A substation intelligent detection method and satellite image processing system

Abstract

The present invention relates to the technical field of intelligent operation and inspection of power transmission lines, and in particular to an intelligent detection method for substations and a satellite image processing system. The method comprises: collecting target data information, and constructing a power station detection data set by acquiring a historical multidimensional feature data set, establishing a multi-level feature extraction network to perform multi-target feature extraction on the target data information, performing feature fusion on the multi-target feature mapping by setting a void convolution module, constructing a power station comprehensive detection model based on the multi-level feature extraction network and the void convolution module, and performing power station detection on the target data information through the power station comprehensive detection model to obtain power station detection results. The present invention effectively solves the problem that traditional detection methods are easily interfered with by factors such as weather and terrain in complex environments, resulting in unstable recognition effect and low accuracy, effectively improves the fusion capability of multi-level features, and thus ensures high-precision recognition of substations and their equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and inspection of power transmission lines, and in particular to an intelligent detection method for a substation and a satellite image processing system. Background Art

[0002] With the rapid development of science and technology, electricity has become an indispensable part of modern life and industrial production. The existence of substations has greatly improved the flexibility and reliability of the power system. By adjusting the voltage levels in different regions, substations can flexibly allocate power resources according to actual needs, effectively avoiding large-scale power outages caused by supply and demand mismatches. This function is crucial for ensuring basic electricity needs for residents in urban and remote areas. Therefore, ensuring the safe and stable operation of substations is particularly important.

[0003] However, substations are currently usually located in outdoor environments, and there may be various obstructions such as trees and buildings around them. They are also affected by factors such as light changes and weather conditions. Existing detection methods often find it difficult to accurately distinguish substations from other similar objects, resulting in inaccurate data information and posing a huge safety hazard to the operation and maintenance of transmission lines.

[0004] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present disclosure and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] The present invention provides a substation intelligent detection method and a satellite image processing system, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A substation intelligent detection method, the method comprising:

[0008] Collecting target data information and obtaining a historical multidimensional feature data set, and constructing a power plant detection data set based on the historical multidimensional feature data set;

[0009] Establishing a multi-level feature extraction network, performing multi-target feature extraction on the target data information according to the multi-level feature extraction network to obtain a multi-target feature map, and setting a hole convolution module to perform feature fusion on the multi-target feature map to obtain comprehensive power station features;

[0010] Building a comprehensive power plant detection model based on the multi-level feature extraction network and the dilated convolution module, and training the comprehensive power plant detection model based on the power plant detection dataset;

[0011] The power station detection is performed on the target data information according to the trained power station comprehensive detection model to obtain a power station detection result.

[0012] Furthermore, performing multi-target feature extraction on the target data information according to the multi-level feature extraction network to obtain a multi-target feature map includes:

[0013] constructing a power plant feature database based on the historical multidimensional feature data set, performing feature extraction on the target data information based on the power plant feature database to obtain a plurality of preliminary feature maps, wherein the preliminary feature maps include a plurality of low-level feature maps and a plurality of deep-level feature maps;

[0014] Establishing residual skip paths based on the plurality of preliminary feature maps, respectively, wherein the residual skip paths are used to directly link low-level feature maps with deep-level feature maps;

[0015] A cross-layer feature transfer weight is set according to the plurality of preliminary feature maps, a jump parameter is set according to the residual jump path, and the cross-layer feature transfer weight is dynamically adjusted based on the jump parameter to obtain a multi-target feature map.

[0016] Furthermore, dynamically adjusting the cross-layer feature transfer weight based on the jump parameter includes:

[0017] Calculating feature similarity between the low-layer feature map and the deep-layer feature map based on the power station feature database, and setting initial cross-layer feature transfer weights based on the feature similarity;

[0018] Calculating a cross-layer influence factor according to the initial cross-layer feature transfer weight, and modifying the jump parameter according to the cross-layer influence factor;

[0019] The cross-layer feature transfer weights are redistributed based on the modified skip parameters.

[0020] Furthermore, power station detection is performed on target data information according to the trained power station comprehensive detection model, including:

[0021] Extracting a multi-level feature map of the target data information, wherein a low-level feature map includes texture details of the power equipment and a high-level feature map includes a geometric outline of the power equipment;

[0022] Performing feature weighted integration on the low-level feature map and the high-level feature map to obtain an enhanced feature map;

[0023] Scanning a local area of the enhanced feature map to determine a device area, and extracting physical features and spatial coordinates of the power equipment in the device area;

[0024] The historical equipment relationship of the power station is obtained, the consistency check of the power equipment in the equipment area is performed according to the historical equipment relationship of the power station, and the positioning information of the power station facilities is output.

