Intelligent Monitoring and Progress Management System for High-Altitude Converter Station Site Construction

Abstract

This invention belongs to the field of engineering construction management and relates to an intelligent monitoring and progress management system for the construction of high-altitude converter station aprons. The system includes: a micro-meteorological acquisition node to acquire real-time wind speed and air pressure data; a UAV aerial survey data preprocessing terminal to perform optical distortion compensation and feature point offset correction on aerial survey images based on wind speed and air pressure data, and extract morphological features of the construction surface based on the corrected aerial survey images; and a progress dynamic prediction server that inputs the changes in the morphological features of the construction surface and meteorological attenuation factors into a long short-term memory network, outputs construction progress prediction time-series data, and compares the construction progress prediction time-series data with the baseline progress plan to generate scheduling instructions. This invention overcomes the interference of high-altitude low air pressure and strong winds on aerial survey images, improves the accuracy of construction surface feature extraction, eliminates the nonlinear impact of meteorological fluctuations on construction efficiency, realizes pre-construction progress prediction and dynamic tracking, and solves the problems of delayed progress assessment and distorted control.

Description

Technical Field

[0001] This invention belongs to the field of engineering construction management and relates to an intelligent monitoring and progress management system for the construction of high-altitude converter station site. Background Technology

[0002] The construction environment of high-altitude converter station aprons is characterized by low air pressure, low oxygen, and frequent strong winds, posing high health risks to construction workers. Manual inspections are difficult and dangerous. Current technologies increasingly employ unmanned aerial vehicle (UAV) aerial surveying to replace manual inspections for apron construction monitoring in such harsh environments. Conventional UAV monitoring systems periodically photograph the construction site along preset flight paths, acquiring surface image data, which is then transmitted back to a backend server. The backend server uses image stitching and comparison technology to analyze the aerial survey images from different time periods, extracting changes in the area and elevation of the construction surface, which are then used as a basis for assessing construction progress. This conventional approach can meet basic progress tracking needs in plains areas or environments with stable weather conditions, achieving partial automation of the construction process.

[0003] At the implementation level, existing UAV monitoring and progress management systems typically use camera calibration models under standard atmospheric parameters for image geometric correction and distortion processing, assuming a calm or light wind environment with uniform atmospheric refractive index. Progress management modules often calculate construction deviations by subtracting the static baseline progress plan from the actual acquired images. However, high-altitude areas present complex micro-meteorological conditions. Low air pressure alters air density and refractive index, while strong winds cause high-frequency attitude jitter in UAVs during aerial photography. Existing systems do not incorporate air pressure and wind speed parameters into the image preprocessing workflow, resulting in optical distortion and feature point shifts in the acquired aerial survey images. Furthermore, progress calculations do not consider the degrading effect of sudden weather changes on mechanical operation efficiency, still employing a linear evaluation model.

[0004] Because existing technologies do not incorporate high-altitude micro-meteorological parameters into aerial survey image preprocessing and progress projection calculations, conventional camera calibration models and static progress comparison logic cannot adapt to the unique high-altitude environment. Low air pressure causes changes in air refractive index, and strong winds induce UAV attitude deflection, resulting in uncorrected optical distortion and feature point shifts in aerial survey images. This leads to discrepancies between the morphological features of the construction surface extracted in the background and the actual surface conditions. Progress calculations based on distorted morphological features, coupled with a static evaluation model that does not consider the nonlinear decay of mechanical operation efficiency due to sudden weather changes, cause the system's output progress data to deviate from the actual construction situation, resulting in delayed progress control. This constitutes the core technical problem of delayed progress assessment and distorted control in high-altitude site construction. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring and progress management system for the construction of high-altitude converter station aprons, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The intelligent monitoring and progress management system for high-altitude converter station apron construction includes: a high-altitude micro-meteorological acquisition node for acquiring real-time wind speed and air pressure data; a UAV aerial survey data preprocessing terminal for performing optical distortion compensation and feature point offset correction on aerial survey images based on the wind speed and air pressure data, and extracting morphological features of the construction surface based on the corrected aerial survey images; and a progress dynamic projection server for inputting the changes in the morphological features of the construction surface and the meteorological attenuation factor into a long short-term memory network, outputting construction progress prediction time series data, and comparing the construction progress prediction time series data with the baseline progress plan to generate scheduling instructions.

[0008] Preferably, the UAV aerial survey data preprocessing end performs optical distortion compensation and feature point offset correction on the aerial survey image based on the wind speed and air pressure data, including: calculating the air refractive index gradient distribution based on the air pressure data, and constructing an air pressure deformation compensation matrix in combination with the camera intrinsic parameter matrix to perform optical distortion compensation on the aerial survey image; calculating the yaw angle and pitch angle changes at the moment of UAV aerial photography based on the wind speed data, mapping the yaw angle and pitch angle changes to the pixel coordinate system, and generating a feature point offset vector to perform feature point offset correction on the aerial survey image.

[0009] Preferably, the UAV aerial survey data preprocessing end extracts morphological features of the construction surface based on the corrected aerial survey image, including: inputting the corrected aerial survey image into a multi-scale feature extraction network to obtain the initial texture features and boundary contour features of the construction surface under different receptive fields; constructing a morphological constraint kernel based on prior knowledge of the pixel distribution of high-altitude site construction materials; and using the morphological constraint kernel to perform non-maximum suppression filtering on the initial texture features and boundary contour features to generate the morphological features of the construction surface.

[0010] Preferably, the progress dynamic simulation server inputs the changes in the morphological features of the construction surface and the meteorological attenuation factor into the long short-term memory network, including: calculating the feature difference matrix of the morphological features of the construction surface in adjacent time series as the changes; mapping the real-time wind speed and air pressure data to a high-altitude construction efficiency surface and extracting the meteorological attenuation factor; and concatenating the feature difference matrix and the meteorological attenuation factor along the feature channel dimension and inputting them into the temporal coding layer of the long short-term memory network.

[0011] Preferably, the progress dynamic simulation server compares the construction progress prediction time series data with the baseline progress plan to generate scheduling instructions, including: calculating the progress deviation vector between the construction progress prediction time series data and the baseline progress plan; determining whether the progress deviation vector exceeds a preset progress tolerance range, and if it does, extracting the key lagging process identifiers from the progress deviation vector; performing forward and backward tracing in the construction topology network based on the key lagging process identifiers to locate the affected set of related processes, and generating the scheduling instructions according to the resource requirements of the set of related processes.

[0012] Preferably, the calculation of the air refractive index gradient distribution based on the air pressure data, and the construction of an air pressure deformation compensation matrix in conjunction with the camera intrinsic parameter matrix to compensate for optical distortion of the aerial survey image, includes: acquiring surface temperature distribution data of the high-altitude field area, constructing a space aerosol density field model in conjunction with the air pressure data; calculating the deflection angle of the light propagation path based on the space aerosol density field model, performing an orthogonal transformation between the deflection angle and the focal length parameter in the camera intrinsic parameter matrix to generate the air pressure deformation compensation matrix containing a thermal field disturbance term, and performing pixel-by-pixel remapping of the aerial survey image.

