A method and device for identifying the progress of power transmission and transformation projects based on integrated space-air-ground technology

Abstract

This application provides a method and apparatus for identifying the progress of power transmission and transformation projects based on integrated space-air-ground technology, relating to the field of project progress management. The method includes: acquiring target image data of the construction area through multiple acquisition devices and preprocessing the target image data; extracting features from the preprocessed image data to obtain project progress feature information corresponding to the construction area; classifying the project progress feature information; matching the classified project progress feature information with a project progress feature library to output the current project progress corresponding to the construction area; and triggering a project progress warning if the deviation between the current project progress and the planned project progress is confirmed to be greater than a preset deviation value. This application solves the problem of low accuracy in identifying the progress of power transmission and transformation projects using a single monitoring technology.

Description

Technical Field

[0001] This application relates to the field of engineering progress management, and in particular to a method and device for identifying the progress of power transmission and transformation projects based on integrated space-air-ground technology. Background Technology

[0002] Traditional power transmission and transformation project progress monitoring mainly relies on manual on-site inspection records. This method consumes a lot of manpower and resources, has a long inspection cycle, cannot achieve real-time monitoring, and is prone to missing key progress nodes.

[0003] Traditional methods for identifying the progress of power transmission and transformation projects typically rely on single technologies, such as ground-based cameras, which have limited coverage and struggle to provide comprehensive coverage for complex terrain and large construction areas, resulting in limitations in information collection. Some methods utilize drone aerial photography as a single monitoring technology, but this method performs poorly when monitoring progress in shadowed areas, also limiting information acquisition. Therefore, relying on single monitoring technologies to identify the progress of power transmission and transformation projects generally suffers from low accuracy.

[0004] Therefore, there is an urgent need for methods and devices for identifying the progress of power transmission and transformation projects based on integrated air-space-ground technology. Summary of the Invention

[0005] This application provides a method and device for identifying the progress of power transmission and transformation projects based on integrated air-space-ground technology, which solves the problem of low accuracy in identifying the progress of power transmission and transformation projects when using a single monitoring technology.

[0006] The first aspect of this application provides a method for identifying the progress of power transmission and transformation projects based on integrated air-space-ground technology. The method includes: acquiring target image data of the construction area through multiple acquisition devices and preprocessing the target image data; extracting features from the preprocessed image data to obtain project progress feature information corresponding to the construction area; classifying the project progress feature information; matching the classified project progress feature information with a project progress feature library to output the current project progress corresponding to the construction area; and triggering a project progress warning if the deviation between the current project progress and the planned project progress is confirmed to be greater than a preset deviation value.

[0007] Optionally, target image data of the construction area can be acquired through multiple acquisition devices, specifically including: periodically taking pictures of the construction area using satellite remote sensing equipment to acquire satellite image data; taking aerial pictures of the construction area from multiple angles using drone equipment to acquire aerial image data; taking close-up pictures of the construction area using intelligent robot equipment to acquire complex area image data; and using satellite image data, aerial image data, and complex area image data as target image data.

[0008] Optionally, preprocessing operations are performed on the target image data, specifically including: performing radiometric correction on the satellite image data to eliminate radiometric errors, and performing geometric correction on the satellite image data to correct the satellite image data to the correct geographic coordinate system; stitching and fusion of aerial image data, and denoising the fused image data; extracting frames from complex area image data to decompose the video into single-frame images, and normalizing the size of the complex area image data and each frame image.

[0009] Optionally, feature extraction is performed on the preprocessed image data to obtain the project progress feature information corresponding to the construction area. Specifically, this includes: inputting the preprocessed target image data into a convolutional neural network model and outputting the project progress feature information based on the convolutional neural network model; and extracting advanced semantic features from the preprocessed target image data based on the project progress feature information. The advanced semantic features include substation building features and transmission line features.

[0010] Optionally, the project progress feature information is classified, specifically including: inputting the project progress feature information into a pre-trained convolutional neural network model constructed with a ResNet50 network as the backbone; connecting a classification head consisting of two fully connected layers to the back end of the ResNet50 network, wherein the first fully connected layer contains multiple neurons and uses the ReLU activation function, and the number of neurons in the second fully connected layer is consistent with the number of project progress stages; performing linear transformation processing on the project progress feature information through the first fully connected layer; and classifying the linearly transformed project progress feature information through the second fully connected layer.

[0011] Optionally, based on the classified project progress feature information, a matching process is performed using a project progress feature database to output the current project progress corresponding to the project construction area. Specifically, this includes: establishing a project progress feature database based on the construction process and characteristics of the project construction area; using the data in the project progress feature database as training data to construct a project progress recognition model; and outputting the current project progress based on the project progress recognition model.

[0012] Optionally, the current project progress is output based on the project progress recognition model, specifically including: generating a normalized probability distribution for each project progress stage using the Softmax activation function based on the project progress recognition model; performing structured processing on the normalized probability distribution and selecting the project progress corresponding to the highest probability value as the first candidate project progress; obtaining a historical project stage sequence based on historical image data and constructing a stage confidence evaluation function based on the dynamic evolution trend between the first candidate project progress and the historical project stage sequence; outputting a first confidence score corresponding to the first candidate project progress based on the stage confidence evaluation function; if it is confirmed that the first confidence score is less than a preset threshold, returning to the image acquisition step and reacquiring the target image data; outputting a second confidence score corresponding to the second candidate project progress based on the target image data through the project progress recognition model; if it is confirmed that the second confidence score is greater than or equal to a preset threshold, outputting the second candidate project progress as the current project progress.