[0025] Furthermore, a comprehensive detection model for power plants is constructed, including:

[0026] Matching and comparing the comprehensive power station features according to the power station facility positioning information to obtain a spatial correspondence between the power station facility positioning information and the comprehensive power station features;

[0027] If the power station facility location information does not match the comprehensive power station feature, then calculate the feature contribution weight based on the power station facility location information, and modify the category distribution of the comprehensive power station feature according to the feature contribution weight to match the power station facility location information, wherein the feature contribution weight is allocated according to the number of pixels of the regional category;

[0028] If the power station facility positioning information matches the comprehensive power station characteristics, the power station facility positioning information and the comprehensive power station characteristics are integrated to construct a comprehensive power station detection model.

[0029] Furthermore, obtaining a power station detection result and performing image optimization on the power station detection result include:

[0030] Acquire a false detection image and a standard image according to the power station detection result, reshape the features of the false detection image, and verify the standard image;

[0031] Performing multiple image detail optimizations on the misdetected image to obtain a number of optimized comparison images of different degrees;

[0032] Comparing the plurality of optimized comparison images with the false detection images respectively, and marking the feature difference areas of the false detection images;

[0033] The characteristic difference region is merged with the power station detection dataset to update the power station detection dataset.

[0034] Furthermore, the standard image is verified, including:

[0035] Collecting historical record coordinates, and extracting the centroid coordinates of the power station area according to the standard image, comparing the historical record coordinates with the centroid coordinates of the power station area to obtain a plane distance difference;

[0036] An offset threshold is set according to the historical equipment relationship of the power station. If the plane distance difference exceeds the offset threshold, a secondary verification is performed on the standard image to obtain a corrected offset;

[0037] The standard image is corrected according to the corrected offset and calibrated again until the plane distance difference meets the offset threshold.

[0038] Furthermore, the power plant comprehensive detection model is trained according to the power plant detection data set, including:

[0039] Cleaning the historical multidimensional feature data set to obtain an original training set;

[0040] Performing data information processing on the plurality of historical multidimensional feature data sets according to the original training set to construct a power station detection data set, wherein the data information processing includes target recognition and manual correction;

[0041] The power station detection data set is divided into a power station training set and a power station test set, the power station comprehensive detection model is trained according to the power station training set, and the power station comprehensive detection model is verified according to the power station test set.

[0042] A satellite image processing system, comprising:

[0043] A data information construction module, which uses remote sensing satellites to collect target data information and obtain historical multidimensional feature data sets, and constructs a power station detection data set based on the historical multidimensional feature data sets;

[0044] A feature extraction and fusion module establishes a multi-level feature extraction network, performs multi-target feature extraction on the target data information according to the multi-level feature extraction network, obtains a multi-target feature map, and sets a hole convolution module to perform feature fusion on the multi-target feature map to obtain comprehensive power station features;

[0045] A recognition model construction training module is used to construct a comprehensive power plant detection model based on the multi-level feature extraction network and the dilated convolution module, and to train the comprehensive power plant detection model according to the power plant detection dataset;

[0046] The power station identification and output module performs power station detection on the target data information according to the trained power station comprehensive detection model to obtain power station detection results.

[0047] Furthermore, the feature extraction and fusion module includes:

[0048] a power plant feature extraction and analysis unit, which constructs a power plant feature database based on the historical multidimensional feature data set, performs feature extraction on the target data information based on the power plant feature database, and obtains a plurality of preliminary feature maps, wherein the preliminary feature maps include a plurality of low-level feature maps and a plurality of deep-level feature maps;

[0049] A residual connection feature transfer unit is configured to establish a residual skip path based on the plurality of preliminary feature maps, wherein the residual skip path is used to directly link the low-level feature map with the deep-level feature map;

[0050] A cross-layer weight dynamic adjustment unit sets a cross-layer feature transfer weight according to the plurality of preliminary feature maps, sets a jump parameter according to the residual jump path, and dynamically adjusts the cross-layer feature transfer weight based on the jump parameter to obtain a multi-target feature map.