[0013] Preferably, the corrected aerial survey image is input into a multi-scale feature extraction network to obtain the initial texture features and boundary contour features of the construction surface under different receptive fields. This includes: introducing an altitude gradient perception branch into the multi-scale feature extraction network; fusing the UAV flight altitude data and digital elevation model data corresponding to the corrected aerial survey image into an altitude gradient feature map; and performing cross-channel interactive fusion of the altitude gradient feature map and the deep semantic features of the corrected aerial survey image to generate the initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information.

[0014] Preferably, mapping the real-time wind speed and air pressure data to a high-altitude construction efficiency surface and extracting the meteorological attenuation factor includes: acquiring historical meteorological time-series data and corresponding historical mechanical operation efficiency data of the high-altitude converter station apron; using the historical meteorological time-series data and the historical mechanical operation efficiency data to fit and construct the high-altitude construction efficiency surface; inputting the real-time wind speed and air pressure data as coordinate indices into the high-altitude construction efficiency surface, interpolating to solve for the dynamic attenuation weight at the current moment, and using the dynamic attenuation weight as the meteorological attenuation factor.

[0015] Preferably, the system also includes an adaptive waypoint planning terminal, which acquires the scheduling instructions and the meteorological attenuation factor output by the progress dynamic simulation server; parses the scheduling instructions to extract the coordinates of the area to be monitored; adjusts the UAV's cruise altitude and overlap rate parameters based on the meteorological attenuation factor; when the meteorological attenuation factor exceeds a preset attenuation threshold, reduces the cruise altitude and increases the overlap rate parameter to generate an adaptive flight path, and sends the adaptive flight path to the UAV flight control system.

[0016] Preferably, it also includes an edge computing node deployed at the construction site. The edge computing node receives the initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information output by the UAV aerial survey data preprocessing terminal. When the communication link between the edge computing node and the progress dynamic simulation server is interrupted, the initial texture features and boundary contour features of the construction surface are stored in a local circular queue buffer. After the communication link is restored, the stored initial texture features and boundary contour features of the construction surface are retransmitted to the progress dynamic simulation server in the order of timestamps.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] 1. By performing optical distortion compensation and feature point offset correction on aerial survey images based on real-time wind speed and air pressure data at the UAV aerial survey data preprocessing end, the interference of high altitude, low air pressure, and strong winds on image acquisition quality is overcome, and the true morphological features of the construction surface are obtained. In the progress dynamic extrapolation server, the morphological feature changes and meteorological attenuation factors are input into the long short-term memory network to output construction progress prediction time series data. The construction progress prediction time series data is compared with the baseline progress plan to generate scheduling instructions, eliminating the nonlinear impact of meteorological fluctuations on construction efficiency, realizing the pre-examination and dynamic tracking of construction progress, and solving the problems of lagging assessment and distorted control of construction progress at high altitudes. The meteorological attenuation factor and morphological feature changes are concatenated along the feature channel dimension in the long short-term memory network and input into the temporal coding layer to capture the temporal dependency between construction status and meteorological disturbances, so that the output prediction time series data includes the trend prediction under future meteorological intervention. The construction progress forecast time series data is compared with the baseline schedule plan to generate a progress deviation vector. When the progress deviation vector exceeds the preset progress tolerance range, the key lagging process identifier is extracted and traced forward and backward in the construction topology network to locate the set of related processes affected. Scheduling instructions are generated based on the resource requirements of the set of related processes, realizing the physical and logical transformation from passive response to active intervention.

[0019] 2. Based on air pressure data, the air refractive index gradient distribution is calculated, and an air pressure deformation compensation matrix is ​​constructed by combining it with the camera intrinsic parameter matrix. Additionally, feature point offset vectors are generated by calculating the changes in UAV yaw and pitch angles based on wind speed data, eliminating the perturbation of the high-altitude atmospheric environment on the imaging optical path and camera attitude from a pixel-level physical mechanism. A spatial aerosol density field model is constructed by combining surface temperature distribution data and air pressure data. The light deflection angle and the focal length parameter in the camera intrinsic parameter matrix are orthogonally transformed to generate an air pressure deformation compensation matrix containing thermal field perturbation terms, achieving fine correction of the aerial survey image pixel-by-pixel remapping. Real-time wind speed and air pressure data are mapped to a high-altitude construction efficiency surface constructed by fitting historical meteorological time-series data and historical mechanical operation efficiency data. Dynamic attenuation weights are solved by interpolation to quantify the attenuation law of micro-meteorology on mechanical operation efficiency. The UAV waypoint adaptive planning terminal adjusts the cruise altitude and overlap rate parameters according to the meteorological attenuation factor and scheduling instructions. When the meteorological attenuation factor exceeds the preset attenuation threshold, the cruise altitude is reduced and the overlap rate is increased to ensure the quality of data sources under severe weather conditions. When the communication link is interrupted, the edge computing node stores the initial texture features and boundary contour features of the construction surface in a local circular queue buffer. After the communication link is restored, the interrupted data is resumed in the order of timestamps, ensuring the continuity of data flow in high-altitude and remote areas. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the overall operation of the intelligent monitoring and progress management system for high-altitude converter station construction.

[0021] Figure 2 This is a flowchart of the aerial survey image optical distortion compensation and feature point offset correction process of the present invention;

[0022] Figure 3 This is a flowchart of the construction surface morphological feature extraction process of the present invention;

[0023] Figure 4 This is a flowchart illustrating the process of generating time-series data for construction progress prediction according to the present invention.

[0024] Figure 5 A flowchart for generating construction scheduling instructions according to the present invention;

[0025] Figure 6 This is a flowchart of the UAV waypoint adaptive planning and edge computing data transmission process of the present invention. Detailed Implementation

[0026] refer to Figure 1In one embodiment, the intelligent monitoring and progress management system for the construction of a high-altitude converter station apron includes high-altitude micro-meteorological data acquisition nodes, a UAV aerial survey data preprocessing terminal, and a progress dynamic simulation server. The high-altitude micro-meteorological data acquisition nodes are deployed at multiple key locations on the construction apron. Each node integrates a barometric pressure sensor and a wind speed sensor, with the sensor sampling frequency set to 1Hz. The acquired data is transmitted to the UAV aerial survey data preprocessing terminal via a LoRa communication module. The UAV aerial survey data preprocessing terminal is deployed at a temporary command center on the construction site, receiving real-time wind speed and barometric pressure data from the high-altitude micro-meteorological data acquisition nodes, and simultaneously receiving aerial survey image data transmitted back from the UAV flight control system. Based on the received wind speed and barometric pressure data, the UAV aerial survey data preprocessing terminal performs optical distortion compensation and feature point offset correction on the aerial survey images. Based on the corrected aerial survey images, it extracts the morphological features of the construction surface and transmits these features to the progress dynamic simulation server via an optical fiber communication link. The progress dynamic simulation server is deployed in a cloud data center. It receives the morphological features of the construction surface from the UAV aerial survey data preprocessing terminal, calculates the changes in the morphological features of the construction surface in adjacent time series, and simultaneously maps real-time wind speed and air pressure data to a high-altitude construction efficiency surface to extract meteorological attenuation factors. The changes in the morphological features of the construction surface and the meteorological attenuation factors are input into a long short-term memory network, and the server outputs the construction progress prediction time series data. The progress dynamic simulation server compares the construction progress prediction time series data with the pre-stored baseline schedule plan, generates scheduling instructions, and sends the scheduling instructions to the scheduling terminal at the construction site.

