Unmanned aerial vehicle image automatic progress and material statistics method and system for capital construction project

Abstract

The application discloses a kind of unmanned aerial vehicle image automation progress and material statistics method and system for capital construction, belong to capital construction management technical field.The method includes: control unmanned aerial vehicle to collect multi-angle image of construction area and generate three-dimensional real scene model;Through model registration and semantic segmentation, identify the progress deviation value of each construction subarea;Material piles are identified simultaneously and the material inventory is calculated;Based on the principle of spatial distance nearest, establish the spatial attribution relationship table between construction subarea and material piles;Query the relationship table to calculate the actual consumption rate of each subarea associated material;Align the progress deviation value and consumption rate, when the same subarea meets progress lag and material consumption rate is lower than threshold value, send early warning;When detecting subarea or pile changes, automatically update the relationship table.The application realizes the linkage analysis of progress and material and subarea-level accurate early warning, solves the problem of progress and material data fragmentation in construction management, improves monitoring efficiency and intelligent level.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of infrastructure project management and UAV remote sensing technology, specifically involving a method and system for automated progress and material statistics using UAV imagery for infrastructure projects. Background Technology

[0002] Construction progress management and material statistics for infrastructure projects (such as roads, bridges, tunnels, dams, and construction sites) are core aspects of project management. Traditional methods primarily rely on manual on-site inspections, manual recording, and comparison with two-dimensional drawings, which have the following technical shortcomings:

[0003] (1) Inefficient and long cycle: Manual inspection requires a lot of manpower. For large-scale or complex construction sites, it is difficult to achieve high-frequency and full-coverage monitoring. The progress statistics cycle is usually in weeks or months, and problems cannot be detected in time.

[0004] (2) Insufficient accuracy and strong subjectivity: There are large errors in estimating the progress and the volume of material piles by manual visual estimation, especially for irregularly shaped piles of sand, gravel, earth and other materials, traditional measurement methods are difficult to accurately calculate their volume and quantity.

[0005] (3) Separation of schedule and material data: In the existing technology, schedule management and material management are often carried out separately by different departments using different tools. There is a lack of unified time and space benchmarks and correlation analysis mechanisms, making it impossible to determine whether the root cause of the schedule delay is a problem of construction efficiency or a problem of material supply shortage.

[0006] (4) High safety risks: There are safety hazards such as falls from heights and mechanical injuries when people enter the construction area for inspection and measurement.

[0007] In recent years, drone aerial photography technology has begun to be applied to engineering monitoring. However, existing solutions mostly remain at the level of image acquisition and simple stitching, lacking an integrated process for automatic comparison of multiple images, calculation of three-dimensional material volume, and linkage analysis of progress and materials. Although some solutions can output progress comparison results or material inventory estimates separately, they fail to establish a spatial mapping relationship between the two, and cannot achieve accurate attribution and early warning at the regional level.

[0008] Therefore, there is an urgent need for a technical solution that can automatically and accurately acquire project progress and material data synchronously, and establish a correlation analysis mechanism between the two. Summary of the Invention

[0009] The present invention aims to provide an automated method and system for tracking progress and materials using UAV imagery in infrastructure projects, in order to solve the technical problems in the prior art, such as the separation of construction progress and material data, the inability to achieve zone-level linkage early warning, and the lack of spatial mapping relationships.

[0010] In a first aspect, embodiments of this application provide an automated method for tracking progress and materials using UAV imagery in infrastructure projects, the method comprising:

[0011] Control the drone to collect multi-angle images of the construction area at regular intervals along the planned route, and generate a three-dimensional real-scene model based on the images;

[0012] The current cycle's 3D reality model is registered with the previous cycle's 3D reality model. The actual construction outline of each construction zone is identified through deep learning semantic segmentation and compared with the planned construction surface to output the progress deviation value of each zone.

[0013] In the same three-dimensional real scene model, the boundaries of the material piles are identified and extracted. The volume of each pile relative to the reference ground is calculated by the spatial triangulation method, and the volume is converted into material inventory based on the material density parameter.

[0014] Based on the principle of closest spatial distance, each construction zone is associated with at least one material pile that is closest to it in the three-dimensional real scene model, and a spatial affiliation table is generated. The table records the material pile identifier associated with each construction zone.

[0015] Query the spatial attribution table to obtain the material stockpile associated with each construction zone. Calculate the actual consumption rate of the material stockpile associated with the construction zone by dividing the difference in material inventory between two consecutive periods by the time interval.

[0016] The spatial affiliation table is queried to align the progress deviation value of each construction zone with the actual consumption rate of the material pile associated with that zone. When the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material pile is lower than the second threshold, an early warning signal is automatically issued for that zone.

[0017] When a newly added or disappeared construction zone or material pile is detected in the three-dimensional reality model, the step of establishing an association based on the principle of closest spatial distance is re-executed, and the spatial affiliation table is automatically updated.

[0018] Secondly, embodiments of this application provide an automated progress and material statistics system for unmanned aerial vehicle (UAV) imagery used in infrastructure projects, applied to the automated progress and material statistics method for UAV imagery used in infrastructure projects as described in the first aspect, the system comprising:

[0019] The data acquisition and modeling module is used to control the drone to collect multi-angle images of the construction area at regular intervals along the planned route, and generate a three-dimensional real-scene model based on the images;

[0020] The progress recognition module is used to register the current cycle's 3D real-world model with the previous cycle's 3D real-world model, identify the actual construction outline of each construction zone through deep learning semantic segmentation, compare it with the planned construction surface, and output the progress deviation value of each zone.

[0021] The material statistics module is used to identify and extract the boundaries of material piles in the same 3D reality model, calculate the volume of each pile relative to the reference ground using the spatial triangulation method, and convert the volume into material inventory based on the material density parameter.

[0022] The relationship table generation module is used to associate each construction zone with at least one material pile that is spatially closest to it in the three-dimensional real scene model based on the principle of spatial proximity, and generate a spatial ownership relationship table. The relationship table records the material pile identifier associated with each construction zone.

[0023] The consumption rate calculation module is used to query the spatial attribution table, obtain the material piles associated with each construction zone, and calculate the actual consumption rate of the material piles associated with the construction zone based on the difference between the material inventory of two consecutive periods divided by the time interval.

