Adaptive robust detection method and system for corrosion of power transmission tower in complex environment

Abstract

The present application belongs to the technical field of image data processing, and provides a transmission tower corrosion adaptive robust detection method and system under complex environment, which acquires a synthetic image dataset generated under multi-dimensional complex environment; trains an environment-adaptive robust detection model based on the synthetic image dataset to obtain a trained environment-adaptive robust detection model; detects a to-be-detected image based on the trained environment-adaptive robust detection model to obtain a corrosion area detection result; registers the detection results obtained in different periods, and performs change detection based on the registered detection results to obtain a change detection result; and inputs the obtained change detection result sequence data into the trained time series prediction model to predict the future corrosion expansion rate and direction. The present application realizes the leap from static diagnosis to predictive maintenance.

Description

Technical Field

[0001] This invention belongs to the field of image data processing technology, and in particular relates to an adaptive robust detection method and system for corrosion of transmission towers under complex environments. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Rust detection of steel components on transmission towers is a crucial step in ensuring the safe operation of the power grid. However, existing visual inspection methods exhibit significant instability under different environmental conditions, severely impacting their accuracy.

[0004] These methods typically quantify the degree of corrosion by segmenting images of component regions and projecting pixel colors to a specific color space based on a fixed threshold. While effective, this method has fundamental limitations: first, the fixed threshold is difficult to adapt to dynamically changing environments, such as rain, fog, shadows, seasonality, and differences in day and night lighting, and has weak anti-interference capabilities; second, it lacks temporal modeling of the corrosion evolution process, making it impossible to analyze the path of pitting initiation, expansion, and even the evolution into cracks, thus failing to achieve predictive early warning. Summary of the Invention

[0005] To address at least one of the technical problems mentioned above, this invention provides an adaptive robust detection method and system for corrosion of transmission towers in complex environments. This method not only improves the robustness and accuracy of corrosion detection in complex environments, but also achieves a leap from static diagnosis to predictive maintenance.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides an adaptive robust detection method for corrosion of transmission towers under complex environments, comprising the following steps: Obtain a dataset of synthetic images generated in a multi-dimensional and complex environment; The trained environment-adaptive robust detection model is obtained by training an environment-adaptive robust detection model based on a synthetic image dataset. The trained environment-adaptive robust detection model is used to detect the image to be detected and obtain the rust region detection result. The detection results from different periods are registered, and the change detection results are obtained by performing change detection based on the registered detection results. The acquired transformation detection result sequence data is input into the trained time-series prediction model to predict the future corrosion expansion rate and direction.

[0007] Furthermore, the acquisition of the synthetic image dataset generated in a multi-dimensional complex environment includes: Construct 3D models of the transmission tower and its various components; Based on the constructed 3D model, the physical materials in the model are defined to ensure the realism of the interaction between light and shadow; Generate various environmental conditions, including different weather, different lighting, different viewing angles, and occlusion conditions; Based on the various environmental conditions generated, a large amount of diverse synthetic image data is produced, along with pixel-level labels.

[0008] Furthermore, the construction process of the environment-adaptive robust detection model includes: Multi-level feature maps are extracted from synthetic image data based on a constructed multi-scale feature pyramid. Based on each feature level and combined with the structural characteristics of the transmission tower, the structural guidance weight of the transmission tower's structural features is calculated; By performing environment-invariant feature learning on the features at each feature level, the essential material features are separated from the environmental interference features to obtain the essential material features. By combining the structural guidance weights of the transmission tower's structural features with the material essence features of each feature level, adaptive feature fusion is performed to obtain the fused multi-scale features; Based on the fused multi-scale features, corrosion area prediction is performed to obtain the output corrosion probability map.

[0009] Furthermore, during the training of the environment-adaptive robust detection model, an environment classifier is introduced as an adversarial network to improve the model's environment invariance. The training loss function is expressed as: , , in, Indicates detection loss, Indicates the balance hyperparameters, Indicates environmental damage and loss. Indicates the first i Layer features, Indicates the essential characteristics of the material. This represents the environment discriminator. This refers to a material essential feature extractor.