[0027] In this embodiment, the deployment locations of the high-altitude micro-meteorological data acquisition nodes are determined based on the terrain features and meteorological distribution patterns of the construction site. Data acquisition nodes are deployed at the four corners, the center, and areas with significant terrain undulations, forming a micro-meteorological monitoring network covering the entire construction area. Each data acquisition node uses a capacitive barometric pressure sensor with a measurement range of 30 kPa to 110 kPa and a resolution of 0.01 kPa; and an ultrasonic anemometer with a measurement range of 0 m / s to 60 m / s and a resolution of 0.1 m / s. The data acquisition nodes have built-in solar power modules and lithium battery energy storage modules, enabling them to operate continuously for more than 30 days without an external power supply. After timestamping the collected real-time wind speed and pressure data, the data acquisition nodes broadcast the data via a LoRa communication module, achieving a communication distance of up to 5 km, covering most high-altitude converter station construction areas.

[0028] The UAV aerial survey data preprocessing unit utilizes an industrial-grade computer equipped with a multi-core processor and large-capacity memory, enabling simultaneous processing of multiple UAV aerial survey image data streams. A LoRa gateway establishes a communication connection between the UAV aerial survey data preprocessing unit and high-altitude micro-meteorological acquisition nodes, receiving real-time wind speed and air pressure data from all acquisition nodes. Based on the location information of the acquisition nodes, the preprocessing unit performs spatial interpolation processing on the data to generate wind speed and air pressure field data covering the entire construction area. A 5G communication link connects the UAV flight control system, receiving aerial survey image data captured by the UAV, along with corresponding aerial photography time, UAV position, attitude, and other auxiliary information. For each aerial survey image, the preprocessing unit extracts the corresponding wind speed and air pressure data from the generated wind speed and air pressure field data based on the aerial photography time and shooting location, for subsequent image correction processing.

[0029] The progress dynamic simulation server adopts a distributed cloud computing architecture, consisting of multiple computing nodes, capable of processing large amounts of construction surface morphological feature data and progress simulation calculation tasks in parallel. The server pre-stores a baseline schedule for the high-altitude converter station apron construction. This baseline schedule is represented in the form of a construction topology network, including the start time, end time, duration, resource requirements, and logical dependencies between each process. After receiving the construction surface morphological features from the UAV aerial survey data preprocessing terminal, the server stores them in a time-series database according to timestamp order and calculates the changes in morphological features of adjacent time-series construction surfaces. Simultaneously, the server receives real-time wind speed and air pressure data from high-altitude micro-meteorological acquisition nodes, mapping them to a high-altitude construction efficiency surface to extract meteorological attenuation factors. The server then inputs the changes in construction surface morphological features and meteorological attenuation factors into a pre-trained long short-term memory network, outputting time-series data predicting the construction progress for the next 7 days. The progress dynamic simulation server compares the construction progress prediction time series data with the baseline progress plan, calculates the progress deviation vector, and when the progress deviation vector exceeds the preset progress tolerance range, it extracts the key lagging process identifiers and performs forward and backward tracing in the construction topology network to locate the set of affected related processes. Based on the resource requirements of the set of related processes, it generates scheduling instructions.

[0030] In this embodiment, the Long Short-Term Memory (LSTM) network structure includes an input layer, a temporal coding layer, a hidden layer, and an output layer. The number of neurons in the input layer equals the sum of the dimensions of the construction surface morphology features and the meteorological attenuation factor. The temporal coding layer consists of three stacked LSM units, each containing 128 hidden neurons. The hidden layer contains 64 fully connected neurons. The number of neurons in the output layer equals the length of the construction progress prediction time-series data. The LSM network is trained using historical construction data and meteorological data. The training dataset includes morphological feature data of the high-altitude converter station's construction surface, meteorological data, and corresponding actual construction progress data from the past 12 months. During training, mean squared error is used as the loss function, and the Adam optimization algorithm is used for parameter updates. The learning rate is set to 0.001, the batch size to 32, and the number of training epochs to 100. After training, the LSM network can accurately predict the construction progress over a future period based on the changes in the input construction surface morphology features and the meteorological attenuation factor.

[0031] Table 1 Deployment Parameters for High-Altitude Micrometeorological Data Acquisition Nodes

[0032]

[0033] In this embodiment, Table 1 shows the deployment parameters of the high-altitude micro-meteorological data acquisition nodes, including deployment location, node number, sensor model, communication module model, and power supply method. By deploying multiple acquisition nodes at key locations on the construction site, real-time wind speed and air pressure data of the construction area can be obtained comprehensively and accurately, providing reliable meteorological data support for subsequent aerial survey image correction and progress prediction.

[0034] In this embodiment, the system workflow is as follows: High-altitude micro-meteorological data acquisition nodes collect real-time wind speed and air pressure data of the construction area and send the data to the UAV aerial survey data preprocessing terminal; the UAV conducts aerial surveys of the construction site according to a preset route, captures aerial survey images, and sends the image data and auxiliary information to the UAV aerial survey data preprocessing terminal; the UAV aerial survey data preprocessing terminal performs optical distortion compensation and feature point offset correction on the aerial survey images based on real-time wind speed and air pressure data, extracts morphological features of the construction surface based on the corrected aerial survey images, and sends the feature data to the progress dynamic simulation server; the progress dynamic simulation server calculates the changes in the morphological features of the construction surface, extracts the meteorological attenuation factor, inputs both into a long short-term memory network, and outputs construction progress prediction time-series data; the progress dynamic simulation server compares the construction progress prediction time-series data with the baseline progress plan, generates scheduling instructions, and sends them to the scheduling terminal at the construction site. This embodiment overcomes the interference of high-altitude low air pressure and strong winds on image acquisition quality by introducing high-altitude micro-meteorological parameters into aerial survey image preprocessing and progress prediction calculations, eliminates the nonlinear impact of meteorological fluctuations on construction efficiency, and realizes the pre-deduction and dynamic tracking of construction progress.