[0013] A second aspect of this application provides a power transmission and transformation project progress identification device based on integrated air-space-ground technology. The device includes an acquisition module and a processing module, wherein...

[0014] The acquisition module is used to acquire target image data in the construction area through multiple acquisition devices and to perform preprocessing operations on the target image data.

[0015] The processing module is used to extract features from the preprocessed image data to obtain the project progress feature information corresponding to the construction area; classify the project progress feature information; match the classified project progress feature information with the project progress feature library to output the current project progress corresponding to the construction area; if it is confirmed that the progress deviation between the current project progress and the planned project progress is greater than the preset deviation value, a project progress warning is triggered.

[0016] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described above.

[0017] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which is executed by a processor using the method described in any of the foregoing descriptions.

[0018] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0019] 1. Acquire target image data in the construction area through multiple acquisition devices and preprocess the target image data; extract features from the preprocessed image data to obtain the corresponding project progress feature information of the construction area; classify the project progress feature information; based on the classified project progress feature information, match it with the project progress feature library to output the current project progress corresponding to the construction area; if it is confirmed that the progress deviation between the current project progress and the planned project progress is greater than the preset deviation value, a project progress warning is triggered, thereby realizing the accurate identification and dynamic monitoring of the progress status of the power transmission and transformation project construction area in a full-process, full-space, and automated manner. It can efficiently grasp the changes in the project stage without relying on manual inspection, improve the real-time and intelligent level of project management, promptly detect construction delays or abnormal progress, assist the project scheduling system in decision adjustment and resource optimization, ensure consistency and controllability between the construction progress and the planned target, and effectively reduce the risk of delays and construction costs.

[0020] 2. Radiometric correction is performed on satellite imagery data to eliminate radiometric errors, and geometric correction is performed to align the satellite imagery data to the correct geographic coordinate system. Aerial imagery data is stitched and fused, and the fused imagery data is denoised. Frame extraction is performed on complex area imagery data to decompose the video into single-frame images, and the size of complex area imagery data and each frame image is normalized. This constructs a multi-source target imagery dataset with a unified spatial reference, clear visual quality, and standard input format. This provides a high-quality, highly consistent input foundation for subsequent feature extraction, project progress classification, and stage identification by convolutional neural network models, ensuring the model's stability and generalization ability when processing images across scales, devices, and time periods. This improves the accuracy and robustness of progress identification and provides reliable image representation support for full-cycle project monitoring and data-driven scheduling.

[0021] 3. Based on the construction process and characteristics of the construction area, establish a project progress feature library; use the data in the project progress feature library as training data to construct a project progress recognition model, and output the current project progress based on the project progress recognition model, thereby realizing deep modeling and automatic recognition of the semantic features and progress stage information hidden in the project construction images. This enables the project progress recognition model to have end-to-end stage judgment capability. By learning the typical structural features of different construction stages in the image dimension, it can efficiently and accurately infer the current project stage status of the construction area in actual operation, reduce the reliance on manual annotation and rule design, improve the adaptability and judgment accuracy of the recognition system in complex scenes, multi-stage overlap and fuzzy boundary conditions, enhance the system's generalization ability to multi-source image input, and provide a stable and intelligent recognition foundation for dynamic perception and scheduling control of project progress. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the method for identifying the progress of power transmission and transformation projects based on integrated air-space-ground technology provided in this application embodiment;

[0023] Figure 2 This is a schematic diagram of a power transmission and transformation project progress identification device based on integrated air-space-ground technology provided in an embodiment of this application;

[0024] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0025] Explanation of reference numerals in the attached drawings: 21. Acquisition module; 22. Processing module; 301. Processor; 302. Communication bus; 303. User interface; 304. Network interface; 305. Memory. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0029] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0030] Please refer to Figure 1 The flowchart illustrates the progress identification method for power transmission and transformation projects based on integrated air-space-ground technology provided in the embodiments of this application. The flowchart mainly includes the following steps: S101 to S105.

[0031] Step S101: Acquire target image data in the construction area through multiple acquisition devices, and perform preprocessing operations on the target image data.

[0032] Specifically, target image data of the construction area is acquired through multiple acquisition devices, including satellite remote sensing equipment, drones, and intelligent robots. Different preprocessing operations are performed on the target image data according to the different acquisition devices to construct input data with analytical value.

[0033] In one possible implementation, step S101 further includes: periodically taking pictures of the construction area using satellite remote sensing equipment to obtain satellite image data; taking multi-angle aerial pictures of the construction area using drone equipment to obtain aerial image data; taking close-up pictures of the construction area using intelligent robot equipment to obtain complex area image data; and using the satellite image data, aerial image data, and complex area image data as target image data.