[0010] Further, the step of inputting the acquired transformation detection result sequence data into the trained time-series prediction model to predict the future corrosion propagation rate and direction includes: Key evolutionary features are extracted from the temporal change detection results to construct a corrosion evolution feature descriptor; Based on the corrosion evolution feature descriptor, spatiotemporal evolution modeling is performed to obtain the corrosion evolution feature sequence; Multi-scale trend analysis was performed on key evolutionary features to separate long-term trend components and short-term fluctuation components, and then a trend feature vector was constructed. The corrosion propagation rate and main propagation direction are predicted based on the corrosion evolution feature sequence and trend feature vector.

[0011] Furthermore, the key evolutionary features extracted include the rate of change of rust area, the rate of rust perimeter expansion, the rust centroid movement vector, the rust density change, and the shape complexity factor.

[0012] A second aspect of the present invention provides an adaptive robust detection system for corrosion of transmission towers under complex environments, comprising: The data acquisition module is used to acquire synthetic image datasets generated in multi-dimensional complex environments. The adaptive robust detection module is used to train an environment-adaptive robust detection model based on a synthetic image dataset to obtain a trained environment-adaptive robust detection model; and to detect the rusted area based on the trained environment-adaptive robust detection model in the image to be detected. The transformation detection module is used to register the detection results acquired at different times, and then perform change detection based on the registered detection results to obtain the transformation detection result; The corrosion prediction module is used to input the acquired transformation detection result sequence data into the trained time series prediction model to predict the future corrosion expansion rate and direction.

[0013] A third aspect of the present invention provides a computer-readable storage medium.

[0014] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the adaptive robust detection method for corrosion of transmission towers under complex environments as described above.

[0015] A fourth aspect of the present invention provides a computer device.

[0016] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described above.

[0017] A fifth aspect of the present invention provides a program product.

[0018] A program product, which is a computer program product, includes a computer program that, when executed by a processor, implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described above.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a complex environmental dataset covering multiple climates, lighting conditions, and weather conditions. A corrosion prediction model is trained based on this dataset, learning to remove environmental interference factors during training to achieve an essential understanding of the component's material condition and improve its adaptability to the environment. Furthermore, based on acquired time-series image data from regular inspections, and combining this data with the constructed time-series model, the invention analyzes the corrosion propagation rate and morphological evolution, enabling modeling and prediction of key propagation paths such as pitting corrosion to cracks. This not only improves the robustness and accuracy of corrosion detection in complex environments but also represents a leap from static diagnosis to predictive maintenance.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of the adaptive robust detection method for corrosion of transmission towers under complex environments provided in the embodiments of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Traditional methods for detecting corrosion of steel components in transmission towers, as mentioned in the background section, typically involve segmenting the component area image and projecting pixel colors onto a specific color space based on a fixed threshold to quantify the degree of corrosion. While effective, this method has fundamental limitations: the fixed threshold is difficult to adapt to dynamically changing environments and has weak anti-interference capabilities; it lacks temporal modeling of the corrosion evolution process, making it impossible to analyze the path of pitting corrosion initiation, propagation, and even the evolution into cracks, thus failing to achieve predictive early warning.

[0027] This invention proposes a novel robust corrosion detection and evolution analysis method that integrates physically based rendering synthetic data with temporal modeling. First, a complex environmental dataset encompassing multiple climates, lighting conditions, and weather conditions is constructed. A corrosion prediction model is trained based on this dataset, learning to remove environmental interference factors during training to achieve an essential perception of the component's material state and improve its adaptability to the environment. Furthermore, based on acquired temporal image data from regular inspections, the method combines the temporal image data with the constructed temporal model to analyze the corrosion propagation rate and morphological evolution, achieving modeling and prediction of key propagation paths such as pitting corrosion to cracks. This not only improves the robustness and accuracy of corrosion detection in complex environments but also represents a leap from static diagnosis to predictive maintenance.

[0028] Example 1 like Figure 1 As shown, this embodiment provides an adaptive robust detection method for corrosion of transmission towers in complex environments, including the following steps: Step 1: Obtain the synthetic image dataset generated in a multi-dimensional complex environment; Specifically, the steps include the following: Step 101: Construct 3D models of the transmission tower and its various components; In this embodiment, 3D models of each component (including main materials, diagonal materials, bolts, anchor bolts, stiffening plates, etc.) are created using Revit software and then assembled into a 3D model of the overall structure of the transmission tower.