[0035] refer to Figure 2 In one embodiment, the UAV aerial survey data preprocessing unit performs optical distortion compensation and feature point offset correction on the aerial survey images based on wind speed and air pressure data. The air refractive index gradient distribution is calculated based on the air pressure data, and a pressure deformation compensation matrix is ​​constructed in conjunction with the camera intrinsic parameter matrix to compensate for optical distortion in the aerial survey images. The changes in yaw and pitch angles at the moment of UAV aerial photography are calculated based on wind speed data, and these changes are mapped to the pixel coordinate system to generate feature point offset vectors for feature point offset correction in the aerial survey images.

[0036] In this embodiment, the air refractive index is calculated using the Edlén formula, which describes the relationship between air refractive index and air pressure, temperature, humidity, and carbon dioxide concentration. In high-altitude areas, the changes in air humidity and carbon dioxide concentration are relatively small; for simplicity, they can be considered constants. Therefore, the air refractive index n can be expressed as:

[0037]

[0038] Where P is air pressure, in Pa; and T is thermodynamic temperature, in K.

[0039] The calculation of the air refractive index gradient distribution is based on spatial interpolation. Using discrete pressure and temperature data acquired from high-altitude micro-meteorological data collection nodes, an inverse distance-weighted interpolation method is employed to generate pressure and temperature field data covering the entire construction area. This data is then used to calculate the air refractive index at each spatial point, forming the air refractive index gradient distribution field. The formula for the inverse distance-weighted interpolation method is as follows:

[0040]

[0041] in, Points to be interpolated The attribute value; The attribute value of the i-th known sampling point; is the distance between the point to be interpolated and the i-th known sampling point; k is the power exponent, usually taken as 2; The number of known sampling points.

[0042] The camera intrinsic parameter matrix K represents the camera's internal geometric parameters, and its form is:

[0043]

[0044] in, and These are the camera's focal lengths along the x and y axes, respectively, in pixels; and These are the coordinates of the camera's principal point along the x-axis and y-axis, respectively, in pixels.

[0045] The construction of the pressure deformation compensation matrix M is based on the geometric optics principle of light propagation. When light passes through a non-uniform medium, refraction occurs, causing the light propagation path to deflect. For aerial survey cameras, after light is reflected from objects on the ground, it passes through the atmosphere to reach the camera lens. Due to the gradient distribution of the air's refractive index, the light propagation path bends, causing an offset in the imaging position. The pressure deformation compensation matrix M is used to describe this offset in the imaging position, and its construction process is as follows: For each pixel in the aerial survey image... The ray direction vector in the camera coordinate system is obtained by back-projecting the camera intrinsic parameter matrix K onto the camera coordinate system. The deflection angle of the ray during propagation is calculated based on the air refractive index gradient distribution field. The ray direction vector is then corrected based on the deflection angle. Finally, the corrected ray direction vector is reprojected onto the pixel coordinate system to obtain the corrected pixel coordinates. Based on the original pixel coordinates With the corrected pixel coordinates The mapping relationship between them is used to construct the pressure deformation compensation matrix M.

[0046] Light deflection angle The calculation formula is:

[0047]

[0048] Where n is the refractive index of air; The air refractive index gradient; Let L be the unit vector representing the direction of light propagation; L is the length of the light propagation path. Let be the infinitesimal length of the light propagation path.

[0049] In this embodiment, to improve computational efficiency, a piecewise integration method is used to calculate the light deflection angle. The light propagation path is divided into N equal-length infinitesimal segments, each with a length of... Then the angle of light refraction can be approximately expressed as:

[0050]

[0051] in, Let be the air refractive index at the midpoint of the i-th infinitesimal segment; Let be the air refractive index gradient at the midpoint of the i-th infinitesimal segment; Let be the unit vector of the direction of light propagation within the i-th infinitesimal segment.

[0052] The changes in yaw and pitch angles at the moment of drone aerial photography are calculated based on wind speed data. During flight, the drone experiences attitude jitter due to wind forces, causing changes in the camera's attitude at the moment of aerial photography and resulting in the displacement of feature points in the image. The drone's attitude changes are mainly manifested in changes in yaw, pitch, and roll angles, among which the change in roll angle has a relatively small impact on imaging and can be ignored. Therefore, this embodiment mainly considers the impact of changes in yaw and pitch angles on imaging.

[0053] Change in yaw angle during drone aerial photography With pitch angle change The calculations are based on aerodynamic principles. The wind moment acting on the drone is proportional to the square of the wind speed and the drone's frontal area. According to the drone's dynamic model, the wind moment causes angular acceleration, which in turn leads to changes in attitude angles. The change in yaw angle is also considered. With pitch angle change The calculation formula is:

[0054]

[0055]

[0056] in, This is the yaw moment coefficient; This is the pitch moment coefficient; ρ is air density; v is wind speed; S is the frontal area of ​​the UAV; L is the wingspan of the UAV; c is the average aerodynamic chord of the UAV. Let be the moment of inertia of the UAV about the z-axis; Let be the moment of inertia of the UAV about the y-axis; This refers to the response time of the UAV attitude stabilization system.

[0057] The changes in yaw and pitch angles are mapped to the pixel coordinate system to generate feature point offset vectors. The pixel coordinate offsets caused by camera attitude changes can be calculated using a rotation matrix. (Yaw angle change...) Corresponding rotation matrix for:

[0058]

[0059] Pitch angle change Corresponding rotation matrix for:

[0060]

[0061] The total rotation matrix R after the camera pose change is:

[0062]

[0063] For any feature point in the image, its coordinates in the camera coordinate system are: The coordinates after transformation by rotation matrix R are Projecting the transformed coordinates onto the pixel coordinate system yields the transformed pixel coordinates. Feature point offset vector ,in These are the original pixel coordinates of the feature points.

[0064] In this embodiment, surface temperature distribution data of the high-altitude construction site area is acquired, and a spatial aerosol density field model is constructed by combining it with air pressure data. Surface temperature distribution data is acquired using an infrared thermal imager deployed at a high point on the construction site, capable of capturing infrared thermal images covering the entire construction area. The infrared thermal images have a resolution of 640×512, a temperature measurement range of -40℃ to 150℃, and a measurement accuracy of ±0.5℃. The infrared thermal images are registered with the digital elevation model data to obtain surface temperature data for each spatial point. Combined with air pressure data acquired from high-altitude micro-meteorological acquisition nodes, a Gaussian process regression method is used to construct the spatial aerosol density field model. Gaussian process regression is a nonparametric Bayesian method that can effectively handle high-dimensional data and nonlinear relationships, making it suitable for constructing spatial distribution field models.