[0024] The early warning module is used to query the spatial affiliation table, align the progress deviation value of each construction zone with the actual consumption rate of the material pile associated with that zone, and automatically issue an early warning signal for that zone when the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material pile is lower than the second threshold.

[0025] The relationship table update module is used to automatically update the spatial affiliation relationship table when a newly added or disappeared construction zone or material pile is detected in the three-dimensional real scene model.

[0026] Thirdly, embodiments of this application provide an electronic device, including:

[0027] processor;

[0028] Memory used to store processor-executable instructions;

[0029] The processor is configured to implement, when executing the instructions, the automated progress and material statistics method for unmanned aerial vehicle (UAV) imagery used in infrastructure projects as described in the first aspect.

[0030] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that instructs a device to execute the unmanned aerial vehicle imagery-based automated progress and material statistics method for infrastructure projects as described in the first aspect.

[0031] Compared with existing technologies, this invention has the following advantages: By constructing a spatial attribution table, it establishes for the first time a spatial mapping relationship between construction zones and material stacks in a unified 3D real-world model, enabling correlation analysis of schedule deviations and material consumption under the same spatiotemporal benchmark, filling the gap in the pyramid-style management of schedule and materials in existing technologies. Employing a dual-condition judgment logic of "schedule lag + abnormal material consumption," it can automatically distinguish the root cause of schedule lag (construction efficiency issues or material supply shortages) and issue early warnings for specific construction zones, avoiding general alarms and improving the accuracy and operability of early warnings. When construction zones or material stacks change, the system can automatically detect the changes and re-establish spatial attribution relationships without manual intervention, adapting to the actual needs of dynamic changes in infrastructure projects. Using automated drone data collection and AI intelligent analysis replaces traditional manual inspections and measurements, significantly reducing labor costs and safety risks, while achieving high-frequency, high-precision continuous monitoring. The generated schedule deviation values ​​and material consumption rates can be correlated to digital twin scenarios or time-series databases, supporting trend prediction, material procurement decisions, and construction scheduling optimization, thus facilitating the digital transformation of infrastructure project management. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of a method for automated progress and material statistics of unmanned aerial vehicle (UAV) imagery for infrastructure projects, provided as an embodiment of this application.

[0033] Figure 2 This is an architecture diagram of an automated progress and material statistics system for unmanned aerial vehicle (UAV) imagery used in infrastructure projects, provided as an embodiment of this application.

[0034] Figure 3 A schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0036] It should be noted that in the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0037] Based on the embodiments described in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] To facilitate understanding of the technical solution of this application, the following terms are defined:

[0039] Aerial triangulation: a photogrammetric technique that identifies corresponding feature points in multiple images and calculates the camera position, attitude parameters, and three-dimensional coordinates of the feature points when each image was captured. It is a fundamental step in 3D reconstruction.

[0040] Dense matching: Based on sparse point clouds, the process of performing stereo matching on each pixel in an image to generate a dense 3D point cloud. Commonly used algorithms include multi-view stereo vision and semi-global matching.

[0041] Digital orthophoto: A geometrically corrected vertical projection image that eliminates geometric distortions caused by terrain undulations and camera tilt, with each pixel corresponding to the precise planar coordinates of the actual ground features.

[0042] Digital surface model: A raster data recording the elevation of the top of the Earth's surface and all objects on the surface (including buildings, vegetation, mounds, etc.). Each raster cell stores the elevation value of the highest point at that location.

[0043] Iterative nearest point algorithm: A point cloud registration algorithm that iteratively finds corresponding point pairs between two point clouds, minimizes the sum of squares of the distances between all corresponding point pairs, and solves for the optimal rotation matrix and translation vector to align the two point clouds to the same coordinate system.

[0044] Point cloud registration: The process of aligning two or more point cloud datasets in different coordinate systems to the same reference coordinate system through geometric transformation.

[0045] Semantic segmentation: a technique that classifies each pixel in an image or each point in a point cloud and assigns it a semantic label (such as "completed structure", "material pile", "ground", etc.).

[0046] Spatial Boolean operations: operations such as intersection, union, and difference on three-dimensional entities, used to calculate the overlapping volume or difference between two three-dimensional entities. In this application, it is used to calculate the difference volume between the actual construction outline and the planned model.

[0047] Irregular triangular mesh: A continuous curved surface model composed of a series of non-overlapping triangular facets, used to express the surface morphology in three-dimensional space, and used in this application to construct a three-dimensional surface model of a material pile.

[0048] Reference ground: The original ground elevation surface at the location of the material pile before it is covered by the stockpile, used to deduct the influence of the original terrain when calculating the net volume of the stockpile.

[0049] Euclidean distance: the straight-line distance between two points in three-dimensional space. It is calculated by taking the square root of the sum of the squared differences of the coordinates of the two points in the X, Y, and Z directions.

[0050] Time series database: A database system specifically designed for storing and querying timestamped data, supporting efficient time series data writing, querying, and aggregation analysis. In this application, it is used to store the progress deviation values ​​of each construction zone's historical periods.

[0051] Example 1

[0052] Figure 1 This is a schematic diagram illustrating a method for automated progress and material statistics using drone imagery in infrastructure projects, provided as an embodiment of this application. Figure 1 As shown, a method for automated progress and material statistics using drone imagery for infrastructure projects includes:

[0053] S1. Control the drone to periodically collect multi-angle images of the construction area along a planned flight path, and generate a 3D reality model based on the images. The drone automatically acquires image data of the construction area through aerial photography, and uses photogrammetry technology to reconstruct the 2D images into a 3D digital model with spatial coordinates, serving as the basic data carrier for all subsequent analyses.

[0054] Specifically, in this embodiment, step S1 includes:

[0055] Step 1.1: Based on the construction schedule and monitoring requirements, plan the periodic inspection routes of the UAV in advance.

[0056] Based on the key nodes and monitoring accuracy requirements of different construction stages in the construction schedule, inspection routes are pre-set in the UAV ground station software. For example, for high-rise building construction sites, a daily circular route can be set at a flight altitude of 50 meters to ensure coverage of the entire construction area; for linear projects such as roads or bridges, a strip route can be set twice a week, flying along the construction path. Route parameters include takeoff point, waypoint coordinates, flight speed, camera shooting interval, etc., and are saved as task templates for automatic recall later.