[0051] The technical solution of the present invention can achieve the following technical effects:

[0052] It effectively solves the problem that traditional detection methods are easily interfered by factors such as weather and terrain in complex environments, resulting in unstable recognition effects and low accuracy. By combining a multi-level feature extraction network with a dilated convolution module, the fusion capability of multi-level features is effectively improved, thereby ensuring high-precision recognition of substations and their equipment. At the same time, the training and optimization of the deep learning model enable the present invention to operate stably in complex environments, significantly improving the accurate positioning of power station facilities and meeting the needs of high-precision recognition.

[0053] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 This is a flow chart of the intelligent detection method for substations;

[0056] Figure 2 Schematic diagram of the multi-target feature map extraction process;

[0057] Figure 3 Schematic diagram of the process of dynamically adjusting weights for cross-layer feature transfer;

[0058] Figure 4 Optimize the schematic diagram for power station test results;

[0059] Figure 5 Schematic diagram of satellite image processing system. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0061] 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 invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0062] Embodiment 1;

[0063] like Figure 1 As shown, the present application provides a substation intelligent detection method, the method comprising:

[0064] S100: Using remote sensing satellites to collect target data information and obtain historical multidimensional feature datasets, and constructing power station detection datasets based on the historical multidimensional feature datasets;

[0065] S200: Establishing a multi-level feature extraction network, performing multi-target feature extraction on target data information according to the multi-level feature extraction network, obtaining a multi-target feature map, and setting a dilated convolution module to perform feature fusion on the multi-target feature map to obtain comprehensive power station features;

[0066] S300: Build a comprehensive power plant detection model based on a multi-level feature extraction network and a dilated convolution module, and train the model based on the power plant detection dataset.

[0067] S400: Performing power station detection on target data information according to the trained power station comprehensive detection model to obtain power station detection results.

[0068] Specifically, first, remote sensing satellites can be used to obtain data information of the target area. Remote sensing satellites can obtain geographic and power infrastructure data of a large area, including the location of substations, surrounding environment, distribution information of obstacles such as buildings and trees. After collecting the target data information, the system integrates these data with historical data sets, and constructs a power station detection data set for subsequent detection based on the historical multidimensional feature data set. This data set contains multi-dimensional features of the substation, such as power facility type, relative position, surrounding environment type, etc., and constructs a multi-level feature extraction network. The network can extract multiple target features from the target data information, including structural features of the substation, surrounding environment features, etc. This feature extraction network combines deep learning technology to gradually extract features from low levels to high levels through multi-layer convolutional networks. Then, the extracted multi-target features are mapped by setting a void convolution module. For feature fusion, the hole convolution module can capture target features of different sizes at different sampling intervals, which enhances the adaptability of the network to substations of different scales. This fusion process can effectively integrate the key features of the substation and its surrounding environment, eliminate noise interference in the environment, and construct a comprehensive detection model for power stations based on multi-target feature mapping and combined with the feature fusion results of the hole convolution module. The goal of this model is to automatically identify substations in target data through training, and accurately distinguish substations from surrounding buildings, trees and other similar objects. During the training process, the model can continuously adjust parameters to minimize the error function and gradually improve the detection accuracy. Through the trained comprehensive detection model of the power station, the target data information can be used to perform substation detection, and finally obtain the detection results of the substation. This detection result includes the specific location of the substation, the description of the surrounding environment and related power facility information.

[0069] The technical solution of the present invention effectively solves the problem that traditional detection methods are easily interfered by factors such as weather and terrain in complex environments, resulting in unstable recognition effects and low accuracy. By combining a multi-level feature extraction network with a dilated convolution module, the fusion capability of multi-level features is effectively improved, thereby ensuring high-precision recognition of substations and their equipment. At the same time, the training and optimization of the deep learning model enable the present invention to operate stably in complex environments, significantly improving the accurate positioning of power station facilities and meeting the needs of high-precision recognition.

[0070] Further, if Figure 2 As shown, multi-target feature extraction is performed on target data information according to a multi-level feature extraction network to obtain a multi-target feature map, including:

[0071] S210: constructing a power plant feature database based on the historical multidimensional feature data set, performing feature extraction on the target data information based on the power plant feature database, and obtaining a number of preliminary feature maps, wherein the preliminary feature maps include a number of low-level feature maps and a number of deep-level feature maps;

[0072] S220: establishing residual skip paths based on the plurality of preliminary feature maps, respectively, where the residual skip paths are used to directly link the low-layer feature maps with the deep-layer feature maps;

[0073] S230: setting cross-layer feature transfer weights according to a number of preliminary feature maps, setting jump parameters according to the residual jump path, dynamically adjusting the cross-layer feature transfer weights based on the jump parameters, and obtaining multi-target feature maps.