[0029] Step 102: Based on the constructed 3D model, define the physical materials in the model to ensure the realism of the interaction between light and shadow; In this embodiment, the material color and texture are defined based on the characteristics of steel using Unreal Engine (UE Virtual Engine). Step 103: Generate various environmental conditions, including different weather, different lighting, different viewing angles, and occlusion conditions; In this embodiment, the UE virtual engine can construct an environment system through multiple components such as sky and atmosphere, clouds, lighting, post-processing volume, exponential height fog, and particle system, including different environmental conditions such as sky, atmosphere, clouds, fog, lighting, wind and rain. Step 104: Generate massive and varied synthetic image data based on the various environmental conditions generated during the construction, and attach pixel-level labels; This invention addresses the problem of weak anti-interference ability of traditional fixed thresholds in adapting to dynamically changing environments by first constructing a complex environmental dataset covering multiple climates, lighting conditions, and weather conditions.

[0030] Step 2: Train the environment-adaptive robust detection model based on the synthetic image dataset to obtain the trained environment-adaptive robust detection model. Then, use the trained environment-adaptive robust detection model to detect the image to be detected and obtain the rust region detection results. Specifically, the steps include the following: Step 201: Extract multi-level feature maps from the synthetic image data based on the constructed multi-scale feature pyramid; In this embodiment, the multi-scale pyramid is constructed using the Swin Transformer model, and the corresponding first feature map, second feature map, third feature map and fourth feature map are obtained after four levels of hierarchical processing. Let the original input image be... Feature maps are obtained after four levels of hierarchical processing. ,in: , , , , Where C1, C2, C3, and C4 represent the number of feature channels. Indicates the height of the feature map. Indicates the width of the feature map; Step 202: Based on each feature level and combined with the structural characteristics of the transmission tower, calculate the structural guidance weight of the transmission tower's structural features; In this embodiment, the formula for calculating the structure-guided weight of each feature level is as follows: , in, Indicates structure-oriented weights, Indicates the first i Layer features, This represents the spatial attention mechanism. This represents the sigmoid function; [;] indicates channel concatenation. Step 203: Perform environment-invariant feature learning on the features at each feature level to separate the essential material features from the environmental interference features, and obtain the essential material features; Material essential characteristics Represented as: , , Here, ⊙ represents element-wise multiplication. An environmental mask used to suppress interference from factors such as light and weather.

[0031] Step 204: Combine the structural guidance weights of the transmission tower's structural features with the material essence features of each feature level to perform adaptive feature fusion and obtain the fused multi-scale features; An adaptive feature fusion mechanism is used to integrate multi-scale features, which is represented as: , , in, For the fused multi-scale features, and The i-th and j-th learnable scale weight parameters are optimized through backpropagation.

[0032] Step 205: Based on the fused multi-scale features, predict the corrosion area to obtain the output corrosion probability map and bounding box regression parameters; In this embodiment, based on the FCOS detection head, the fused features are used to predict the corrosion region, and the final corrosion probability map is represented as follows: , The bounding box regression parameters are expressed as follows: , in, This represents a corrosion probability diagram. This represents the bounding box regression parameters.

[0033] In this embodiment, when training the environment-adaptive robust detection model, an environment classifier is introduced as an adversarial network to improve the model's environment invariance. The training loss function is expressed as follows: , , in, Indicates detection loss, Indicates the balance hyperparameters, Indicates environmental damage and loss. This represents the environment discriminator. This represents a material essential feature extractor, used to extract essential features from... i Material characteristics such as rust can be separated from the layers.

[0034] It should be noted that, The environment discriminator is a classification network whose goal is to determine the environment category of an image as accurately as possible based on the material features of the input, such as sunny or rainy weather. This represents a material-intrinsic feature extractor, which is part of an environment-adaptive robust detection model. Its goal is to generate features that can "fool" the environment discriminator. The characteristic is to extract essential features that are unaffected by the environment and are only related to the material (such as corrosion). .