[0065] The deflection angle of the light propagation path is calculated based on a space aerosol density field model. This deflection angle is then orthogonally transformed with the focal length parameter in the camera's intrinsic parameter matrix to generate a pressure deformation compensation matrix that includes a thermal field disturbance term. This matrix is ​​then used for pixel-by-pixel remapping of the aerial survey images. Aerosol particles scatter light, causing further deflection of the light propagation path. The space aerosol density field model can describe the spatial distribution of aerosol density within the construction area, thus calculating the contribution of aerosol particles to the light deflection angle. The light deflection angle caused by aerosols is added to the light deflection angle caused by the air refractive index gradient to obtain the total light deflection angle. This total light deflection angle is then orthogonally transformed with the focal length parameter in the camera's intrinsic parameter matrix to generate a pressure deformation compensation matrix that includes a thermal field disturbance term. This compensation matrix is ​​then used for pixel-by-pixel remapping of the aerial survey images, achieving fine optical distortion compensation.

[0066] Table 2 Comparison of Air Refractive Index and Light Refraction Angle under Different Air Pressure Conditions

[0067]

[0068] In this embodiment, Table 2 shows the air refractive index and the deflection angle of light traveling 1 km under different air pressure and temperature conditions. As can be seen from Table 2, the air refractive index increases with increasing air pressure and decreases with increasing temperature; the light deflection angle is directly proportional to the air refractive index, increasing with increasing air pressure and decreasing with increasing temperature. In high-altitude areas, where air pressure is typically between 60 kPa and 80 kPa, although the light deflection angle is relatively small, it can still cause significant imaging errors for high-precision aerial survey images, necessitating correction.

[0069] In this embodiment, the processing flow for optical distortion compensation and feature point offset correction is as follows: The UAV aerial survey data preprocessing terminal receives real-time air pressure and wind speed data sent by the high-altitude micro-meteorological acquisition node, as well as surface temperature distribution data sent by the infrared thermal imager; the inverse distance weighted interpolation method is used to generate air pressure field, temperature field, and wind speed field data covering the entire construction area; based on the air pressure field and temperature field data, the Edlén formula is used to calculate the air refractive index gradient distribution field; combining the surface temperature distribution data and air pressure data, the Gaussian process regression method is used to construct a space aerosol density field model; Based on the air refractive index gradient distribution field and space aerosol density field model, the total deflection angle of the light propagation path is calculated. Combined with the camera intrinsic parameter matrix, a pressure deformation compensation matrix including thermal field disturbance terms is constructed, and the aerial survey image is remapped pixel-by-pixel to complete optical distortion compensation. Based on wind speed field data, the changes in yaw and pitch angles at the moment of UAV aerial photography are calculated using aerodynamic principles. These changes are mapped to the pixel coordinate system to generate feature point offset vectors. The feature point offset vectors are then used to correct feature points in the aerial survey image, completing feature point offset correction. This embodiment eliminates the disturbances of the high-altitude atmospheric environment on the imaging optical path and camera attitude from a pixel-level physical mechanism, improving the geometric accuracy of the aerial survey image and providing high-quality image data for subsequent extraction of morphological features of the construction surface.

[0070] refer to Figure 3 In one embodiment, the UAV aerial survey data preprocessing unit extracts morphological features of the construction surface based on the corrected aerial survey image. The corrected aerial survey image is input into a multi-scale feature extraction network to obtain the initial texture features and boundary contour features of the construction surface under different receptive fields. A morphological constraint kernel is constructed based on prior knowledge of the pixel distribution of high-altitude site construction materials. The morphological constraint kernel is used to perform non-maximum suppression filtering on the initial texture features and boundary contour features to generate the morphological features of the construction surface.

[0071] In this embodiment, the multi-scale feature extraction network uses ResNet-50 as the backbone network, extracting feature maps of different scales at different stages of the backbone network. The ResNet-50 network contains five convolutional stages: conv1, conv2_x, conv3_x, conv4_x, and conv5_x. The size of the feature map output by each stage decreases sequentially, while the receptive field increases sequentially. Feature maps are extracted from the conv2_x, conv3_x, conv4_x, and conv5_x stages respectively, resulting in four feature maps of different scales, with sizes of 1 / 4, 1 / 8, 1 / 16, and 1 / 32 of the original image, and receptive fields of 28×28, 56×56, 112×112, and 224×224, respectively. These four feature maps of different scales can capture texture and boundary contour information of the construction surface at different scales.

[0072] An altitude gradient sensing branch is introduced into the multi-scale feature extraction network to fuse the UAV flight altitude data corresponding to the corrected aerial survey image with the digital elevation model (DEM) data into an altitude gradient feature map. The UAV flight altitude data is obtained from auxiliary information transmitted back from the UAV flight control system, while the DEM data is pre-stored in the UAV aerial survey data preprocessing unit with a resolution of 1m × 1m. The UAV flight altitude data and the DEM data are registered to obtain the absolute altitude of each pixel. The altitude difference between adjacent pixels is calculated to generate the altitude gradient feature map. The size of the altitude gradient feature map is the same as that of the corrected aerial survey image, and the value of each pixel represents the magnitude of the altitude gradient at that location.

[0073] Cross-channel interactive fusion of the elevation gradient feature map and the deep semantic features of the corrected aerial survey image is performed to generate initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information. The cross-channel interactive fusion is implemented using an attention mechanism, specifically as follows: The elevation gradient feature map is input into a 1×1 convolutional layer, and its number of channels is adjusted to be the same as the number of channels in the deep semantic feature map; the adjusted elevation gradient feature map and the deep semantic feature map are added element-wise to obtain a fused feature map; the fused feature map is input into a global average pooling layer to obtain a channel attention vector; the channel attention vector is multiplied element-wise with the fused feature map to obtain a weighted fused feature map; the weighted fused feature map is input into a 3×3 convolutional layer for feature fusion and dimensionality adjustment to generate initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information.

[0074] A morphological constraint kernel is constructed based on prior knowledge of the pixel distribution of construction materials for high-altitude converter station pavement. High-altitude converter station pavement construction primarily involves earthwork, using materials such as soil, gravel, and concrete. Different construction materials exhibit different pixel distribution characteristics in RGB images; for example, soil typically appears brown or yellow, gravel usually appears gray or white, and concrete usually appears light gray. Through statistical analysis of numerous high-altitude pavement construction images, prior knowledge of the pixel distribution of different construction materials is obtained, including the RGB pixel mean, variance, and pixel distribution range for each material. Based on this prior knowledge, a morphological constraint kernel is constructed. The morphological constraint kernel is a two-dimensional matrix with the same size as the feature map, where each element represents the probability that the location belongs to the construction surface. Locations with pixel values ​​matching the pixel distribution characteristics of the construction materials are assigned higher probability values, while locations with pixel values ​​not matching these characteristics are assigned lower probability values.

[0075] Non-maximum suppression (NMS) filtering is applied to the initial texture features and boundary contour features using a morphological constraint kernel to generate the morphological features of the construction surface. The NMS filtering process is as follows: the initial texture features and boundary contour features are multiplied element-wise with the morphological constraint kernel to obtain weighted texture features and weighted boundary contour features; NMS processing is then applied to both weighted texture features and weighted boundary contour features, retaining local maxima and suppressing non-maxima; finally, the processed weighted texture features and weighted boundary contour features are concatenated to generate the morphological features of the construction surface. The morphological features of the construction surface are high-dimensional vectors containing texture information, boundary contour information, and three-dimensional spatial scale information, enabling a comprehensive and accurate description of the morphological characteristics of the construction surface.