[0057] Step 1.2: Control the drone to fly along the planned route and collect overlapping image sequences of the construction area from multiple angles using circumferential or oblique photography.

[0058] The drone is controlled to fly automatically along a preset route, acquiring images during flight using either orbiting or oblique photography. Orbiting photography is suitable for single buildings or concentrated material storage areas, where the drone flies in a circle around the target area. Oblique photography is suitable for large construction sites, where the camera alternates between multiple pitch angles (e.g., -45°, 0°, +45°) as the drone flies along the route. During acquisition, the overlap between adjacent images must meet the requirements for subsequent processing, with a forward overlap of no less than 80% and a lateral overlap of no less than 70%, to ensure the stability of image matching and the integrity of the 3D reconstruction.

[0059] Step 1.3: Perform aerial triangulation and dense matching processing on the acquired multi-angle images.

[0060] The acquired multi-angle image sequences are imported into 3D reconstruction software. First, aerial triangulation is performed. By identifying corresponding feature points on different images, the camera position, pose parameters, and 3D coordinates of the feature points at the time of image capture are calculated, forming a sparse point cloud. Then, dense matching is performed. Using a multi-view stereo vision algorithm or a semi-global matching algorithm, pixel-by-pixel matching is conducted on the region surrounding each feature point in the sparse point cloud, generating a dense 2D point matrix, where each point contains 3D coordinate information. The multi-view stereo vision algorithm improves matching robustness by utilizing redundant information from multiple viewpoints, while the semi-global matching algorithm aggregates costs in multiple directions to balance matching accuracy and computational efficiency.

[0061] Step 1.4: Synchronously generate digital orthophotos that reflect the true colors and textures of the Earth's surface.

[0062] Based on dense matching, the texture information of multiple images is corrected and fused to eliminate geometric distortions caused by terrain undulations and camera tilt, generating an orthophoto from a vertical perspective. This image has realistic surface colors and textures, and each pixel corresponds to the precise planar coordinates of actual ground features, which can be used for two-dimensional planar analysis and visualization of construction areas.

[0063] Step 1.5: Simultaneously generate a 3D real-world model or digital surface model that expresses the detailed geometry of the earth's surface.

[0064] Based on point cloud data generated through dense matching, a 3D mesh model is constructed using a triangulation algorithm. The texture of the original image is then mapped onto the mesh surface, generating a 3D reality model with realistic color and texture. This model can intuitively display the three-dimensional morphology of the construction area, the 3D shape of completed structures, and the surface undulations of material piles. Alternatively, a digital surface model can be directly generated, which records raster data of the ground surface and the top elevations of all objects on it, for elevation analysis and volume calculation.

[0065] S2. Register the current 3D reality model with the previous 3D reality model. Use deep learning semantic segmentation to identify the actual construction outline of each construction zone and compare it with the planned construction surface to output the progress deviation value of each zone. Spatially align the 3D models from different periods and use artificial intelligence image recognition technology to automatically identify which areas have completed construction and compare them with the preset construction plan to calculate the gap between the actual progress and the plan for each construction area.

[0066] Specifically, in this embodiment, the schedule deviation value is calculated in the following way:

[0067] Step 2.1: When constructing a digital twin scene, perform high-precision point cloud registration between the current period's 3D real scene model and the baseline period's 3D real scene model, and unify them to the same coordinate system.

[0068] Import the 3D reality model generated during the current period and the 3D reality model from the baseline period (such as when the project started or at the previous key node) into the digital twin platform. Use the iterative nearest point algorithm to register the point cloud data in the two models. The basic principle of this algorithm is: first, find the closest corresponding point pairs in the two point clouds; then, solve for the optimal rotation matrix and translation vector by minimizing the sum of the squares of the distances between all corresponding point pairs, transforming the source point cloud to the target point cloud coordinate system. Repeat the above process until convergence. Using unchanged fixed features within the construction area as the registration reference, continuously iterate and optimize the rotation matrix and translation vector between the two point cloud sets to make the spatial position error between them less than centimeters, ultimately unifying the two models into the same construction coordinate system.

[0069] Step 2.2: Using a pre-trained deep learning semantic segmentation model, the registered current cycle model is identified, and the actual three-dimensional contours of the completed construction parts in each construction zone are segmented.

[0070] A pre-trained deep learning semantic segmentation model, trained on a large dataset of construction site images, is loaded. This model employs network structures such as U-Net or DeepLabV3+. U-Net uses a symmetrical encoder-decoder structure, fusing shallow detail features with deep semantic features through skip connections, making it suitable for small-sample training scenarios. DeepLabV3+ uses dilated convolutional pyramid pooling modules to capture multi-scale contextual information, working in conjunction with the decoder to restore spatial resolution, making it suitable for large-scale complex scenes. This model can identify completed structures within construction zones, such as poured concrete piers, installed steel beams, and constructed walls.

[0071] Step 2.3: Obtain the planned construction model or design model of each construction zone from the digital twin scenario.

[0072] In the digital twin scenario, pre-stored planned construction models or BIM design models for each construction zone are invoked. These models contain information on the quantities of work to be completed for each construction zone at planned milestones, such as the volume of concrete to be poured, the area to be excavated, and the number of structural components to be installed. The planned model and the real-world model use the same coordinate system and scale to ensure the accuracy of subsequent comparisons.

[0073] Step 2.4: In three-dimensional space, calculate the volume or area enclosed by the actual three-dimensional contour, compare it item by item with the planned construction volume or area of ​​the corresponding partition, and automatically calculate and output the progress completion percentage and absolute deviation value of each partition.

[0074] For each construction zone, calculate the spatial volume or projected area enclosed by the actual 3D contour, as well as the planned volume or area corresponding to the planned construction model. Divide the actual value by the planned value to obtain the percentage of progress completed for that zone; subtract the planned value from the actual value to obtain the absolute deviation value (positive value indicates over-schedule, negative value indicates under-schedule). Output the calculation results for each zone, for example, "Pier No. 3 zone: planned concrete volume 120 cubic meters, actual completed 98 cubic meters, completion percentage 81.7%, absolute deviation -22 cubic meters", for subsequent use by the early warning module.

[0075] The progress deviation value is further integrated into the digital twin scenario in the following manner:

[0076] Step 2.4.1: Introduce a timeline into the digital twin scenario and automatically overlay and compare the current period's 3D real-world model with the previous period's or design baseline model.