[0074] As a preferred embodiment of the above embodiment, first, a power station feature database is constructed based on a historical multidimensional feature data set. The database contains remote sensing data information of various power station facilities, and annotates different power equipment, facility layout and structural information. Subsequently, the constructed power station feature database is used to extract features of the target data information. The preliminary feature map is divided into a low-level feature map and a deep-level feature map. The low-level feature map mainly captures detail information (such as the texture and color of the equipment), while the deep-level feature map focuses on higher-level spatial relationships (such as the geometric shape and arrangement of the equipment). Based on the preliminary feature map, a residual jump path is established. The residual jump path is used to directly link the low-level feature map and the deep feature map to solve the loss problem in information transmission. The residual jump path can be established in the following manner: In some embodiments, the low-level features and the deep features are merged through an addition formula. The addition formula is: ,in, Represents low-level features, Represents deep features, Represents the merged feature map; the low-level feature map and the deep feature map are fused through residual connections. The residual connection ensures that the low-level features are directly passed to the deep network without multiple convolution and pooling operations. This connection method helps the network maintain attention to low-level details at a deeper level; then a weight coefficient is introduced to control the contribution of low-level and deep features in the fusion process. Initially, a fixed cross-layer transfer weight is set, and then these weights are dynamically adjusted according to the training data. During the training process, the weight of cross-layer feature transfer can be dynamically adjusted through the backpropagation algorithm. All optimized features are weighted fused to obtain the final multi-target feature map. This feature map will contain comprehensive information about multiple power facilities (such as substations, transformers, switchgear, etc.) in the remote sensing image.

[0075] Furthermore, if Figure 3 As shown in Figure 2, the cross-layer feature transfer weights are dynamically adjusted based on the jump parameters, including:

[0076] S231: Calculating feature similarity between the low-layer feature map and the deep-layer feature map based on the power station feature database, and setting initial cross-layer feature transfer weights based on the feature similarity;

[0077] S232: Calculate a cross-layer influence factor based on the initial cross-layer feature transfer weight, and modify the jump parameter based on the cross-layer influence factor;

[0078] S233: Redistribute the cross-layer feature transfer weights based on the corrected skip connection parameters.

[0079] In this embodiment, a low-level feature map and a deep-level feature map of the target data information are obtained through a feature extraction network (e.g., a convolutional neural network). Subsequently, the spatial and semantic similarity between the low-level feature map and the deep-level feature map is evaluated by calculating the similarity between the two. The following similarity calculation method can be used: , Represents low-level features, Represents deep features, represents the dot product operation, represents the norm of a vector; based on the calculated feature similarity, initial cross-layer feature transfer weights are assigned to low-level and deep-level features. The cross-layer influence factor is then calculated and the skipping parameters are modified. The cross-layer influence factor quantifies the strength of the relationship between low-level and deep-level features. Backpropagation is used to calculate the gradient of the loss function with respect to the features at each layer of the network. During backpropagation, the chain rule is used to calculate the gradient layer by layer. The magnitude of the gradient reflects the sensitivity of the feature to the loss, that is, the impact of feature changes on the final output. A large gradient indicates that the feature at that layer has a greater impact on the final prediction; conversely, a small gradient indicates that the feature at that layer has a smaller impact. The calculated gradient size can be used to define the cross-layer influence factor, which quantifies the relative influence of low-level and deep-level features on the final output. The skipping parameters are then modified based on these cross-layer influence factors. Throughout training, the cross-layer influence factor is continuously adjusted based on the backpropagated gradient information. After each training iteration, the model recalculates the gradients of low-level and deep-level features and adjusts the cross-layer influence factor and skipping parameters accordingly.

[0080] Furthermore, the trained power plant comprehensive detection model is used to perform power plant detection on the target data information, including:

[0081] Extracting a multi-level feature map of target data information, wherein the low-level feature map contains texture details of the power equipment and the high-level feature map contains the geometric outline of the power equipment;

[0082] Perform feature weighted integration on the low-level feature map and the high-level feature map to obtain an enhanced feature map;

[0083] Scan the enhanced feature map in a local area to determine the equipment area, and extract the physical features and spatial coordinates of the power equipment in the equipment area;

[0084] Obtain the historical equipment relationship of the power station, perform consistency check on the power equipment in the equipment area based on the historical equipment relationship of the power station, and output the location information of the power station facilities.