[0076] In this embodiment, edge computing nodes are also deployed at the construction site. These edge computing nodes utilize embedded industrial computers equipped with low-power processors and large-capacity solid-state drives, enabling stable operation in harsh high-altitude environments. A fiber optic communication link is established between the edge computing nodes and the UAV aerial survey data preprocessing unit, receiving the initial texture features and boundary contour features of the construction surface, containing three-dimensional spatial scale information, output by the UAV aerial survey data preprocessing unit. A 5G communication link is established between the edge computing nodes and the progress dynamic simulation server, forwarding the received initial texture features and boundary contour features of the construction surface to the progress dynamic simulation server.

[0077] When the communication link between the edge computing node and the progress dynamic simulation server is interrupted, the edge computing node stores the initial texture features and boundary contour features of the construction surface in a local circular queue buffer. The circular queue buffer is a first-in, first-out (FIFO) data structure that efficiently handles the storage and retrieval of time-series data. The size of the circular queue buffer is set to store 7 days' worth of initial texture features and boundary contour features of the construction surface. The edge computing node timestamps each received feature data and stores it in the circular queue buffer in timestamp order. When the circular queue buffer is full, it automatically overwrites the oldest stored data, ensuring that the buffer always stores the latest feature data.

[0078] Once the communication link between the edge computing node and the progress dynamic simulation server is restored, the edge computing node resumes the transmission of the initial texture features and boundary contour features of the construction surface stored in its local circular queue buffer to the progress dynamic simulation server in timestamp order. The resuming process is as follows: The edge computing node sends a resuming request to the progress dynamic simulation server, which includes the timestamp of the earliest untransmitted data in its local buffer; upon receiving the resuming request, the progress dynamic simulation server queries the timestamp of the latest data stored in its local database and returns its timestamp to the edge computing node; based on the timestamp returned by the progress dynamic simulation server, the edge computing node determines the range of data to be transmitted; the edge computing node transmits the data within the specified range to the progress dynamic simulation server in timestamp order; after the transmission is complete, the edge computing node sends a transmission completion confirmation message to the progress dynamic simulation server; upon receiving the transmission completion confirmation message, the progress dynamic simulation server sends a confirmation response to the edge computing node, and the resuming process ends.

[0079] Table 3 Output feature parameters of the multi-scale feature extraction network at each stage

[0080]

[0081] In this embodiment, Table 3 shows the feature parameters output by the multi-scale feature extraction network at each stage, including feature map size, number of channels, receptive field size, and extracted feature information. As can be seen from Table 3, with the increase of network depth, the feature map size gradually decreases, the number of channels gradually increases, and the receptive field gradually increases, enabling the extraction of feature information at different levels, from fine-grained to coarse-grained. The elevation gradient perception branch can extract the three-dimensional spatial scale information of the construction surface, and by fusing it with the two-dimensional image features extracted by the multi-scale feature extraction network, it can generate more comprehensive and accurate morphological features of the construction surface.

[0082] In this embodiment, the workflow of the construction surface morphological feature extraction and edge computing node is as follows: The UAV aerial survey data preprocessing end inputs the corrected aerial survey image into the multi-scale feature extraction network; the multi-scale feature extraction network extracts feature maps of different scales from different stages, while the altitude gradient perception branch fuses the UAV flight altitude data and digital elevation model data into an altitude gradient feature map; the altitude gradient feature map and deep semantic features are cross-channel interactively fused to generate initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information; a morphological constraint kernel is constructed based on prior knowledge of the pixel distribution of high-altitude site construction materials; the morphological constraint kernel is used to perform non-maximum suppression filtering on the initial texture features and boundary contour features to generate morphological features of the construction surface; the UAV aerial survey data preprocessing end sends the initial texture features and boundary contour features of the construction surface to the edge computing node; the edge computing node forwards the received feature data to the progress dynamic simulation server; when the communication link is interrupted, the edge computing node stores the feature data in a local circular queue buffer; when the communication link is restored, the edge computing node resumes the transmission of the stored feature data to the progress dynamic simulation server according to the timestamp order. This embodiment improves the expressive power of features by introducing an altitude gradient perception branch and integrating three-dimensional spatial scale information into the morphological features of the construction surface. By deploying edge computing nodes and adopting a circular queue buffer and breakpoint resume mechanism, the continuity of data flow in high-altitude and remote areas is ensured.

[0083] refer to Figure 4 In one embodiment, the progress dynamic simulation server inputs the changes in the morphological features of the construction surface and the meteorological attenuation factor into a long short-term memory network. The feature difference matrix of the morphological features of the construction surface in adjacent time series is calculated as the change. Real-time wind speed and air pressure data are mapped to a high-altitude construction efficiency surface, and the meteorological attenuation factor is extracted. The feature difference matrix and the meteorological attenuation factor are concatenated along the feature channel dimension and input into the temporal coding layer of the long short-term memory network.

[0084] In this embodiment, the morphological features of the construction surface are a high-dimensional vector, denoted as... Where t represents the time series and d represents the feature dimension. The feature difference matrix of the morphological features of the construction surface between adjacent time series. The calculation formula is:

[0085]

[0086] Feature difference matrix It reflects the changes in the morphology of the construction surface between two adjacent moments, and includes information on the construction progress.

[0087] Real-time wind speed and air pressure data are mapped to a high-altitude construction efficiency surface, and meteorological attenuation factors are extracted. The high-altitude construction efficiency surface is a two-dimensional surface where the horizontal axis represents wind speed, the vertical axis represents air pressure, and the height of the surface is the corresponding construction efficiency attenuation coefficient. The construction of the high-altitude construction efficiency surface is based on historical meteorological time-series data and corresponding historical mechanical operation efficiency data. Historical meteorological time-series data for the high-altitude converter station apron is obtained, including hourly wind speed and air pressure data for the past 12 months; simultaneously, corresponding historical mechanical operation efficiency data is obtained, including hourly operation data for major construction machinery such as excavators, bulldozers, and road rollers. The historical data is preprocessed to remove outliers and missing values, obtaining valid data samples.

[0088] A surface representing high-altitude construction efficiency was constructed by fitting historical meteorological time-series data with historical mechanical operation efficiency data. Bicubic spline interpolation was employed for surface fitting; bicubic spline interpolation is a high-precision surface interpolation method capable of generating smooth and continuous surfaces. The basic idea of ​​bicubic spline interpolation is to divide the entire data region into several rectangular cells, constructing a bicubic polynomial within each cell, such that the polynomials between adjacent cells have continuous first and second derivatives at the boundaries. The form of the bicubic polynomial is:

[0089]

[0090] in, These are the polynomial coefficients; Wind speed; This refers to air pressure.