[0077] A four-dimensional scene with a time dimension is constructed within the digital twin platform. The timeline is marked in monitoring cycles, such as "Week 1, Week 2, Week 3" or "April 1, 2026, April 8, 2026". Users can slide along the timeline to view the construction status of different cycles. The system automatically overlays the 3D reality model generated in the current cycle with the reality model of the previous cycle or the initial design baseline model, using semi-transparent or layered rendering methods to make the differences in construction between the two periods immediately apparent.

[0078] Step 2.4.2: Use the deep learning semantic segmentation model to identify newly added structures in the model and accurately extract the actual construction outline of each construction zone in the latest cycle.

[0079] The deep learning semantic segmentation model described in step 2.2 is invoked, and the 3D reality model of the current period and the 3D reality model of the previous period are input simultaneously. By comparing the point cloud or mesh data of the two periods, the model identifies structures that have newly appeared in the current period but did not exist in the previous period, such as newly poured concrete layers, newly installed steel structures, and newly stacked materials. For each construction zone, the model accurately extracts the 3D contour boundaries of these newly added structures, ensuring that only the actual amount of work completed in the current period is counted, avoiding double counting.

[0080] Step 2.4.3: Perform spatial Boolean operations or volume difference calculations on the actual construction outline and the planned schedule model of the corresponding zone to dynamically generate the schedule deviation value expressed as a percentage or quantity.

[0081] The extracted 3D contours of the new structures are compared with the planned schedule models of the corresponding construction zones using spatial Boolean operations. Specifically, the intersection volume or union difference between the two is calculated. For example, if the planned schedule model requires the top pouring of a certain bridge pier to be completed in this cycle, the model calculates the volume of the overlapping part between the new structure and the planned model, which is the actual compliant work volume completed. If the new structure exceeds the scope of the planned model, the excess part is considered over-fulfillment or construction deviation. The calculation results are dynamically generated as a percentage (the proportion of actual completion to planned volume) or a specific work volume value (such as the actual cubic meters of concrete poured).

[0082] Step 2.4.4: Automatically associate the generated schedule deviation value as a key attribute with the corresponding construction zone object in the digital twin scenario to drive the visual early warning in the scenario or directly update the time series database to support more accurate trend prediction.

[0083] The calculated schedule deviation value for each construction zone is stored and bound as an attribute field of that zone object. In the digital twin scenario, zones lagging behind are automatically highlighted in red or flashing, zones on the right track are displayed in green, and zones ahead of schedule are displayed in blue. Simultaneously, the system automatically writes the schedule deviation value for each period into a time-series database, forming a historical progress trajectory curve for each construction zone. Based on this historical data, the system can invoke trend prediction algorithms (such as linear regression or time series analysis) to predict the progress trend of each zone over several future periods, assisting managers in identifying risks in advance and developing intervention measures.

[0084] S3. Within the same 3D reality model, identify and extract the boundaries of material piles. Calculate the volume of each pile relative to the reference ground using the spatial triangulation method, and convert the volume into material inventory based on material density parameters. Automatically identify the location and boundaries of material piles such as sand, gravel, earthwork, and steel reinforcement in the 3D model. Calculate the volume of each pile using geometric calculation methods, and then convert it into the actual quantity of materials based on the material density.

[0085] Step S3 specifically includes:

[0086] Step 3.1: Using a point cloud semantic segmentation algorithm or a 3D instance segmentation network, process the 3D real-world model to identify material piles belonging to different categories and extract the 3D boundary point cloud of each pile. 3D instance segmentation network: A deep learning network capable of simultaneously detecting multiple independent object instances in point cloud data and generating pixel-level semantic labels and instance identifiers for each instance, used in this application to distinguish different individual material piles.

[0087] The generated 3D reality model is converted into point cloud data format and input into a pre-trained point cloud semantic segmentation algorithm (such as PointNet++, RandLA-Net) or a 3D instance segmentation network. PointNet++ extracts local features hierarchically, making it suitable for fine segmentation of irregularly distributed point clouds; RandLA-Net employs random downsampling and local feature aggregation modules, making it suitable for real-time processing of large-scale point clouds. These models can automatically distinguish different types of material piles, such as sand and gravel piles, earthwork piles, rebar piles, and pipe piles, based on the geometric shape, color texture, and spatial distribution characteristics of the point cloud. For each identified independent pile, the algorithm extracts its 3D boundary point cloud, which is the set of all points constituting the outer surface contour of the pile, while simultaneously recording the material category label of each point.

[0088] Step 3.2: Based on the three-dimensional boundary point cloud, construct its corresponding three-dimensional surface model using an irregular triangular mesh, and obtain the net packing volume of each material pile by calculating the volume of the space enclosed between the three-dimensional surface model and the reference ground below it.

[0089] Using the extracted point cloud of the pile boundary as input, an irregular triangular mesh is constructed using the Delaunay triangulation algorithm to generate a 3D surface model of the pile. Delaunay triangulation is a classic triangulation algorithm whose core constraint is that the circumcircle of each triangle does not contain any other points. This characteristic ensures that the generated triangles are as close to equilateral as possible, avoiding the appearance of elongated triangles, thus obtaining the highest quality surface representation. The model consists of a large number of interconnected triangular facets, accurately reproducing the undulating shape of the pile. Simultaneously, the original ground point cloud unaffected by the material pile is extracted from the 3D reality model to construct a reference ground model. Then, the volume of the spatial envelope between the 3D surface model and the reference ground model is calculated, i.e., the volume of the space enclosed vertically from the pile surface to the reference ground, yielding the net packing volume of the material pile. This method can eliminate the influence of the original terrain undulations on the volume calculation.

[0090] Step 3.3: In the material database, preset the standard density parameters for each type of material.

[0091] A material density parameter table is pre-established in the system's material database, setting standard density parameters for each type of material, such as sand and gravel, earthwork, steel reinforcement, concrete, and pipes. For example, the preset density for crushed stone piles is 1.6 tons per cubic meter, for earthwork piles it is 1.4 tons per cubic meter, and for steel reinforcement piles it is 7.85 tons per cubic meter. These parameters can be customized according to the material characteristics of different regions, and users can also modify the parameters or add new material types through the management interface.