[0035] Step 3: Register the detection results from different periods, and perform change detection based on the registered detection results to obtain the transformation detection results; Specifically, the steps include the following: Step 301: High-precision registration of multi-period data. Geometric registration is performed on the detection results acquired at different times to eliminate geometric distortion between images. To eliminate geometric distortions between images from different time periods and provide a basis for pixel-level alignment for change detection, the following registration process needs to be implemented: Feature extraction and matching: SIFT is used to extract robust feature points (such as corner points and edges) from the two images. Transformation model estimation: Based on successfully matched feature point pairs, the RANSAC algorithm is used to filter out mismatched points and estimate an optimal projection transformation model to describe the spatial relationship between the two images. Image resampling: Based on the estimated optimal projection transformation model, the image to be registered (usually called the moving image) is resampled (usually bilinear or cubic convolution interpolation) and geometrically corrected to a coordinate system consistent with the reference image (active image).

[0036] Step 302: Change detection analysis. Based on the registered detection results, change detection is performed to obtain a binarized change detection result image, specifically including: Generate difference map: For the two registered images, the spectral feature differences are calculated pixel by pixel using the image difference method; Analysis of the difference map: The difference map is divided into two categories, "changed" and "unchanged", by K-means threshold segmentation; the gray value of the difference map represents the degree of change, and the higher the gray value, the more significant the change.

[0037] Post-processing optimization: Morphological algorithms are used to post-process the preliminary change detection results to remove small particle noise and fill holes, finally obtaining a smooth, continuous and accurate binary change detection result image.

[0038] Step 4: Input the acquired transformation detection result sequence data into the trained time series prediction model to predict the future corrosion expansion rate and direction.

[0039] Specifically, the steps include the following: Step 401: Extract key evolution features from the temporal change detection results and construct a corrosion evolution feature descriptor. ; In this embodiment, the key evolutionary features extracted The factors including the rate of change of rusted area, the rate of rust perimeter expansion, the rust centroid movement vector, the rust density change, and the shape complexity factor are expressed as: ; Among them, the rate of change of the area of ​​rusted region The calculation formula is: , Corrosion perimeter expansion rate The calculation formula is: , Corrosion centroid movement vector The calculation formula is: , Rust density variation The calculation formula is: , Shape complexity factor The calculation formula is: , in, and These represent the areas of the corroded region at the current moment and the previous moment, respectively. and These are the perimeters of the corroded areas at the current and previous moments, respectively. and These represent the difference in centroid coordinates of the corroded region at the current and previous moments, respectively. and Δt represents the corrosion density at the current moment and the previous moment, respectively, and Δt is the time interval.

[0040] Step 402: Based on the corrosion evolution feature descriptor, perform spatiotemporal evolution modeling to obtain the corrosion evolution feature sequence; In this embodiment, a time-series prediction model is used for spatiotemporal evolution modeling. Specifically, a time-series prediction model is constructed based on an LSTM-Transformer hybrid architecture, including: The corrosion evolution feature descriptor sequence is input into the LSTM unit to calculate the hidden state. , is represented as: ; The hidden state is transformed linearly to obtain the query matrix Q, the key matrix K, and the value matrix V. Calculate attention weights: ,in, The feature dimension scaling factor; Calculate the attention-weighted output corrosion evolution feature sequence: .

[0041] Step 403: Perform multi-scale trend analysis on key evolutionary features to separate the long-term trend component and the short-term fluctuation component, and further construct the trend feature vector; Specifically, the formula for calculating the long-term trend component is: ,in, It is a moving average function. This is the long-term window size; The formula for calculating the short-term volatility component is: , The formula for calculating the trend eigenvector is: , in, It is a first-order difference operator. ² represents the second-order difference operator.