[0034] Specifically, the construction area is periodically photographed using satellite remote sensing equipment. This equipment, typically an imaging payload system mounted on an orbiting satellite platform, includes visible light, multispectral, or panchromatic sensors. Through periodic orbital coverage, these devices acquire wide-area, panoramic images of the construction site without requiring on-site deployment. Periodic photography involves repeatedly observing the same construction area at set time intervals, such as daily, weekly, or every ten days, to obtain time-series image data reflecting the macroscopic progress of the project. Satellite imagery data offers broad coverage, strong georeferencing capabilities, and can provide information on the overall progress of the construction area, site layout, and changes in material stacking.

[0035] Drones are used to conduct multi-angle aerial photography of the construction area. These drones, including fixed-wing and multi-rotor drones, are equipped with high-resolution optical cameras and possess the ability to flexibly perform low-altitude imaging tasks at different altitudes, angles, and paths. Multi-angle aerial photography refers to the drones acquiring images of the target area from multiple perspectives, including perpendicular, tilted, and circling directions, according to a preset mission path. This enhances the image's ability to represent three-dimensional structures and the visibility of target details. Drone mission scheduling is centrally configured by a management platform, allowing for selection of immediate execution mode for emergency shooting, scheduled execution mode for planned sampling, and periodic execution mode for continuous monitoring. Options include immediate execution, scheduled execution, and periodic execution (daily, weekly, monthly), ensuring continuous coverage of key stages and critical structures during construction.

[0036] Intelligent robotic equipment is used to take close-up photos of the construction area. This equipment refers to a work platform deployed on the construction site, capable of autonomous walking, positioning and navigation, and image acquisition. Examples include wheeled robots, tracked robots, and biomimetic gait robots. This type of equipment can enter complex areas difficult for humans to access, such as inside pipe corridors, around foundation pits, and between tower foundations, to perform high-precision, pinpoint photography of detailed areas. Close-up photography refers to the robot acquiring high-texture, high-definition local images of the area within tens of centimeters to several meters of the target structural unit using its onboard high-magnification imaging module. These images are used to identify construction details and construct feature markers.

[0037] The satellite imagery, aerial imagery, and complex area imagery acquired by the three types of equipment are combined as target imagery data to form a raw image set with spatially hierarchical distribution characteristics. This set covers multiple granular perspectives in the spatial dimension, from macroscopic (satellite), mesoscopic (drone), to microscopic (robot) dimensions. In the temporal dimension, it achieves multi-timescale fusion, from periodic monitoring (satellite) to medium-frequency inspection (drone) and real-time capture (robot), providing a comprehensive, detailed, and timely data foundation for subsequent engineering progress analysis.

[0038] In one possible implementation, step S101 further includes: performing radiometric correction on the satellite image data to eliminate radiometric errors, and performing geometric correction on the satellite image data to correct the satellite image data to the correct geographic coordinate system; stitching and fusion of the aerial image data, and denoising the fused image data; extracting frames from the complex area image data to decompose the video into single-frame images, and performing size normalization processing on the complex area image data and each frame image.

[0039] Specifically, firstly, satellite imagery data suffers from significant radiometric errors and geometric distortions due to the influence of atmospheric conditions, solar radiation angle, and sensor response differences during its acquisition process. To improve image radiometric consistency, radiometric correction is necessary. Radiometric correction includes atmospheric correction, sensor gain normalization, and surface reflectance inversion, and its mathematical expression can be:

[0040]

[0041]

[0042] in, This refers to the radiance of the top of the atmosphere. For pixel grayscale values, and These are the sensor's calibration offset and gain coefficient, respectively. For surface reflectance, The distance between the Earth and the Sun (astronomical units). This represents the solar irradiance in this wavelength band. This refers to the solar zenith angle. The above correction steps can significantly reduce image brightness deviations caused by differences in imaging time or sensor.

[0043] Geometric correction is then performed to eliminate spatial offsets caused by factors such as changes in imaging angle and orbital disturbances. By selecting Ground Control Points (GCPs) with defined spatial coordinates, a polynomial mapping relationship is constructed between the imagery and the geographic reference map. Commonly used geometric transformation models are quadratic or cubic polynomial functions.

[0044]

[0045] in, Indicates the pixel coordinates of the original image. Indicates the coordinates of the corrected image. and These are the fitting coefficients. This indicates the order of the polynomial geometric transformation. This process aligns the imagery with the GIS geographic coordinate system, ensuring spatial consistency across time-series data.

[0046] Secondly, for aerial imagery data, due to the overlapping areas and viewpoint differences caused by multi-angle and multi-trajectory imaging by drones, stitching and fusion processing is required. The stitching process first extracts key points from the image using feature matching algorithms such as SIFT (Scale Invariant Feature Transform) or SURF (Accelerated Robust Feature Transform) and calculates feature descriptors; then, feature point pairing is performed using nearest neighbor search and ratio testing, and the RANSAC algorithm is used to estimate the affine transformation matrix.

[0047]

[0048] Among them, the affine matrix parameters , , The image is obtained by fitting matching points. After geometric alignment, image fusion is performed to eliminate brightness faults in overlapping areas. Fusion methods can include pyramid fusion, multi-resolution weighted averaging, or Poisson fusion algorithms. The final output is a complete, continuous image of the construction area with a unified viewpoint.