[0085] Specifically, by acquiring target data information, the target recognition algorithm is used to identify the target data information, and the improved Yolov8 network can be used to build a target recognition model: EfficientNet is used as the backbone network for feature extraction. EfficientNet can extract low-level features in the image through its efficient convolutional layer design, such as the texture and local details of the power equipment. Through multiple convolutional layers, local information of the power equipment such as surface texture, details, edges, etc. is extracted. The deep part of the YOLOv8 network uses deep convolutional layers and Transformer Blocks to enhance the global feature capture capability, extract the geometric contours and spatial relationships of the power equipment, and then use BiFPN (Bidirectional Feature Pyramid The YOLOv8 network performs weighted integration of low-level and high-level features, combining them to generate an enhanced feature map through weighted methods. Based on the fused enhanced feature map, the target detection module of the YOLOv8 network is used to perform local area scanning to identify the specific location and area of the power equipment. Finally, the historical equipment relationship data of the power station is used to associate the spatial location, equipment type, facility layout and other information between devices. Based on the historical equipment relationship database, the detected equipment area is checked for consistency. The verification process includes verifying the type, location, size and other information of the equipment to ensure the consistency of the identification results with the historical data. After the consistency check, the correct location and type of the power equipment are finally determined, and the power station facility positioning information, including the specific coordinates, type, function, etc. of the equipment, is output.

[0086] Furthermore, a comprehensive power plant detection model is constructed, including:

[0087] According to the power station facility positioning information, the comprehensive power station characteristics are matched and compared to obtain the spatial correspondence between the power station facility positioning information and the comprehensive power station characteristics;

[0088] If the power station facility location information does not match the comprehensive power station features, the feature contribution weight is calculated based on the power station facility location information, and the category distribution of the comprehensive power station features is corrected according to the feature contribution weight to match the power station facility location information. The feature contribution weight is allocated according to the number of pixels in the regional category.

[0089] If the power station facility positioning information matches the comprehensive power station characteristics, the power station facility positioning information and the comprehensive power station characteristics are integrated to construct a comprehensive power station detection model.

[0090] As a preferred embodiment of the above, first, the location of the power station facilities (such as substations, transformers, switchgear, etc.) is marked on the collected target data information and matched with the comprehensive power station features (such as global features and local features extracted based on deep learning). A feature embedding method can be used to project the power station facility positioning information and the power station features into the same embedding space. In this shared space, the similarity of the features will be directly measured. When the power station facility positioning information does not match the comprehensive power station features, it is necessary to further adjust the comprehensive power station features. The category distribution of the comprehensive power station features can be corrected by feature contribution weights. This can be done in the following way: based on the number of pixels in the area where the power station facilities are located (i.e., the size of the area or the occupied area), and the distribution of different categories of power equipment in the area, the contribution weight of each category feature is calculated, and the category distribution of the comprehensive power station features is corrected according to the feature contribution weights. For each category, the feature contribution weight can be obtained by the ratio of the number of pixels in the category to the total number of pixels. When the power station facility positioning information matches the comprehensive power station features, the two are merged to construct a complete comprehensive power station detection model. This model combines the physical characteristics, spatial coordinates and comprehensive power station characteristics of power station facilities. It can also adopt feature-level fusion method to merge low-level and high-level features through weighted fusion, splicing or addition operations. After the above steps, the fused comprehensive feature model of the power station is obtained, completing the precise positioning and identification of the power station facilities.

[0091] Furthermore, if Figure 4 As shown, the power station detection results are obtained and image optimization is performed on the power station detection results, including:

[0092] Obtain false positive images and standard images based on the power station inspection results, reshape the features of the false positive images, and verify the standard images;

[0093] Perform multiple image detail optimizations on the misdetected image to obtain several optimized comparison images of different degrees;

[0094] Compare several optimized comparison images with the false detection images respectively, and mark the feature difference areas of the false detection images;

[0095] The feature difference areas are merged with the power station detection dataset to update the station detection dataset.