[0042] Step 404: Attention-weighted output With trend eigenvectors The expansion rate and prediction confidence are predicted. Specifically, the attention-weighted output With trend eigenvectors After splicing, the data is input into a multilayer perceptron (MLP) and represented as follows: , The overall expansion rate is calculated and expressed as: , The prediction confidence level is calculated and expressed as: , in, To predict the rate of area change, Perimeter expansion rate , Represents the weight matrix; Step 405: Attention-weighted output With trend eigenvectors Construct a directional probability distribution model to predict the main expansion direction of corrosion; Calculate the historical expansion direction based on the historical centroid movement vector: , The attention-weighted output With trend eigenvectors Input direction prediction multilayer perceptron The dominant direction of the prediction is obtained. and directional distribution standard deviation ; Dominant direction based on prediction and directional distribution standard deviation The constructed directional probability distribution model predicts the main direction of corrosion expansion; Construct a directional probability distribution model: , in, The dominant direction represents the angle at which corrosion is most likely to spread, and the direction is predicted by the multilayer perceptron. Output, The standard deviation of the directional distribution represents the degree of dispersion in the direction of expansion; the larger the value, the more uncertain the direction. Indicating in the dominant direction and directional distribution standard deviation Below, the probability density of corrosion propagation direction at angle θ; During the training phase, negative log-likelihood (NLL) loss is typically used to optimize the algorithm. and This maximizes the probability of observed historical directions under this distribution.

[0043] Of course, other models can also be used for the directional probability distribution model in other embodiments, such as the von Mises distribution, depending on the situation. Step 406: Use Kalman filtering to smooth and optimize the predicted spread rate and spread direction to obtain the optimized spread rate and spread direction.

[0044] In this embodiment, the specific optimization process is the existing process and will not be described in detail. Step 406 introduces Kalman filtering to incorporate the static probability prediction results obtained in step 405 into a dynamic system estimation framework. By utilizing the temporal continuity of corrosion expansion, the direct prediction values ​​with noise are smoothed and corrected in a Bayesian optimal manner. Finally, a set of more stable and reliable optimized expansion rates and directions is output, providing higher quality input for subsequent corrosion area growth simulation or early warning.

[0045] Step 5: Calculate the corrosion risk index by combining the information on the rate of spread and the direction, and assess the corrosion risk level; The formula for calculating the corrosion risk index is: , The formula for calculating the corrosion risk level is:

[0046] Where β1, β2, and β3 are weighting coefficients, and τ1 and τ2 are risk thresholds.

[0047] Example 2 This embodiment provides an adaptive robust detection system for corrosion of transmission towers in complex environments, including: The data acquisition module is used to acquire synthetic image datasets generated in multi-dimensional complex environments. The adaptive robust detection module is used to train an environment-adaptive robust detection model based on a synthetic image dataset to obtain a trained environment-adaptive robust detection model; and to detect the rusted area based on the trained environment-adaptive robust detection model in the image to be detected. The transformation detection module is used to register the detection results acquired at different times, and then perform change detection based on the registered detection results to obtain the transformation detection result; The corrosion prediction module is used to input the acquired transformation detection result sequence data into the trained time series prediction model to predict the future corrosion expansion rate and direction.

[0048] It should be noted that the specific implementation of the adaptive robust detection system for transmission tower corrosion in complex environments in this embodiment of the invention is similar to the specific implementation of the adaptive robust detection method for transmission tower corrosion in complex environments in this embodiment of the invention. For details, please refer to the description in the method section. To reduce redundancy, it will not be repeated here.

[0049] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the adaptive robust detection method for transmission tower corrosion under complex environments as described above.

[0050] Example 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described above.

[0051] Example 5 This embodiment provides a program product, which is a computer program product, including a computer program. When the computer program is executed by a processor, it implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described above.

[0052] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0053] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0054] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0055] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0056] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive and robust detection method for corrosion of transmission towers under complex environments, characterized in that, include: Obtain a dataset of synthetic images generated in a multi-dimensional and complex environment; The trained environment-adaptive robust detection model is obtained by training an environment-adaptive robust detection model based on a synthetic image dataset. The construction process of the environment-adaptive robust detection model includes: Multi-level feature maps are extracted from synthetic image data based on a constructed multi-scale feature pyramid. Based on each feature level and combined with the structural characteristics of the transmission tower, the structural guidance weight of the transmission tower's structural features is calculated; By performing environment-invariant feature learning on the features at each feature level, the essential material features are separated from the environmental interference features to obtain the essential material features. By combining the structural guidance weights of the transmission tower's structural features with the material essence features of each feature level, adaptive feature fusion is performed to obtain the fused multi-scale features; Based on the fused multi-scale features, the corrosion area is predicted to obtain the output corrosion probability map; The trained environment-adaptive robust detection model is used to detect the image to be detected and obtain the rust region detection result. The detection results from different periods are registered, and the change detection results are obtained by performing change detection based on the registered detection results. The acquired transformation detection result sequence data is input into the trained time-series prediction model to predict the future corrosion expansion rate and direction.