[0049] After image fusion and stitching, local noise accumulation occurs. To improve visual clarity, denoising operations need to be performed on the images. Common methods include median filtering and Gaussian filtering. The expression for Gaussian filtering is:

[0050]

[0051] in, Represents the position of a two-dimensional Gaussian function. The weight value at that time, The standard deviation of the Gaussian kernel determines the degree of blur; noise smoothing is achieved by convolving the image with this kernel.

[0052] Finally, for image data of complex areas, since the acquisition method is often video stream data acquired by robotic equipment, frame extraction is required to meet the needs of static image processing. Frame extraction is based on a set frame rate. Extracting image frame sequences at equal intervals from the video stream , so that:

[0053]

[0054] in, This represents the time interval between two consecutive frames. Both the retrieved frame image and the original image need to be normalized to the specified input size. Commonly used methods are bilinear interpolation or region interpolation.

[0055]

[0056] in For the current interpolation calculation, the first... Line 1 Column interpolation weights This is the original image. This is the scaled image. By unifying the image resolution and aspect ratio, we ensure batch processing consistency and feature scale stability during subsequent model processing.

[0057] Step S102: Perform feature extraction on the preprocessed image data to obtain the project progress feature information corresponding to the construction area.

[0058] In one possible implementation, step S102 further includes: inputting the preprocessed target image data into a convolutional neural network model and outputting project progress feature information based on the convolutional neural network model; and extracting advanced semantic features from the preprocessed target image data based on the project progress feature information, wherein the advanced semantic features include substation building features and transmission line features.

[0059] Specifically, the convolutional neural network model employs a deep network architecture with ResNet50 as its backbone. Through multi-level convolutional layers, residual connection structures, and a global receptive field construction mechanism, it extracts features such as edges, textures, structures, and spatial distribution from the input image layer by layer. The model receives image input in a uniform format and sequentially processes it through operations such as convolutional kernel scanning, activation function mapping, pooling dimensionality reduction, and residual jump connections. Finally, it forms a compact engineering progress feature representation in the high-level feature map to capture the key semantic features of the target image in the current engineering state.

[0060] Subsequently, based on the project progress characteristics, targeted high-level semantic feature extraction was performed to further identify the typical structural differences and visual morphological changes exhibited by substation buildings and transmission lines at different construction stages. High-level semantic features refer to high-order spatial structural information and image content semantic labels that can be used to distinguish different progress states, possessing both clear project semantics and directional identification of construction stages. Specifically, the features of substation buildings at different project stages include: site leveling stage: ground texture features, color distribution features, and obstacle outline features; foundation construction stage: pit shape features, soil color and texture features, and construction machinery appearance features; civil engineering construction stage: building structure outline features, building material color and texture features, and distribution features of construction personnel and equipment; equipment installation stage: electrical equipment shape features. The features of transmission lines at different project stages include: foundation construction stage: foundation shape and size features, foundation surface texture features, and surrounding soil disturbance features; tower erection stage: tower shape features, tower color and material features, and tower erection construction scene features; and line stringing stage: visual features of conductors and lightning protection wires, line stringing equipment features, and line sag features.

[0061] Step S103: Classify the project progress characteristic information.

[0062] In one possible implementation, step S103 further includes: inputting the project progress feature information into a pre-trained convolutional neural network model constructed with a ResNet50 network as the backbone network; connecting a classification head including two fully connected layers to the back end of the ResNet50 network, wherein the first fully connected layer contains multiple neurons and uses the ReLU activation function, and the number of neurons in the second fully connected layer is consistent with the number of project progress stages; performing linear transformation processing on the project progress feature information through the first fully connected layer; and classifying the linearly transformed project progress feature information through the second fully connected layer.

[0063] Specifically, the project progress feature information extracted in the previous stage is used as input and fed into a pre-trained convolutional neural network model built with ResNet50 as the backbone to perform project progress classification. ResNet50 is a deep residual structure network containing 50 layers and residual connection units. Its core advantage lies in effectively mitigating the gradient vanishing and degradation problems in deep network training, thereby ensuring the effective transmission and expression of deep semantic features. This model has been pre-trained on large-scale image recognition datasets (such as ImageNet) and exhibits excellent image understanding and feature generalization capabilities. The model training parameters are as follows: Optimizer: The Adam optimizer is selected. The Adam optimizer combines the advantages of the Adaptive Gradient Algorithm (AdaGrad) and the Root Mean Square Propagation Algorithm (RMSProp), enabling rapid convergence in the early stages of training and stable optimization of model parameters in the later stages; Loss function: The cross-entropy loss function is used, with the formula:

[0064]

[0065] in, This represents the total cross-entropy loss. The total number of samples, The total number of categories, Indicates the first Does the _ sample belong to the _ ... Class (real tags); The model predicts the first... The sample belongs to the first The probability of a class This is the logarithm of the predicted probability.