[0096] In this embodiment, based on the power station detection results, false positive images and standard images are obtained. False positive images are incorrectly identified during the model recognition process, which may include incorrectly identified equipment areas or category errors. Standard images are images obtained based on the annotation information of real power station facilities (such as the actual location and category of power station equipment). The false positive images and standard images can be compared using image difference analysis methods, and image quality assessment methods such as structural similarity indicators or mean square error are used for comparison to find the false positive areas. Subsequently, the image quality can be optimized through multiple Gaussian blurring and sharpening processes: for false positive images, the Gaussian blur standard deviation is selected based on historical experience, and a Gaussian kernel is constructed based on the standard deviation. And calculate the weight value of each pixel point, then convolve the Gaussian kernel with the false detection image, traverse each pixel point in the image, use Gaussian weights for weighted averaging, and output the blurred image; after Gaussian blurring, perform sharpening filtering on the blurred image, and in at least one embodiment, use the Laplace operator to perform the sharpening operation, and through repeated optimization, obtain multiple optimized comparison images, which differ in the degree of detail restoration and can include comparison results of high-quality optimization, medium-quality optimization, and low-quality optimization: the optimization process can be processed through a hierarchical optimization strategy, and each time optimization is performed, the network hyperparameters (such as learning rate, batch size, etc.) can be adjusted to control the optimization effect. Through multiple optimizations, contrast images at different levels are obtained. The multiple optimized contrast images are compared with the original false detection images one by one, and the characteristic difference areas of the false detection images are marked. The difference areas refer to the obvious differences between the optimized image and the false detection image. These areas usually contain part of the false detection information and can be more accurately identified after optimization. The comparison process can be carried out through pixel-level difference comparison; the characteristic difference areas marked in the false detection images can be merged with the power station detection dataset using an incremental learning method. The updated dataset will contain more optimization information, especially through learning and improvement in error recognition, thereby providing more training data for the model.

[0097] Furthermore, the standard image is calibrated, including:

[0098] Collect historical record coordinates, extract the centroid coordinates of the power station area based on the standard image, compare the historical record coordinates with the centroid coordinates of the power station area, and obtain the plane distance difference;

[0099] An offset threshold is set based on the historical equipment relationship of the power station. If the plane distance difference exceeds the offset threshold, a secondary verification is performed on the standard image to obtain the corrected offset.

[0100] The standard image is corrected according to the correction offset and recalibrated until the plane distance difference meets the offset threshold.

[0101] As a preferred embodiment of the above, historical record coordinates are first collected using a geographic information system (GIS) or manual calibration. Image segmentation techniques (such as threshold segmentation and edge detection) can be used to accurately determine the power station area. The center of mass of the area is then calculated. The center of mass can be calculated by taking a weighted average of the coordinates of all pixels in the image. The historical record coordinates are compared with the coordinates of the power station area center of mass, and the planar distance difference between them is calculated. The planar distance difference can be calculated using the Euclidean distance formula. An initial offset threshold can then be set based on historical experience. A calibration test can be conducted on some power stations to verify whether the offset threshold is appropriate. If excessive errors are found to be uncorrected, the offset threshold may need to be lowered. If the errors are small, the threshold can be appropriately increased. If the planar distance difference exceeds the offset threshold, a secondary calibration is performed. During this process, the location of the power station area is re-examined to obtain a corrected offset. The power station area can be moved by the calculated corrected offset through image translation to match the historical record coordinates. Verification is then performed using specific landmarks in the image (such as the specific location of the transformer or equipment identification) to ensure that the corrected power station location is more accurate. After obtaining the corrected offset, the standard image is corrected according to the corrected offset. After the correction is completed, the plane distance difference is recalculated and re-calibrated. The process is iterated until the plane distance difference meets the offset threshold.

[0102] Furthermore, the power plant comprehensive detection model is trained based on the power plant detection dataset, including:

[0103] Clean the historical multidimensional feature dataset to obtain the original training set;

[0104] According to the original training set, several historical multi-dimensional feature data sets are processed to construct a power station detection data set. The data processing includes target recognition and manual correction.

[0105] The power plant detection dataset is divided into a power plant training set and a power plant test set. The power plant comprehensive detection model is trained based on the power plant training set, and verified based on the power plant test set.

[0106] In this embodiment, before starting model training, the historical multidimensional feature dataset needs to be cleaned. For images of poor quality, image enhancement techniques (such as Gaussian blur denoising and contrast enhancement) can be used to remove noise and improve image quality. At the same time, the accuracy of the annotation of each image is checked. If an annotation error is found (such as incorrectly annotating the location of power station facilities), it needs to be manually corrected. If some images cannot clearly identify power station facilities or the annotated area is too small, they can be removed from the dataset. Then, an object recognition algorithm (feature pyramid network (FPN)) is used to identify objects in each image and annotate the locations of power station facilities (for example, transformers, switchgear, substation buildings, etc.). For power station equipment that is more difficult to identify, the diversity of the training set can be increased by enhancing the data (such as image flipping, rotation, and cropping) to obtain the final training set. The final training dataset is divided into the dataset in a ratio of 6:4 to obtain the processed training set and test set, and the comprehensive power station detection model is trained and verified.

