spectrum information. The building mask information of the target area is obtained by weighted accumulation based on the polarization building index and the spectral building index.

[0163] In one embodiment, the remote sensing image and meteorological data to be input are used as input data and input into a pre-built convolutional model, including:

[0164] Perform outlier detection on meteorological data and remove outliers in meteorological data;

[0165] After removing outliers, the meteorological data is adjusted to spatial grid data with the same resolution as the remote sensing image, and the spatial grid data is determined as multi-dimensional meteorological data;

[0166] The multi-dimensional meteorological data and the remote sensing image to be input are used as input data.

[0167] In one embodiment, a convolution model is used to perform inversion monitoring based on input data and output multiple meteorological parameter prediction values, including:

[0168] The remote sensing image to be input and the multi-dimensional meteorological data in the input data are spliced ​​in the channel dimension through the convolution model to obtain a spliced ​​image;

[0169] Multiple convolution kernels are used to extract multi-scale features in the spliced ​​image and obtain a shared feature map vector;

[0170] Perform local pooling operations along the rows or columns of the shared feature map vector to obtain the pooled feature map;

[0171] The graph vector obtained by flattening the pooled feature map is input into the fully connected layer, and multiple meteorological parameter prediction values ​​are output through the fully connected layer.

[0172] In one embodiment, after outputting the plurality of meteorological parameter forecast values, the method further includes:

[0173] The degree of similarity between the predicted values ​​and the observed values ​​of the meteorological parameters is determined. When the degree of similarity is higher than a preset threshold, the model inversion result is determined to be good.

[0174] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present invention, and can implement any one of the above-mentioned method embodiments of the present invention to provide a transmission tower environment inversion monitoring method.

[0175] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Furthermore, in the drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which may be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement the present invention without inventive effort.

[0176] Based on the above-mentioned embodiment of the transmission tower environment inversion monitoring method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the transmission tower environment inversion monitoring method of any embodiment of the present invention is implemented.

[0177] For example, in this embodiment, the computer program may be divided into one or more modules, one or more of which are stored in a memory and executed by a processor to implement the present invention. One or more module elements may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0178] The terminal device can be a computing device such as a desktop computer, notebook computer, PDA, or cloud server. The terminal device may include, but is not limited to, a processor and memory.

[0179] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0180] On the basis of the above-mentioned method embodiments, another embodiment of the present invention provides a computer-readable storage medium, comprising a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the transmission tower environment inversion monitoring method of any one of the above-mentioned method embodiments of the present invention.

[0181] Among them, the module / unit integrated in the device / terminal equipment, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0182] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for inversion monitoring of transmission tower environment, characterized in that: include: Acquiring multi-source data of a target area of ​​a transmission tower, wherein the multi-source data includes optical remote sensing images, dual-polarization remote sensing images, and meteorological data; extracting building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and removing the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain a remote sensing image to be input; Inputting the remote sensing image to be input and the meteorological data as input data into a pre-built convolutional model; The convolution model is used to perform inversion monitoring based on the input data and output a plurality of meteorological parameter prediction values.

2. The transmission tower environment inversion monitoring method according to claim 1, characterized in that: The extracting the building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image includes: Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization; Selecting a band based on the optical remote sensing image to determine band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information; Building mask information of the target area is determined according to the normalized backscatter coefficient and the band spectrum information.

3. The transmission tower environment inversion monitoring method according to claim 2, characterized in that: The determining of the building mask information of the target area according to the normalized backscatter coefficient and the band spectrum information includes: A polarization building index is determined according to the normalized backscatter coefficient, a spectral building index is determined according to the band spectrum information, and weighted accumulation is performed based on the polarization building index and the spectral building index to obtain building mask information of the target area.

4. The transmission tower environment inversion monitoring method according to claim 1, characterized in that: The step of inputting the remote sensing image to be input and the meteorological data as input data into a pre-built convolutional model includes: performing outlier detection on the meteorological data and removing outliers in the meteorological data; Adjusting the meteorological data after removing outliers to spatial grid data with the same resolution as the remote sensing image, and determining the spatial grid data as multi-dimensional meteorological data; The multi-dimensional meteorological data and the remote sensing image to be input are used as input data.

5. The transmission tower environment inversion monitoring method according to claim 4, characterized in that: The inversion monitoring is performed according to the input data by the convolution model to output a plurality of meteorological parameter prediction values, including: splicing the remote sensing image to be input in the input data and the multi-dimensional meteorological data in the channel dimension through the convolution model to obtain a spliced ​​image; Using multiple convolution kernels to extract multi-scale features in the spliced ​​image to obtain a shared feature map vector; Performing a local pooling operation along the rows or columns of the shared feature map vector to obtain a pooled feature map; The graph vector obtained by flattening the pooled feature graph is input into a fully connected layer, and multiple meteorological parameter prediction values ​​are output through the fully connected layer.

6. The transmission tower environment inversion monitoring method according to claim 1, characterized in that: After outputting multiple meteorological parameter forecast values, it also includes: The degree of similarity between the predicted value and the observed value of the meteorological parameter is determined, and when the degree of similarity is higher than a preset threshold, the model inversion result is determined to be good.

7. A transmission tower environment inversion monitoring device, characterized in that: include: A multi-source data acquisition module is used to acquire multi-source data of the target area of ​​the transmission tower, wherein the multi-source data includes optical remote sensing images, dual-polarization remote sensing images and meteorological data; a module for acquiring a remote sensing image to be input, configured to extract building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image, and remove the building mask information from the dual-polarization remote sensing image and the optical remote sensing image to obtain a remote sensing image to be input; A data input module, configured to input the remote sensing image to be input and the meteorological data as input data into a pre-built convolution model; The inversion monitoring module performs inversion monitoring based on the input data through the convolution model and outputs a plurality of meteorological parameter prediction values.

8. The transmission tower environment inversion monitoring device according to claim 7, characterized in that: The extracting the building mask information of the target area based on the optical remote sensing image and the dual-polarization remote sensing image includes: Converting the dual-polarized remote sensing image into a backscatter coefficient, and normalizing the backscatter coefficient to obtain a normalized backscatter coefficient; wherein the normalized backscatter coefficient includes a normalized backscatter coefficient of VV polarization and a normalized backscatter coefficient of VH polarization; Selecting a band based on the optical remote sensing image to determine band spectrum information, wherein the band spectrum information includes near-infrared band spectrum information and mid-infrared band spectrum information; Building mask information of the target area is determined according to the normalized backscatter coefficient and the band spectrum information.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for inversion monitoring of the transmission tower environment according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that include: A stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the transmission tower environment inversion monitoring method according to any one of claims 1 to 6.

Images