[0066] After the project progress feature information undergoes convolution, normalization, and pooling processing in the ResNet50 backbone network, a customized classification head is appended to the end of the network. The classification head consists of two fully connected layers. The first fully connected layer performs a linear transformation on the high-dimensional feature vector output by the convolutional network. Its number of neurons is set to an intermediate dimension (e.g., 512 or 1024), and it utilizes the ReLU activation function to achieve non-linear enhancement of the feature space. Its mathematical expression is as follows:

[0067]

[0068] in This represents the project progress feature vector output by ResNet50. and These are the weight matrix and bias vector of the first fully connected layer, respectively. This represents the output features of the first layer.

[0069] then, The feature vector is input to the second fully connected layer for classification. The number of neurons in this layer is equal to the total number of project progress classification categories (e.g., 8 stages). Instead of using non-linear activation, it directly outputs the original classification vector, and then calculates the probability distribution for each category using the Softmax function.

[0070]

[0071] in and For the first The weights and biases corresponding to the categories, The total number of categories, Indicates that the input feature belongs to the first... The probability of a project progress stage.

[0072] Finally, the model output is a value of length [length missing]. The probability vector represents the matching probability of the current input image at each stage of the construction progress. The index position with the highest probability value is taken as the construction progress stage to which the current image belongs, thus completing the classification and recognition of the preprocessed target image. This step uses a combination of linear transformation and nonlinear mapping to accurately map complex image semantic features to a predefined construction stage space, realizing an end-to-end reasoning process from image to engineering semantic state. It should be noted that during training, before each round of training, the model is set to training mode (model.train()), and Dropout and BN layers are enabled; after each round of training, the model is switched to evaluation mode (model.eval()), Dropout and BN layer random operations are disabled, and the model performance is evaluated on the validation set to monitor changes in loss value and accuracy. If the validation set accuracy does not improve significantly in 10 consecutive rounds of training, a learning rate decay strategy (learning rate multiplied by 0.5) is adopted to avoid the model getting trapped in local optima until the model reaches a stable and high accuracy on the validation set.

[0073] Step S104: Based on the classified project progress feature information, match it with the project progress feature library to output the current project progress corresponding to the project construction area.

[0074] Specifically, based on the classified project progress feature information, the pre-built project progress feature library is invoked to perform feature matching operations, determine the standard progress template most similar to the current feature information, and output the current project progress stage of the construction area accordingly. The project progress feature library contains typical image feature representation structures under different stages, which are used to provide semantic comparison and judgment support for the classification results. The accuracy of the stage output by the classification is verified by the matching results, and the stability and interpretability of the recognition are improved.

[0075] In one possible implementation, step S104 further includes: establishing a project progress feature library based on the construction process and characteristics of the project construction area; using the data in the project progress feature library as training data to construct a project progress recognition model; and outputting the current project progress based on the project progress recognition model.

[0076] Specifically, in this step: different construction stages in substation construction have different characteristics. For example, the foundation construction stage can be judged by the pouring status of the tower foundation, while the main structure construction stage can be judged by the completion status of the building structure. These characteristics are quantified and classified to establish a project progress characteristic database. Please refer to Table 1 for the characteristic information of the project progress characteristic database.

[0077] [Table 1]

[0078]

[0079] A large amount of aerial and ground image data of the power transmission and transformation project area at different stages were collected, including satellite remote sensing images (resolution up to 0.5 meters), UAV aerial images (resolution 1-5 centimeters), and high-definition photos collected by ground equipment.

[0080] A team of professional power engineers and image annotation specialists was organized to annotate the images according to the project progress stages (foundation construction, tower erection, line stringing, equipment installation, commissioning and acceptance, etc., a total of 8 stages).

[0081] The LabelMe tool is used to accurately label key areas and features in images related to project progress, such as the pouring area of ​​tower foundations and the construction structure of transmission towers, to form a dataset with accurate progress stage labels, thus creating a project progress feature library.

[0082] Next, the data in the project progress feature library was used as training data to construct a project progress recognition model. The specific construction method of the model is as follows: First, based on the multi-stage construction sample images contained in the project progress feature library, the images were classified and organized according to the labeled project progress tags to form a standard training sample set with stage identifiers; Second, a convolutional neural network model with ResNet50 as the backbone network was adopted, which received the above sample images as input, extracted multi-scale image semantic features through multiple convolutional layers, and connected a fully connected classification layer at the end of the model to output a classification vector corresponding to the number of project stage categories; During the training phase, the model used the cross-entropy loss function as the objective function to minimize the error between the predicted classification and the true label.

[0083] To enhance the model's generalization ability, data augmentation strategies are introduced during training, including image rotation, horizontal flipping, brightness perturbation, and Gaussian blur, to construct a richer input distribution. After each training round, the accuracy and loss metrics are evaluated on the validation set. If the accuracy does not improve for several consecutive rounds of validation, the learning rate is automatically adjusted and the optimal model weights are saved.

[0084] After training, the model is capable of taking any preprocessed image as input and outputting its corresponding engineering stage prediction result, providing a stable prediction basis for subsequent stage classification, confidence inference, and dynamic judgment. The stage probability distribution output by the model serves as the source of the Softmax normalized probability.