[0092] Step 3.4: Based on the calculated net bulk volume and the standard density parameter, automatically calculate the estimated material inventory for each material stockpile, and associate and store this inventory value, material category, calculation time, and unique identifier in the material inventory record.

[0093] For each identified material pile, its net volume is multiplied by the standard density parameter of the corresponding material category to calculate the estimated material inventory, in tons or kilograms. For materials measured by quantity, such as pipes and reinforcing bars, the number of pieces or units is further calculated by combining the standard weight of a single piece. The system automatically generates a unique identifier for each pile and stores this identifier, material category, calculated inventory value, and the calculation timestamp of the current monitoring period in the material inventory record table. This record table supports historical data querying and tracing based on conditions such as time, material category, and pile identifier, providing a data foundation for subsequent consumption rate calculations.

[0094] S4. Based on the principle of closest spatial distance, each construction zone is associated with at least one material pile that is spatially closest to it in the 3D reality model, generating a spatial attribution table. This table records the material pile identifier associated with each construction zone. Following the logic of closest spatial distance, each construction area is bound to nearby material piles, generating a mapping table that records which construction area used materials from which material pile.

[0095] Step S4 specifically includes:

[0096] Step 4.1: For each construction zone identified in the 3D reality model, calculate the 3D spatial Euclidean distance from its geometric center point to the geometric center point of each material pile.

[0097] For each construction zone identified in the 3D reality model, the coordinates of its geometric center point are first calculated, which is the average of the coordinates of all point clouds within the zone. Similarly, the coordinates of the geometric center point of each identified material pile are calculated. Then, for each pair of construction zones and material piles, the 3D Euclidean distance between their geometric center points is calculated, which is the square root of the sum of the squares of the differences in X, Y, and Z coordinates. This distance reflects the proximity of the construction zone and the material pile in actual space; a closer distance indicates more convenient material supply.

[0098] Step 4.2: Based on the calculated distance, establish a candidate set of material stacks for each construction zone and sort them from near to far according to spatial distance.

[0099] For each construction zone, the calculated distance values ​​between it and all material piles are collected to form an initial candidate set of material piles for that zone. Then, using the distance values ​​as the sorting key, all material piles in the candidate set are sorted in ascending order of spatial distance, with the closest piles at the front and the furthest at the back. After sorting, each construction zone has a list of material piles sorted by supply convenience, with piles appearing earlier in the list indicating more direct material supply to that zone.

[0100] Step 4.3: Associate the nearest material pile identifier with the construction zone identifier and record it in the spatial ownership table.

[0101] For each construction zone, select one or more of the closest material piles from the sorted candidate set as the associated objects. For construction zones of normal size, usually only the closest material pile is associated; for large or complex construction zones, two or three of the closest material piles can be associated to handle situations with multiple material sources. Establish a mapping relationship between the selected material pile identifier and the construction zone identifier to form an attribution record.

[0102] Step 4.4: The spatial attribution table shall include at least the following fields: construction zone identifier, list of associated material stack identifiers, and the most recent association update time.

[0103] The spatial attribution table is stored in a structured table format, with each record corresponding to a construction zone. The table contains at least three fields: a construction zone identifier field to uniquely identify each construction zone, which can store a zone number or zone name; a list of associated material pile identifiers field to store unique identifiers of one or more material piles associated with this zone, separated by commas; and a last association update time field to record the timestamp of the most recent establishment or update of this association, accurate to the second. This timestamp is used to determine the timeliness of the association; when a change is detected in a construction zone or material pile, the system can use this field to determine whether the association needs to be recalculated.

[0104] S5. Query the spatial attribution table to obtain the material stockpiles associated with each construction zone. Calculate the actual consumption rate of the material stockpiles associated with each construction zone by dividing the difference in material inventory between two consecutive periods by the time interval. Locate the material stockpiles corresponding to each construction area according to the mapping table. By comparing the changes in material quantity between two adjacent monitoring periods, calculate how much material was actually consumed per unit time in that area.

[0105] The step of calculating the actual consumption rate of the material stockpile associated with the construction zone by dividing the difference between the material inventory in two consecutive periods by the time interval specifically includes:

[0106] Step 5.1: Based on the spatial attribution table, determine at least one material pile that is associated with the target construction zone.

[0107] For each target construction zone where the material consumption rate needs to be calculated, the system first queries the spatial attribution table. It then searches the table for the record corresponding to the construction zone's identifier and reads the list of associated material pile identifiers. For example, the association list for a bridge pier construction zone might contain two identifiers: "Sand and Gravel Pile A" and "Rebar Pile B". If a construction zone is not associated with any material pile in the attribution table, the consumption rate calculation for that zone is skipped, and the anomaly is logged for manual verification.

[0108] Step 5.2: Obtain the inventory records of the associated material stockpile in the current monitoring cycle and the previous monitoring cycle from the material inventory records.

[0109] After identifying the associated material stockpiles, the system accesses the material inventory record table. For each associated stockpile, it queries the inventory records for the current monitoring period and the previous monitoring period based on its unique identifier. The inventory record for the current monitoring period is the latest calculated material inventory value after the completion of this image acquisition and processing. The inventory record for the previous monitoring period is the historical inventory value stored during the previous monitoring. If a stockpile is newly added, the record for the previous period may not exist. In this case, the inventory value for the previous period is set to zero, and the stockpile is marked as a newly added stockpile for subsequent processing reference.

[0110] Step 5.3: Calculate the change in the stock of the associated material pile during the two monitoring periods, and calculate the actual consumption rate of the associated material pile per unit time by combining the actual time interval between the two monitoring periods.

[0111] For each associated material stockpile, the stock value of the previous monitoring period is subtracted from the stock value of the current monitoring period to obtain the material consumption of that stockpile between the two periods. A positive value indicates a decrease in materials, and a negative value indicates an increase in materials. Then, the actual number of days or hours between the two monitoring periods is calculated as the time interval. The material consumption is divided by the time interval to obtain the actual consumption rate per unit time for that stockpile, usually expressed in tons per day or cubic meters per hour. For example, if a sand and gravel stockpile had 500 tons of stock last week and 350 tons this week, with an interval of 7 days, the consumption rate is 21.4 tons per day.