[0091] By solving a system of linear equations to determine the polynomial coefficients within each rectangular element, the entire high-altitude construction efficiency surface is obtained.

[0092] Real-time wind speed and air pressure data are used as coordinate indices and input into the high-altitude construction efficiency surface. The dynamic attenuation weight at the current moment is then calculated through interpolation and used as the meteorological attenuation factor. (Meteorological Attenuation Factor) The value ranges from 0 to 1. When weather conditions are favorable, A value close to 1 indicates that construction efficiency is almost unaffected; however, in adverse weather conditions... A value close to 0 indicates that construction efficiency has been severely affected.

[0093] The feature difference matrix With meteorological attenuation factor Concatenate along the feature channel dimension to obtain the concatenated feature vector. Concatenating feature vectors Its dimension is d+1, and its form is:

[0094]

[0095] Concatenate feature vectors The input is the temporal coding layer of a Long Short-Term Memory (LSTM) network. The temporal coding layer of an LSM network consists of multiple LSM units, capable of capturing temporal dependencies in the input data. The LSM units control the flow of information through input gates, forget gates, and output gates, effectively solving the gradient vanishing and gradient exploding problems inherent in traditional recurrent neural networks, making it suitable for processing long-term temporal data.

[0096] refer to Figure 5 In this embodiment, the progress dynamic simulation server compares the predicted construction progress time series data with the baseline schedule to generate scheduling instructions. It calculates the progress deviation vector between the predicted construction progress time series data and the baseline schedule. The predicted construction progress time series data is a vector, denoted as... Where m is the prediction time step, and each element This represents the predicted construction completion amount at the i-th time step. The baseline schedule is a vector, denoted as . Each element This represents the planned construction completion amount at the i-th time step. Schedule Deviation Vector The calculation formula is:

[0097]

[0098] Schedule Deviation Vector Each element in This represents the construction progress deviation at the i-th time step. When, it indicates that the construction progress is lagging behind; when At this time, it indicates that the construction progress is ahead of schedule.

[0099] Determine if the schedule deviation vector exceeds the preset schedule tolerance range. The preset schedule tolerance range is a pre-defined range, denoted as [missing information]. ,in This represents the schedule tolerance threshold. For each element in the schedule deviation vector... ,if If the deviation exceeds the tolerance range, it means that the construction progress deviation at that time step has exceeded the tolerance range and scheduling intervention is required.

[0100] If the schedule deviation vector exceeds the preset schedule tolerance range, the identifiers of critical lagging processes are extracted from the schedule deviation vector. Critical lagging processes refer to those processes that have the greatest impact on the overall construction schedule. By analyzing the construction topology network, the criticality of each process is determined; the higher the criticality, the greater the impact on the overall construction schedule. The identifiers of the several lagging processes with the highest criticality are extracted as critical lagging process identifiers.

[0101] Based on the identification of critical delayed processes, forward and backward tracing is performed in the construction topology network to locate the set of affected related processes. Forward tracing refers to starting from the critical delayed process and searching along the forward edges of the construction topology network to find all subsequent processes that depend on the critical delayed process; backward tracing refers to starting from the critical delayed process and searching along the backward edges of the construction topology network to find all preceding processes that the critical delayed process depends on. All processes obtained from forward and backward tracing are combined to form the set of affected related processes.

[0102] Scheduling instructions are generated based on the resource requirements of the associated process set. These requirements include human resources, machinery and equipment, and materials. The available resource quantity is determined by analyzing the current resource allocation at the construction site. Based on the resource requirements of the associated process set and the available resource quantity, a resource scheduling plan is formulated, and scheduling instructions are generated. These instructions include information such as the type and quantity of resources to be allocated, the allocation time, and the allocation target.

[0103] refer to Figure 6 In this embodiment, an adaptive waypoint planning terminal for the UAV is also included. Deployed within the UAV flight control system, this terminal automatically generates adaptive flight paths for the UAV based on scheduling instructions and meteorological attenuation factors. The adaptive waypoint planning terminal acquires scheduling instructions and meteorological attenuation factors output by the progress dynamic simulation server. It parses the scheduling instructions to extract the coordinates of the area to be monitored, which is typically an area with lagging construction progress and requires focused monitoring.

[0104] The cruise altitude and overlap rate parameters of the UAV are adjusted based on the meteorological attenuation factor. These parameters directly affect the resolution and overlap of aerial survey images, thus influencing the accuracy of subsequent image stitching and feature extraction. When weather conditions are favorable, the meteorological attenuation factor is small, allowing for a higher cruise altitude and a smaller overlap rate to improve aerial survey efficiency. Conversely, when weather conditions are adverse, the meteorological attenuation factor is large, necessitating a lower cruise altitude and a larger overlap rate to ensure the quality of the aerial survey images.

[0105] When the weather attenuation factor exceeds a preset attenuation threshold, the cruise altitude is reduced and the overlap rate parameter is increased to generate an adaptive route. The preset attenuation threshold is a pre-defined value, denoted as [value to be inserted here]. .when When the weather conditions are severe, it indicates that the UAV's aerial survey parameters need to be adjusted. The adjustment range of the cruising altitude is directly proportional to the weather attenuation factor; the larger the weather attenuation factor, the greater the reduction in cruising altitude. The adjustment range of the overlap rate parameter is also directly proportional to the weather attenuation factor; the larger the weather attenuation factor, the greater the increase in overlap rate. Based on the adjusted cruising altitude and overlap rate parameters, combined with the coordinates of the area to be monitored, an adaptive flight path is generated and sent to the UAV flight control system. The UAV flight control system then controls the UAV to perform aerial survey operations according to the adaptive flight path.

[0106] Table 4. UAV aerial survey parameters corresponding to different meteorological attenuation factors.

[0107]

[0108] In this embodiment, Table 4 shows the UAV aerial survey parameters corresponding to different meteorological attenuation factors, including cruising altitude, forward overlap rate, lateral overlap rate, and aerial survey efficiency. As can be seen from the table, with the increase of the meteorological attenuation factor, the cruising altitude gradually decreases, the forward and lateral overlap rates gradually increase, and the aerial survey efficiency gradually decreases. This parameter adjustment strategy can maximize aerial survey efficiency while ensuring the quality of aerial survey images, achieving a balance between aerial survey quality and efficiency.