2. The adaptive robust detection method for corrosion of transmission towers under complex environments as described in claim 1, characterized in that, The acquisition of the synthetic image dataset generated in a multi-dimensional complex environment includes: Construct 3D models of the transmission tower and its various components; Based on the constructed 3D model, the physical materials in the model are defined to ensure the realism of the interaction between light and shadow; Generate various environmental conditions, including different weather, different lighting, different viewing angles, and occlusion conditions; Based on the various environmental conditions generated, a large amount of diverse synthetic image data is produced, along with pixel-level labels.

3. The adaptive robust detection method for corrosion of transmission towers under complex environments as described in claim 1, characterized in that, During training, the environment-adaptive robust detection model incorporates an environment classifier as an adversarial network to improve its environment invariance. The training loss function is expressed as: ， ， in, Indicates detection loss, Indicates the balance hyperparameters, Indicates environmental damage and loss. Indicates the first i Layer features, Indicates the essential characteristics of the material. This represents the environment discriminator. This refers to a material essential feature extractor.

4. The adaptive robust detection method for corrosion of transmission towers under complex environments as described in claim 1, characterized in that, The step of inputting the acquired transformation detection result sequence data into the trained time-series prediction model to predict the future corrosion propagation rate and direction includes: Key evolutionary features are extracted from the temporal change detection results to construct a corrosion evolution feature descriptor; Based on the corrosion evolution feature descriptor, spatiotemporal evolution modeling is performed to obtain the corrosion evolution feature sequence; Multi-scale trend analysis was performed on key evolutionary features to separate long-term trend components and short-term fluctuation components, and then a trend feature vector was constructed. The corrosion propagation rate and main propagation direction are predicted based on the corrosion evolution feature sequence and trend feature vector.

5. The adaptive robust detection method for corrosion of transmission towers under complex environments as described in claim 4, characterized in that, The key evolutionary features extracted include the rate of change of rust area, the rate of rust perimeter expansion, the rust centroid movement vector, the rust density change, and the shape complexity factor.

6. An adaptive and robust detection system for corrosion of transmission towers under complex environments, characterized in that, include: The data acquisition module is used to acquire synthetic image datasets generated in multi-dimensional complex environments. The adaptive robust detection module is used to train an environment-adaptive robust detection model based on a synthetic image dataset to obtain a trained environment-adaptive robust detection model; and to detect the rusted area based on the trained environment-adaptive robust detection model in the image to be detected. The construction process of the environment-adaptive robust detection model includes: Multi-level feature maps are extracted from synthetic image data based on a constructed multi-scale feature pyramid. Based on each feature level and combined with the structural characteristics of the transmission tower, the structural guidance weight of the transmission tower's structural features is calculated; By performing environment-invariant feature learning on the features at each feature level, the essential material features are separated from the environmental interference features to obtain the essential material features. By combining the structural guidance weights of the transmission tower's structural features with the material essence features of each feature level, adaptive feature fusion is performed to obtain the fused multi-scale features; Based on the fused multi-scale features, the corrosion area is predicted to obtain the output corrosion probability map; The transformation detection module is used to register the detection results acquired at different times, and then perform change detection based on the registered detection results to obtain the transformation detection result; The corrosion prediction module is used to input the acquired transformation detection result sequence data into the trained time series prediction model to predict the future corrosion expansion rate and direction.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described in any one of claims 1-5.

9. A program product, said program product being a computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps in the adaptive robust detection method for corrosion of transmission towers under complex environments as described in any one of claims 1-5.

Images