[0112] Step 5.4: When a target construction zone is associated with multiple material piles, calculate the actual consumption rate of each associated pile separately, and use the sum of the rates of each pile as the overall material consumption rate of the construction zone.

[0113] For construction zones that are associated with multiple material piles, the system calculates the actual consumption rate of each associated pile according to steps 5.2 and 5.3. Then, the consumption rates of all associated piles are summed to obtain the overall material consumption rate for that construction zone. For example, a bridge pier zone is associated with sand and gravel piles and steel bar piles. The sand and gravel pile has a consumption rate of 20 tons per day, and the steel bar pile has a consumption rate of 3 tons per day. Therefore, the overall material consumption rate for this zone is 23 tons per day. This overall rate reflects the combined consumption rate of all materials in this zone and is used for subsequent linked early warning analysis with schedule deviation values.

[0114] S6. Query the spatial attribution table to align the progress deviation value of each construction zone with the actual consumption rate of the associated material stockpile. When the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material stockpile is lower than the second threshold, an automatic warning signal is issued for that zone. Match and analyze the progress difference of each construction area with the consumption rate of the corresponding materials in that area. When an area has both slow construction and abnormally low material consumption, an automatic alarm is issued, indicating that there may be a material shortage problem in that area.

[0115] Step S6 specifically includes:

[0116] Step 6.1: For each construction zone, obtain the current progress deviation value from its attribute record, and obtain the actual consumption rate of its associated material pile from the spatial affiliation table and material consumption calculation record.

[0117] The system iterates through all identified construction zones and performs data acquisition operations for each zone. First, it reads the progress deviation value calculated for the current monitoring period from the zone's attribute records. This value may be negative (indicating lag) or positive (indicating lead). Then, it queries the spatial attribution table to find the list of material stack identifiers associated with the zone. Based on these identifiers, it retrieves the actual consumption rate of each associated stack and the overall material consumption rate for the zone from the material consumption calculation records. After acquisition, the progress deviation value and consumption rate are temporarily stored in memory for subsequent analysis.

[0118] Step 6.2: Establish schedule-consumption correlation analysis rules: Set a first threshold for schedule deviation and a second threshold for consumption rate.

[0119] Two threshold values ​​are pre-defined in the system configuration module. The first threshold is used to judge progress deviation, typically set between -5% and -10%. For example, a first threshold of -10% means that when the progress completion percentage falls behind the plan by more than 10%, it is considered a progress delay. The second threshold is used to judge consumption rate, usually determined by multiplying the average consumption rate under normal construction conditions by a coefficient. For example, a second threshold set to 50% of the normal consumption rate means that when the actual consumption rate is less than half of the normal level, it is considered an abnormal material consumption. These two thresholds can be dynamically adjusted by users according to specific project conditions. Dynamic adjustment methods include: users manually modifying the thresholds through the management interface; the system automatically calculating the statistical distribution of the normal consumption rate based on historical monitoring data and setting the second threshold as the mean minus one standard deviation; or using a machine learning regression model to predict a reasonable consumption rate based on factors such as construction stage and weather conditions, and dynamically calculating the threshold. The second threshold can be set internally by either an absolute rate value (e.g., 10 tons per day) or a percentage coefficient relative to the normal consumption rate (e.g., 50%). The system automatically converts the preset normal consumption rate into an absolute threshold for comparison.

[0120] Step 6.3: For each construction zone, determine whether it simultaneously meets the conditions that the schedule deviation value exceeds the preset first threshold and the actual consumption rate of the associated material pile is lower than the preset second threshold.

[0121] For each construction zone, the system executes a two-condition logical judgment. First, condition a is judged: the obtained progress deviation value is compared with a first threshold. If the progress deviation value is less than the first threshold (i.e., the lag exceeds the set standard), then condition a is true. Then, condition b is judged: the obtained overall material consumption rate is compared with a second threshold. If the consumption rate is less than the second threshold, then condition b is true. An alert is triggered only when both conditions are true; if only one condition is true or neither condition is true, no alert is triggered, and the system continues to check the next construction zone.

[0122] Step 6.4: When the same construction zone simultaneously meets conditions a and b, generate an early warning signal containing the construction zone identifier, current progress deviation value, actual consumption rate, and information on the associated material stockpile.

[0123] For a construction zone that meets both conditions simultaneously, the system generates a structured early warning signal. This signal contains at least four elements: a construction zone identifier to indicate which specific area has a problem; the current progress deviation value to quantify the degree of lag; the actual consumption rate to quantify the degree of abnormal material consumption; and information on associated material stockpiles to indicate which material stockpiles have supply or usage problems. The early warning signal also includes a timestamp and early warning level for subsequent classification, processing, and statistical analysis.

[0124] Step 6.5: Push the warning signal to the digital twin scene to trigger a visual warning of highlighting, flashing or color change of the model object corresponding to the construction zone.

[0125] The system pushes the generated warning signals to the digital twin scene in real time via a data interface. Upon receiving the warning signal, the digital twin scene locates the corresponding 3D model object based on the construction zone identifier in the signal and triggers a visual warning effect. Common warning methods include: rendering the zone model as a bright red to make it highly visible in the scene; setting the model to flash at a certain frequency to attract the operator's attention; and gradually changing the model color from yellow to dark red according to the severity of the lag, achieving tiered warnings. The warning effect remains until the problem is resolved or the warning conditions are confirmed to be lifted in the next monitoring cycle.

[0126] S7. When a newly added or disappeared construction zone or material pile is detected in the 3D reality model, the step of establishing an association based on the principle of closest spatial distance is re-executed, and the spatial attribution table is automatically updated. When a new construction area is detected in the 3D model or an old material pile is used up and disappears, the association step is automatically re-executed, and the mapping table is updated to ensure that subsequent analysis is always based on the latest spatial relationships.

[0127] Specifically, in this embodiment, step S7 includes:

[0128] Step 7.1: After the 3D reality model of the current cycle is processed, compare it with the model or construction zone and material stack list of the previous cycle.

[0129] After completing semantic segmentation and object recognition of the 3D reality model for the current period, the system automatically enters the change detection process. It compares all construction zones and material stacks identified in the current period with the results recorded in the previous period. The reference objects for comparison can be the complete 3D reality model from the previous period, or the list of construction zones and material stacks generated at the end of the previous period. The comparison process combines object identifier matching and spatial location matching. Objects with unique identifiers are directly matched by identifier, while newly appearing objects are identified through spatial location and geometric features.