[0107] Embodiment 2;

[0108] Based on the same inventive concept as the satellite image processing system in the aforementioned embodiment, the present invention also provides a satellite image processing system, such as Figure 5 As shown, the system includes:

[0109] The data information construction module uses remote sensing satellites to collect target data information and obtain historical multidimensional feature data sets, and constructs a power station detection data set based on the historical multidimensional feature data sets;

[0110] The feature extraction and fusion module establishes a multi-level feature extraction network, performs multi-target feature extraction on the target data information based on the multi-level feature extraction network, obtains multi-target feature mapping, and sets a void convolution module to perform feature fusion on the multi-target feature mapping to obtain comprehensive power station features;

[0111] Identification model construction training module, based on the multi-level feature extraction network and the void convolution module to build a comprehensive power plant detection model, and the power plant comprehensive detection model is trained according to the power plant detection data set;

[0112] The power station identification and output module performs power station detection on the target data information based on the trained power station comprehensive detection model to obtain the power station detection results.

[0113] The above-mentioned adjustment system in the present invention can effectively implement the intelligent detection method of the substation, and the technical effects that can be achieved are as described in the above-mentioned embodiments, which will not be repeated here.

[0114] Specifically, the feature extraction and fusion module includes:

[0115] The power plant feature extraction and analysis unit constructs a power plant feature database based on the historical multi-dimensional feature data set, extracts features from the target data information based on the power plant feature database, and obtains several preliminary feature maps, which include several low-level feature maps and several deep-level feature maps;

[0116] The residual connection feature transfer unit establishes residual skip paths based on several preliminary feature maps. The residual skip paths are used to directly link low-level feature maps with deep feature maps.

[0117] The cross-layer weight dynamic adjustment unit sets the cross-layer feature transfer weight according to several preliminary feature maps, sets the jump parameters according to the residual jump path, and dynamically adjusts the cross-layer feature transfer weight based on the jump parameters to obtain multi-target feature maps.

[0118] Similarly, the above-mentioned optimization schemes for the system can also respectively achieve the corresponding optimization effects of the method in Example 1, which will not be repeated here.

[0119] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and drawings are merely illustrative of the present application as defined herein and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, the present application is intended to include such modifications and variations as fall within the scope of the present application and its equivalents.

Claims

1. A substation intelligent detection method, characterized in that: The method comprises: Collect target data information and obtain a historical multidimensional feature data set, and construct a power plant detection data set based on the multidimensional feature data set; Establishing a multi-level feature extraction network, performing multi-target feature extraction on the target data information according to the multi-level feature extraction network to obtain a multi-target feature map, and setting a hole convolution module to perform feature fusion on the multi-target feature map to obtain comprehensive power station features; Building a comprehensive power plant detection model based on the multi-level feature extraction network and the dilated convolution module, and training the comprehensive power plant detection model according to the power plant detection dataset, including: Matching and comparing the comprehensive power station features according to the power station facility positioning information to obtain a spatial correspondence between the power station facility positioning information and the comprehensive power station features; If the power station facility location information does not match the comprehensive power station feature, then calculate the feature contribution weight based on the power station facility location information, and modify the category distribution of the comprehensive power station feature according to the feature contribution weight to match the power station facility location information, wherein the feature contribution weight is allocated according to the number of pixels of the regional category; If the power station facility location information matches the comprehensive power station characteristics, the power station facility location information and the comprehensive power station characteristics are integrated to construct a comprehensive power station detection model; Performing power station detection on the target data information according to the trained power station comprehensive detection model, obtaining a power station detection result, and performing image optimization on the power station detection result, including: Acquire a false detection image and a standard image according to the power station detection result, reshape the features of the false detection image, and verify the standard image; Performing multiple image detail optimizations on the misdetected image to obtain a number of optimized comparison images of different degrees; Comparing the plurality of optimized comparison images with the false detection images respectively, and marking the feature difference areas of the false detection images; merging the characteristic difference region with the power station detection dataset to update the station detection dataset; The standard image is calibrated, including: Collecting historical record coordinates, and extracting the centroid coordinates of the power station area according to the standard image, comparing the historical record coordinates with the centroid coordinates of the power station area to obtain a plane distance difference; An offset threshold is set according to the historical equipment relationship of the power station. If the plane distance difference exceeds the offset threshold, a secondary verification is performed on the standard image to obtain a corrected offset; The standard image is corrected according to the corrected offset and calibrated again until the plane distance difference meets the offset threshold.