[0085] In one possible implementation, step S104 further includes: generating a normalized probability distribution for each stage of the project progress based on the project progress recognition model using a Softmax activation function; performing structured processing on the normalized probability distribution and selecting the project progress corresponding to the highest probability value as the first candidate project progress; acquiring a historical project stage sequence based on historical image data and constructing a stage confidence evaluation function based on the dynamic evolution trend between the first candidate project progress and the historical project stage sequence; outputting a first confidence score corresponding to the first candidate project progress based on the stage confidence evaluation function; if it is confirmed that the first confidence score is less than a preset threshold, returning to the image acquisition step and reacquiring the target image data; outputting a second confidence score corresponding to the second candidate project progress based on the target image data using the project progress recognition model; if it is confirmed that the second confidence score is greater than or equal to a preset threshold, outputting the second candidate project progress as the current project progress.

[0086] Specifically, firstly, the target image is input into the engineering progress recognition model, and the normalized probability distribution predicted at each stage is generated based on the Softmax function of the model's final classification layer. Let the unnormalized classification vector output by the model be... ,in This represents the total number of predefined project schedule stages, i.e., the number of stage categories set in the model classification task, such as site leveling, foundation construction, civil engineering construction, equipment installation, etc. The project progress identification model represents the first... The unnormalized scores output for each stage of the project schedule are then used to calculate the Softmax output probability vector:

[0087]

[0088] The normalized vector This represents the prediction confidence probability for each stage. The normalized image belongs to the first... The predicted confidence probability of a project's progress stage, i.e., the relative likelihood that the image is classified as being in that stage. The model represents all stages ∈{1,…, The output logit score is then processed using structured methods, and the stage index with the highest probability value is defined as the first candidate project progress.

[0089]

[0090] Subsequently, the current target area was extracted from the engineering monitoring system in the past. The historical stage sequence corresponding to each moment , For the first The historical stages corresponding to each moment are identified, and a stage consistency model based on time evolution constraints is constructed. This is to quantify the first candidate stage. Consistency with historical stage sequences, defining a stage confidence evaluation function. as follows:

[0091]

[0092] in, This represents the confidence probability in the Softmax output corresponding to the first candidate stage; For phase consistency indicator functions, when The value is 1 if the condition is met, and 0 otherwise. This is the deep feature vector of the current image. For the first The feature vector of the image at each time step; This represents the cosine similarity function, used to measure the high-dimensional semantic consistency between the current image and historical images; And satisfy The adjustable confidence weighting coefficients are used to calculate the confidence scores for the above stages. With the set threshold If a comparison is made, If the current model output is considered sufficiently stable, then it will be directly output. As the actual project progress output for the target area; if If the current model output is deemed unstable, an image resampling mechanism is triggered, returning to the image acquisition step to acquire target image data from a new perspective or time point. Based on the resampling image, model inference is executed again to obtain the second candidate project progress. and their corresponding stage confidence scores The scoring is still calculated using the formula described above. If the conditions are met... Then The actual project progress output for the current target area is determined as the final identification result; if... If the model re-acquires and performs secondary inference on the target image, it indicates that the output of the second candidate project progress has not yet reached the preset confidence threshold. This means that the current prediction result is uncertain, unstable, or inconsistent with historical evolution trends in multiple aspects, such as probability confidence, historical consistency, feature stability, or inter-class boundary clarity. In this case, at least one of the following processing logics should be executed: Block the current project progress judgment process and do not output any valid stage judgment results to prevent misjudgment; generate abnormal marker information, packaging and recording the input data, Softmax probability distribution, stage confidence, and historical stage sequence corresponding to the image; enter the manual review or intelligent scheduling intervention process, sending a "stage recognition failure" signal to the project monitoring platform through the prompt system interface, and optionally requesting larger viewing angles, higher resolutions, or cross-device supplementary image data; count the frequency of stage recognition failures in the current area, and if multiple failures occur in adjacent consecutive time periods or adjacent spatial units... In such cases, the system is considered to have decreased model robustness or abnormal data acquisition, and can initiate model retraining suggestions or a data quality warning mechanism. This strategy combines the confidence of the current classification output, the continuity of historical stage evolution, and feature stability to significantly improve the robustness of recognition and engineering semantic interpretation capabilities in scenarios with blurred stage boundaries or ambiguous images.

[0093] Step S105: If it is confirmed that the progress deviation between the current project progress and the planned project progress is greater than the preset deviation value, a project progress warning is triggered.

[0094] Specifically, the system compares the current project progress with the planned project progress, calculates the progress deviation, and determines whether to trigger a project progress warning based on whether the deviation exceeds a preset deviation threshold. Specifically, the system extracts the planned project progress stage that should be reached at the current moment based on the time-stage mapping relationship set in the project scheduling plan. It then quantifies the stage-level difference between this stage and the actual project progress stage identified by the model, and calculates the deviation between the actual progress and the planned progress by combining the stage number sequence, duration, or key node extension.

[0095] If the deviation value is determined to be greater than the preset threshold, indicating that the current project is significantly behind schedule or abnormally ahead of schedule, the system will trigger the project progress early warning mechanism, output abnormal prompt information on the monitoring interface, and simultaneously notify project management personnel and on-site construction supervisors through the set notification paths (such as dispatch platform notifications, SMS reminders, and email pushes). At the same time, the time, location unit, and judgment basis of the current deviation will be marked so as to analyze the cause in a timely manner and adjust the construction pace.