[0130] Step 7.2: Based on the comparison results, identify the newly added construction zones, the disappeared construction zones, the newly added material piles, and the disappeared material piles.

[0131] The system categorizes objects into four types based on the comparison results. Newly added construction zones refer to construction areas that exist in the current cycle but not in the previous cycle, typically corresponding to newly excavated foundation pits, newly constructed work surfaces, or newly divided construction sections. Disappeared construction zones refer to construction areas that existed in the previous cycle but no longer exist in the current cycle, typically corresponding to zones that have been completed and merged into a larger area. Newly added material stockpiles refer to newly appearing material stockpiles in the current cycle, possibly corresponding to newly arrived materials or newly stockpiled materials. Disappeared material stockpiles refer to stockpiles that existed in the previous cycle but no longer exist in the current cycle, typically corresponding to material stockpiles that have been completely consumed. The system records the identifiers and location information of these four types of changed objects in the change log.

[0132] Step 7.3: When a newly added or disappeared construction zone or material pile is detected, the relationship reconstruction process is triggered: Based on the latest 3D reality model, the spatial distance between all currently valid construction zones and all currently valid material piles is recalculated, and the relationship between construction zones and material piles is re-established according to the principle of closest spatial distance.

[0133] The system automatically triggers a relationship reconstruction process whenever any of the four types of changes mentioned above is detected. The reconstruction process first obtains a list of all currently valid construction zones and a list of all currently valid material piles in the latest 3D reality model, removing missing objects and adding new ones. Then, following the methods described in steps 4.1 to 4.4, the system calculates the 3D Euclidean distance from the geometric center point of each construction zone to the geometric center point of each material pile, sorts them by distance, selects the closest material pile to establish a relationship, and re-selects the associated material pile for each valid construction zone. After reconstruction, the original relationships are completely replaced with the new relationships, ensuring that the spatial attribution table always reflects the latest construction site layout.

[0134] Step 7.4: Based on the reconstruction results, generate or update the spatial affiliation table, and record the timestamp and changes of this update.

[0135] The system writes the newly reconstructed relationships into the spatial attribution table. If it's the first time the table is generated, an insertion operation is performed; if it's an update, an overwrite operation is performed. While updating the table, the system records the timestamp of this update, accurate to the second. In addition, the system records the changes in a separate change log table, including which construction zones and material stacks were newly associated, which existing associations were deleted, and which relationships remained unchanged. This change log allows managers to trace the historical evolution of relationships and can also be used for auditing and anomaly detection. After the update is complete, the system uses the new spatial attribution table for subsequent consumption rate calculations and early warning analysis.

[0136] Taking a highway bridge construction site as an example, the site is 2 kilometers long and divided into 20 construction zones, including 10 pier zones, 5 beam zones, and 5 deck zones. There are 3 sand and gravel stockpiles, 2 steel bar stockpiles, and 1 cement stockpile on site.

[0137] The system is configured with drones to automatically perform inspection missions every Monday morning at 9:00 AM, flying at an altitude of 80 meters, with a flight path covering the entire line. During the first monitoring cycle, an initial spatial attribution table was established, associating each pier section with the nearest piles of sand and gravel and steel reinforcement. In the fifth week, the system detected a progress deviation of -15% for pier section 7 (exceeding the first threshold of -10%), while the associated sand and gravel pile's consumption rate was 12 tons per day, lower than 50% of the normal value of 25 tons per day (i.e., 12.5 tons), triggering an early warning signal. On-site inspection revealed that the transport route for the sand and gravel pile was temporarily closed, preventing material delivery. After the problem was resolved, the progress deviation for this section returned to -3% in the sixth week, and the early warning was automatically lifted.

[0138] In week 8, the system detected the addition of the 11th pier zone (a newly excavated work face) and a new sand and gravel pile. It automatically triggered the reconstruction of the spatial ownership table, associating the new zone with the nearest new sand and gravel pile, without manual intervention.

[0139] When the following abnormal situations occur, the system will execute the corresponding processing logic:

[0140] (1) Point cloud registration failure: When the iterative nearest point algorithm fails to converge or the registration error exceeds 5 cm, the system automatically uses manual selection of control points for guided registration and records the failure log for algorithm optimization. (2) Low confidence in semantic segmentation: When the confidence of the deep learning model in recognizing a certain construction zone or material pile is lower than 0.7, the system marks the object as "awaiting manual confirmation" and highlights it in the visualization interface for manual review and correction. (3) Incomplete material pile recognition: When the number of detected pile boundary point clouds is less than 100, the system judges it as incomplete recognition, automatically expands the search range for re-recognition, and if it still fails, it is marked as "awaiting reshoot" and a local flight path for the area is added in the next cycle. (4) No associated material piles in the construction zone: When a certain construction zone is not associated with any material piles in the spatial attribution table, the system does not trigger an early warning, but records the zone as "material source unknown" in the log and prompts the management personnel to check.

[0141] Example 2

[0142] like Figure 2As shown, this application provides an architecture diagram of an automated progress and material statistics system for UAV imagery used in infrastructure projects, applied to the automated progress and material statistics method for UAV imagery used in infrastructure projects as described in Embodiment 1, including:

[0143] The data acquisition and modeling module 210 is used to control the UAV to collect multi-angle images of the construction area at regular intervals according to the planned route, and generate a three-dimensional real scene model based on the images.

[0144] The progress recognition module 220 is used to register the current cycle's 3D real scene model with the previous cycle's 3D real scene model, identify the actual construction outline of each construction zone through deep learning semantic segmentation, compare it with the planned construction surface, and output the progress deviation value of each zone.

[0145] The material statistics module 230 is used to identify and extract the boundaries of material piles in the same three-dimensional real scene model, calculate the volume of each pile relative to the reference ground through the spatial triangulation method, and convert the volume into material inventory based on the material density parameter.

[0146] The relationship table generation module 240 is used to establish an association between each construction zone and at least one material pile that is spatially closest to it in the three-dimensional real scene model based on the principle of spatial proximity, and generate a spatial ownership relationship table. The relationship table records the material pile identifier associated with each construction zone.