[0109] In this embodiment, the workflow of dynamic progress simulation and UAV waypoint adaptive planning is as follows: The dynamic progress simulation server receives the morphological features of the construction surface sent by the UAV aerial survey data preprocessing terminal; calculates the feature difference matrix of the morphological features of adjacent time-series construction surfaces; maps real-time wind speed and air pressure data to a high-altitude construction efficiency surface and extracts the meteorological attenuation factor; concatenates the feature difference matrix and the meteorological attenuation factor along the feature channel dimension and inputs it into the Long Short-Term Memory (LSTM) network; the LSM network outputs the construction progress prediction time-series data; calculates the progress deviation vector between the construction progress prediction time-series data and the baseline progress plan; determines whether the progress deviation vector exceeds the preset progress tolerance range; if it does, extracts the key lagging process identifier, performs forward and backward tracing in the construction topology network, and locates the set of affected related processes; generates scheduling instructions based on the resource requirements of the set of related processes; the UAV waypoint adaptive planning terminal obtains the scheduling instructions and the meteorological attenuation factor; parses the scheduling instructions to extract the coordinates of the area to be monitored; adjusts the UAV's cruising altitude and overlap rate parameters based on the meteorological attenuation factor; generates an adaptive flight path and sends it to the UAV flight control system. This embodiment captures the temporal dependency between construction status and meteorological disturbances by inputting meteorological attenuation factors and morphological feature changes into a long short-term memory network, thus enabling advance prediction of construction progress. By performing forward and backward tracing in the construction topology network, the affected related process set is located, realizing the physical logic transformation from passive response to active intervention. Adaptive planning of UAV waypoints ensures the quality of data sources under severe weather conditions.

Claims

1. A smart monitoring and progress management system for the construction of high-altitude converter station site, characterized in that, include: High-altitude micro-meteorological data acquisition nodes are used to obtain real-time wind speed and air pressure data; The UAV aerial survey data preprocessing end is used to perform optical distortion compensation and feature point offset correction on the aerial survey image based on the wind speed and air pressure data, and extract the morphological features of the construction surface based on the corrected aerial survey image. The progress dynamic simulation server is used to input the changes in the morphological characteristics of the construction surface and the meteorological attenuation factor into the long short-term memory network, output the construction progress prediction time series data, and compare the construction progress prediction time series data with the baseline progress plan to generate scheduling instructions.

2. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 1, characterized in that, The UAV aerial survey data preprocessing terminal performs optical distortion compensation and feature point offset correction on the aerial survey image based on the wind speed and air pressure data, including: calculating the air refractive index gradient distribution based on the air pressure data, and constructing an air pressure deformation compensation matrix in combination with the camera intrinsic parameter matrix to perform optical distortion compensation on the aerial survey image. Based on the wind speed data, the changes in yaw and pitch angles at the moment of UAV aerial photography are calculated. The changes in yaw and pitch angles are mapped to the pixel coordinate system to generate feature point offset vectors to correct the feature point offsets in the aerial survey image.

3. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 1, characterized in that, The UAV aerial survey data preprocessing terminal extracts morphological features of the construction surface based on the corrected aerial survey image, including: inputting the corrected aerial survey image into a multi-scale feature extraction network to obtain the initial texture features and boundary contour features of the construction surface under different receptive fields; Based on prior knowledge of the pixel distribution of high-altitude site construction materials, a morphological constraint kernel is constructed. The initial texture features and boundary contour features are then filtered by nonmaximum suppression using the morphological constraint kernel to generate the morphological features of the construction surface.

4. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 1, characterized in that, The progress dynamic simulation server inputs the changes in the morphological features of the construction surface and the meteorological attenuation factor into the long short-term memory network, including: calculating the feature difference matrix of the morphological features of the construction surface in adjacent time series as the changes; The real-time wind speed and air pressure data are mapped to a high-altitude construction efficiency surface, and the meteorological attenuation factor is extracted. The feature difference matrix and the meteorological attenuation factor are concatenated along the feature channel dimension and input into the temporal coding layer of the long short-term memory network.

5. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 4, characterized in that, The progress dynamic simulation server compares the construction progress prediction time series data with the baseline progress plan to generate scheduling instructions, including: calculating the progress deviation vector between the construction progress prediction time series data and the baseline progress plan; Determine whether the schedule deviation vector exceeds the preset schedule tolerance range; if it does, extract the key delayed process identifier from the schedule deviation vector. Based on the key delayed process identifier, forward and backward tracing is performed in the construction topology network to locate the set of affected related processes, and the scheduling instruction is generated according to the resource requirements of the set of related processes.

6. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 2, characterized in that, The air refractive index gradient distribution is calculated based on the air pressure data, and an air pressure deformation compensation matrix is ​​constructed in combination with the camera intrinsic parameter matrix to perform optical distortion compensation on the aerial survey image, including: acquiring surface temperature distribution data of the high-altitude field area, and constructing a space aerosol density field model in combination with the air pressure data. The deflection angle of the light propagation path is calculated based on the space aerosol density field model. The deflection angle is then orthogonally transformed with the focal length parameter in the camera intrinsic parameter matrix to generate the air pressure deformation compensation matrix containing the thermal field disturbance term. The aerial survey image is then remapped pixel by pixel.

7. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 3, characterized in that, The corrected aerial survey image is input into a multi-scale feature extraction network to obtain the initial texture features and boundary contour features of the construction surface under different receptive fields, including: introducing an altitude gradient perception branch into the multi-scale feature extraction network, and fusing the UAV flight altitude data and digital elevation model data corresponding to the corrected aerial survey image into an altitude gradient feature map. The elevation gradient feature map and the deep semantic features of the corrected aerial survey image are cross-channel interactively fused to generate the initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information.

8. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 4, characterized in that, Mapping the real-time wind speed and air pressure data to a high-altitude construction efficiency surface and extracting the meteorological attenuation factor includes: obtaining historical meteorological time series data and corresponding historical mechanical operation efficiency data of the high-altitude converter station apron. The high-altitude construction efficiency surface is constructed by fitting the historical meteorological time-series data with the historical mechanical operation efficiency data. The real-time wind speed and air pressure data are used as coordinate indices and input into the high-altitude construction efficiency surface. The dynamic attenuation weight at the current moment is obtained by interpolation and the dynamic attenuation weight is used as the meteorological attenuation factor.

9. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 8, characterized in that, It also includes an UAV waypoint adaptive planning terminal, which obtains the scheduling instructions and the weather attenuation factor output by the progress dynamic simulation server; The coordinates of the area to be monitored are extracted by parsing the scheduling instructions, and the cruise altitude and overlap rate parameters of the UAV are adjusted based on the meteorological attenuation factor. When the weather attenuation factor exceeds a preset attenuation threshold, the cruise altitude is reduced and the overlap rate parameter is increased to generate an adaptive flight path, which is then sent to the UAV flight control system.

10. The intelligent monitoring and progress management system for high-altitude converter station site construction according to claim 7, characterized in that, It also includes edge computing nodes deployed at the construction site, wherein the edge computing nodes receive the initial texture features and boundary contour features of the construction surface containing three-dimensional spatial scale information output by the UAV aerial survey data preprocessing terminal; When the communication link between the edge computing node and the progress dynamic simulation server is interrupted, the initial texture features and boundary contour features of the construction surface are stored in a local circular queue buffer. After the communication link is restored, the stored initial texture features and boundary contour features of the construction surface are transmitted to the progress dynamic simulation server in the order of timestamps.

Images