[0096] If the deviation value does not exceed the preset threshold, meaning the current project progress is within the planned tolerance range, no warning will be issued. In this case, the system considers the current construction status to be in line with the expected pace. There may be slight misjudgments of stages due to minor time differences, overlapping stage boundaries, or multi-source observation errors, which are insufficient to constitute interference with construction scheduling. The system will record the current status as "normal" or "warning-free state" and continue tracking it, but will not trigger any manual intervention or scheduling response. This strategy helps reduce the false alarm rate and improves the practicality and reliability of the warnings.

[0097] Please refer to Figure 2 This document illustrates a schematic diagram of the modules of a power transmission and transformation project progress identification device based on integrated air-space-ground technology provided in an embodiment of this application. The device includes an acquisition module 21 and a processing module 22.

[0098] The acquisition module 21 is used to acquire target image data in the construction area through multiple acquisition devices and to perform preprocessing operations on the target image data.

[0099] The processing module 22 is used to extract features from the preprocessed image data to obtain the project progress feature information corresponding to the construction area; classify the project progress feature information; match the classified project progress feature information with the project progress feature library to output the current project progress corresponding to the construction area; if it is confirmed that the progress deviation between the current project progress and the planned project progress is greater than the preset deviation value, a project progress warning is triggered.

[0100] In one possible implementation, the acquisition module 21 is used to acquire target image data in the construction area through multiple acquisition devices, specifically including: periodically taking pictures of the construction area through satellite remote sensing equipment to acquire satellite image data; taking aerial pictures of the construction area from multiple angles through drone equipment to acquire aerial image data; taking close-up pictures of the construction area through intelligent robot equipment to acquire complex area image data; and using satellite image data, aerial image data, and complex area image data as target image data.

[0101] In one possible implementation, the acquisition module 21 is used to perform radiometric correction on the satellite image data to eliminate radiometric errors, and to perform geometric correction on the satellite image data to correct the satellite image data to the correct geographic coordinate system; to stitch and fuse the aerial image data, and to perform noise reduction processing on the fused image data; to extract frames from the image data of complex areas to decompose the video into single-frame images, and to perform size normalization processing on the image data of complex areas and each frame image.

[0102] In one possible implementation, the processing module 22 is used to extract features from the preprocessed image data to obtain engineering progress feature information corresponding to the construction area. Specifically, it includes: inputting the preprocessed target image data into a convolutional neural network model and outputting engineering progress feature information based on the convolutional neural network model; and extracting advanced semantic features from the preprocessed target image data based on the engineering progress feature information. The advanced semantic features include substation building features and transmission line features.

[0103] In one possible implementation, the processing module 22 is used to classify the project progress feature information, specifically including: inputting the project progress feature information into a pre-trained convolutional neural network model constructed with a ResNet50 network as the backbone network; connecting a classification head including two fully connected layers to the back end of the ResNet50 network, wherein the first fully connected layer contains multiple neurons and uses the ReLU activation function, and the number of neurons in the second fully connected layer is consistent with the number of project progress stages; performing linear transformation processing on the project progress feature information through the first fully connected layer; and classifying the project progress feature information after linear transformation processing through the second fully connected layer.

[0104] In one possible implementation, the processing module 22 is used to match the classified project progress feature information with the project progress feature library to output the current project progress corresponding to the project construction area. Specifically, it includes: establishing a project progress feature library based on the construction process and characteristics of the project construction area; using the data in the project progress feature library as training data to construct a project progress recognition model; and outputting the current project progress based on the project progress recognition model.

[0105] In one possible implementation, the processing module 22 is used to output the current project progress based on the project progress recognition model, specifically including: generating a normalized probability distribution of each project progress stage using the Softmax activation function based on the project progress recognition model; performing structured processing on the normalized probability distribution and taking the project progress corresponding to the maximum probability value as the first candidate project progress; obtaining a historical project stage sequence based on historical image data, and constructing a stage confidence evaluation function based on the dynamic evolution trend between the first candidate project progress and the historical project stage sequence; outputting a first confidence score corresponding to the first candidate project progress based on the stage confidence evaluation function; if it is confirmed that the first confidence score is less than a preset threshold, returning to the image acquisition step and reacquiring the target image data; outputting a second confidence score corresponding to the second candidate project progress based on the target image data through the project progress recognition model; if it is confirmed that the second confidence score is greater than or equal to the preset threshold, outputting the second candidate project progress as the current project progress.

[0106] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0107] This application also provides an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 301, at least one communication bus 302, a user interface 303, at least one network interface 304, and a memory 305.

[0108] The communication bus 302 is used to enable communication between these components.

[0109] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0110] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0111] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0112] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application for identifying the progress of power transmission and transformation projects based on integrated air-space-ground technology.

[0113] exist Figure 3In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call the power transmission and transformation project progress identification application based on integrated air-space-ground technology stored in the memory 305. When executed by one or more processors 301, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0114] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0115] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0116] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0117] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0118] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0119] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0120] The above description is merely an exemplary embodiment disclosed in this application and should not be construed as limiting the scope of this application. Any equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application.

[0121] This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary technical means in the art that are not described in this application.