2. The intelligent detection method for substation according to claim 1, characterized in that: Performing multi-target feature extraction on the target data information according to the multi-level feature extraction network to obtain a multi-target feature map includes: constructing a power plant feature database based on the historical multidimensional feature data set, performing feature extraction on the target data information based on the power plant feature database to obtain a plurality of preliminary feature maps, wherein the preliminary feature maps include a plurality of low-level feature maps and a plurality of deep-level feature maps; Establishing residual skip paths based on the plurality of preliminary feature maps, respectively, wherein the residual skip paths are used to directly link low-level feature maps with deep-level feature maps; A cross-layer feature transfer weight is set according to the plurality of preliminary feature maps, a jump parameter is set according to the residual jump path, and the cross-layer feature transfer weight is dynamically adjusted based on the jump parameter to obtain a multi-target feature map.

3. The intelligent detection method for substation according to claim 2, characterized in that: Dynamically adjusting the cross-layer feature transfer weight based on the skip parameter includes: Calculating feature similarity between the low-layer feature map and the deep-layer feature map based on the power station feature database, and setting initial cross-layer feature transfer weights based on the feature similarity; Calculating a cross-layer influence factor according to the initial cross-layer feature transfer weight, and modifying the jump parameter according to the cross-layer influence factor; The cross-layer feature transfer weights are redistributed based on the modified skip parameters.

4. The intelligent detection method for substation according to claim 1, characterized in that: Performing power station detection on target data information according to the trained power station comprehensive detection model includes: Extracting a multi-level feature map of the target data information, wherein a low-level feature map includes texture details of the power equipment and a high-level feature map includes a geometric outline of the power equipment; Performing feature weighted integration on the low-level feature map and the high-level feature map to obtain an enhanced feature map; Scanning a local area of the enhanced feature map to determine a device area, and extracting physical features and spatial coordinates of the power equipment in the device area; The historical equipment relationship of the power station is obtained, the consistency check of the power equipment in the equipment area is performed according to the historical equipment relationship of the power station, and the positioning information of the power station facilities is output.

5. The intelligent detection method for substation according to claim 1, characterized in that: Training the power plant comprehensive detection model according to the power plant detection data set includes: Cleaning the historical multidimensional feature data set to obtain an original training set; Performing data information processing on the plurality of historical multidimensional feature data sets according to the original training set to construct a power station detection data set, wherein the data information processing includes target recognition and manual correction; The power station detection data set is divided into a power station training set and a power station test set, the power station comprehensive detection model is trained according to the power station training set, and the power station comprehensive detection model is verified according to the power station test set.

6. A satellite image processing system, characterized in that: According to the substation intelligent detection method as claimed in claim 1, the system includes: A data information construction module, which uses remote sensing satellites to collect target data information and obtain historical multidimensional feature data sets, and constructs a power station detection data set based on the historical multidimensional feature data sets; A feature extraction and fusion module establishes a multi-level feature extraction network, performs multi-target feature extraction on the target data information according to the multi-level feature extraction network, obtains a multi-target feature map, and sets a hole convolution module to perform feature fusion on the multi-target feature map to obtain comprehensive power station features; A recognition model construction training module is used to construct a comprehensive power plant detection model based on the multi-level feature extraction network and the dilated convolution module, and to train the comprehensive power plant detection model according to the power plant detection dataset; The power station identification and output module performs power station detection on the target data information according to the trained power station comprehensive detection model to obtain power station detection results.

7. The satellite image processing system according to claim 6, characterized in that: The feature extraction and fusion module includes: a power plant feature extraction and analysis unit, which constructs a power plant feature database based on the historical multidimensional feature data set, performs feature extraction on the target data information based on the power plant feature database, and obtains a plurality of preliminary feature maps, wherein the preliminary feature maps include a plurality of low-level feature maps and a plurality of deep-level feature maps; A residual connection feature transfer unit is configured to establish a residual skip path based on the plurality of preliminary feature maps, wherein the residual skip path is used to directly link the low-level feature map with the deep-level feature map; A cross-layer weight dynamic adjustment unit sets a cross-layer feature transfer weight according to the plurality of preliminary feature maps, sets a jump parameter according to the residual jump path, and dynamically adjusts the cross-layer feature transfer weight based on the jump parameter to obtain a multi-target feature map.

Images