Claims

1. A method for identifying the progress of power transmission and transformation projects based on integrated space-air-ground technology, characterized in that, The method includes: Target image data of the construction area is acquired through multiple acquisition devices, and the target image data is preprocessed. Feature extraction is performed on the preprocessed image data to obtain the project progress feature information corresponding to the construction area. The project progress characteristic information is classified; Based on the classified project progress feature information, the project progress feature database is used for matching to output the current project progress corresponding to the project construction area. If the deviation between the current project progress and the planned project progress is confirmed to be greater than a preset deviation value, a project progress warning is triggered. The classification of the project progress feature information specifically includes: inputting the project progress feature information into a pre-trained convolutional neural network model constructed with a ResNet50 network as the backbone; connecting a classification head consisting of two fully connected layers to the back end of the ResNet50 network, wherein the first fully connected layer contains multiple neurons and uses the ReLU activation function, and the number of neurons in the second fully connected layer is consistent with the number of project progress stages; performing a linear transformation on the project progress feature information through the first fully connected layer; classifying the linearly transformed project progress feature information through the second fully connected layer; and matching the classified project progress feature information with a project progress feature library to output the current project progress corresponding to the project construction area, specifically including: establishing the project progress feature library based on the construction process and characteristics of the project construction area; using the data in the project progress feature library as training data to construct a project progress recognition model, and outputting the current project progress based on the project progress recognition model.

2. The method according to claim 1, characterized in that, The acquisition of target image data in the construction area through multiple acquisition devices specifically includes: The construction area of ​​the project is periodically photographed using satellite remote sensing equipment to obtain satellite image data; The construction area was photographed from multiple angles using drone equipment to obtain aerial image data; The construction area of ​​the project is photographed at close range using intelligent robotic equipment to obtain image data of complex areas. The satellite image data, the aerial image data, and the complex area image data are used as the target image data.

3. The method according to claim 2, characterized in that, The preprocessing operation on the target image data specifically includes: Radiometric correction is performed on the satellite image data to eliminate radiometric errors, and geometric correction is performed on the satellite image data to correct the satellite image data to the correct geographic coordinate system; The aerial image data is stitched and fused, and the fused image data is then denoised. Frame extraction is performed on the complex area image data to decompose the video into single-frame images, and the size normalization processing is performed on the complex area image data and each frame image.

4. The method according to claim 1, characterized in that, The step of extracting features from the preprocessed image data to obtain project progress feature information corresponding to the construction area specifically includes: The preprocessed target image data is input into a convolutional neural network model, and the project progress feature information is output based on the convolutional neural network model. Based on the project progress feature information, advanced semantic features are extracted from the preprocessed target image data. The advanced semantic features include substation building features and transmission line features.

5. The method according to claim 1, characterized in that, The step of outputting the current project progress based on the project progress recognition model specifically includes: Based on the project progress identification model, the Softmax activation function is used to generate the normalized probability distribution of each project progress stage. The normalized probability distribution is structured, and the project progress corresponding to the maximum probability value is taken as the first candidate project progress. Historical project phase sequences are obtained based on historical image data, and a phase confidence evaluation function is constructed based on the dynamic evolution trend between the progress of the first candidate project and the historical project phase sequences. The first confidence score corresponding to the progress of the first candidate project is output based on the stage confidence evaluation function. If it is confirmed that the first confidence score is less than the preset threshold, then return to the image acquisition step and reacquire the target image data; Based on the target image data, the second confidence score corresponding to the second candidate project progress is output through the project progress recognition model; If it is confirmed that the second confidence score is greater than or equal to the preset threshold, then the second candidate project progress is output as the current project progress.

6. A power transmission and transformation project progress identification device based on integrated air-space-ground technology, characterized in that, The device includes an acquisition module and a processing module, wherein, The acquisition module is used to acquire target image data in the construction area through multiple acquisition devices and to perform preprocessing operations on the target image data. The processing module is used to extract features from the preprocessed image data to obtain project progress feature information corresponding to the construction area; classify the project progress feature information; match the classified project progress feature information with a project progress feature library to output the current project progress corresponding to the construction area; if it is confirmed that the progress deviation between the current project progress and the planned project progress is greater than a preset deviation value, a project progress warning is triggered; the classification of the project progress feature information specifically includes: inputting the project progress feature information into a pre-trained convolutional neural network model constructed with a ResNet50 network as the backbone network; connecting a classification head including two fully connected layers to the back end of the ResNet50 network, wherein... The first fully connected layer contains multiple neurons and uses the ReLU activation function. The number of neurons in the second fully connected layer is consistent with the number of project progress stages. The project progress feature information is linearly transformed through the first fully connected layer. The project progress feature information after linear transformation is classified through the second fully connected layer. The step of matching the classified project progress feature information with the project progress feature library to output the current project progress corresponding to the project construction area specifically includes: establishing the project progress feature library based on the construction process and characteristics of the project construction area; using the data in the project progress feature library as training data to construct a project progress recognition model; and outputting the current project progress based on the project progress recognition model.

7. An electronic device, characterized in that, The device includes a processor, a communication bus, a user interface, a network interface, and a memory. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 5.

Images