[0147] The consumption rate calculation module 250 is used to query the spatial attribution table, obtain the material piles associated with each construction zone, and calculate the actual consumption rate of the material piles associated with the construction zone based on the difference between the material inventory of two consecutive periods divided by the time interval.

[0148] The early warning module 260 is used to query the spatial affiliation table, align the progress deviation value of each construction zone with the actual consumption rate of the material pile associated with that zone, and automatically issue an early warning signal for that zone when the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material pile is lower than the second threshold.

[0149] The relationship table update module 270 is used to automatically update the spatial ownership relationship table when a newly added or disappeared construction zone or material pile is detected in the three-dimensional real scene model.

[0150] Figure 3 This is an electronic device provided in one embodiment of this application. For example... Figure 3 As shown, the electronic device includes at least the following components: processor 301 and memory 300, communication interface 303, and bus 302.

[0151] In this embodiment of the application, memory 300 is used to store executable instructions of processor 301, which, when configured to execute instructions, implements the method as described in the first aspect.

[0152] In embodiments of this application, a computer-readable storage medium includes instructions that instruct a device to perform the method as described in the first aspect. For example, the instructions instruct the device to perform... Figure 1 The method is shown in the process steps.

[0153] In one embodiment of this application, the program operating in the electronic device may be a program that controls a central processing unit (CPU) or similar device to achieve the functions of the above-described embodiments of the present invention (a program that enables the computer to function). Information processed by these systems is then temporarily stored in random access memory (RAM) during processing, and subsequently stored in various ROMs such as read-only memory (FlashROM) and hard disk drives (HDDs), and read, corrected, and written by the CPU as needed.

[0154] It should be noted that a portion of the electronic device described above can also be implemented using a computer. In this case, the program for implementing the control function can be recorded on a computer-readable recording medium, and the program recorded on the recording medium can be read into the computer and executed.

[0155] It should be noted that the computer mentioned here refers to a computer built into an electronic device, employing hardware including an operating system and peripheral devices. Furthermore, computer-readable recording media refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage systems such as hard drives built into the computer.

[0156] Furthermore, computer-readable recording media can include: media that dynamically stores programs for short periods of time, such as communication lines used when transmitting programs via networks like the Internet or communication lines like telephone lines; and media that store programs for fixed periods of time, such as volatile memory inside a computer that serves as a server or client in this case. In addition, the aforementioned program can be a program used to implement the above-mentioned functions, or it can be a program that can implement the above-mentioned functions by combining them with programs already recorded in the computer.

[0157] Furthermore, the electronic device in the above embodiments can also be implemented as an assembly (system group) composed of multiple systems. Each system constituting the system group can possess some or all of the functions or functional blocks of the electronic device in the above embodiments. As a system group, it is sufficient to have all the functions or functional blocks of the electronic device.

[0158] Those skilled in the art should recognize that the above embodiments are only used to illustrate this application and are not intended to limit this application. Any appropriate changes and variations made to the above embodiments within the essential spirit and scope of this application fall within the scope of protection claimed in this application.

Claims

1. A method for automated progress and material statistics using UAV imagery in infrastructure projects, characterized in that, Includes the following steps: Control the drone to collect multi-angle images of the construction area at regular intervals along the planned route, and generate a three-dimensional real-scene model based on the images; The current cycle's 3D reality model is registered with the previous cycle's 3D reality model. The actual construction outline of each construction zone is identified through deep learning semantic segmentation and compared with the planned construction surface to output the progress deviation value of each zone. In the same three-dimensional real scene model, the boundaries of the material piles are identified and extracted. The volume of each pile relative to the reference ground is calculated by the spatial triangulation method, and the volume is converted into material inventory based on the material density parameter. Based on the principle of closest spatial distance, each construction zone is associated with at least one material pile that is closest to it in the three-dimensional real scene model, and a spatial affiliation table is generated. The table records the material pile identifier associated with each construction zone. Query the spatial attribution table to obtain the material stockpile associated with each construction zone. Calculate the actual consumption rate of the material stockpile associated with the construction zone by dividing the difference in material inventory between two consecutive periods by the time interval. The spatial affiliation table is queried to align the progress deviation value of each construction zone with the actual consumption rate of the material pile associated with that zone. When the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material pile is lower than the second threshold, an early warning signal is automatically issued for that zone. When a newly added or disappeared construction zone or material pile is detected in the three-dimensional reality model, the step of establishing an association based on the principle of closest spatial distance is re-executed, and the spatial affiliation table is automatically updated.

2. The method according to claim 1, characterized in that, The steps of controlling the drone to periodically collect multi-angle images of the construction area along a planned route and generating a 3D reality model based on the images specifically include: Based on the construction schedule and monitoring needs, the periodic inspection routes of the drones are planned in advance; the drones are controlled to fly along the planned routes and collect overlapping image sequences of the construction area from multiple angles by means of circling or oblique photography. Aerial triangulation and dense matching processing are performed on the collected multi-angle images to simultaneously generate digital orthophotos that reflect the true colors and textures of the earth's surface, as well as three-dimensional real-scene models or digital surface models that express the fine geometric shapes of the earth's surface.

3. The method according to claim 1, characterized in that, The schedule deviation value is calculated in the following way: When constructing a digital twin scene, the current period's 3D real scene model is registered with the baseline period's 3D real scene model using high-precision point cloud registration, and unified to the same coordinate system; A pre-trained deep learning semantic segmentation model is used to identify the registered current cycle model and segment out the actual three-dimensional contours of the completed construction parts in each construction zone. Obtain the planned construction model or design model of each construction zone from the digital twin scenario; In three-dimensional space, the volume or area enclosed by the actual three-dimensional contour is calculated and compared with the planned construction volume or area of ​​the corresponding partition item by item. The progress completion percentage and absolute deviation value of each partition are automatically calculated and output.

4. The method according to claim 3, characterized in that, The progress deviation value is further integrated into the digital twin scenario in the following manner: A timeline is introduced into the digital twin scenario to automatically overlay and compare the current period's 3D real-world model with the previous period's or design baseline model. The deep learning semantic segmentation model is used to identify newly added structures in the model and to accurately extract the actual construction outline of each construction zone in the latest cycle. The actual construction outline is compared with the planned schedule model of the corresponding zone by performing spatial Boolean operations or volume difference calculations to dynamically generate a schedule deviation value expressed as a percentage or quantity of work. The generated schedule deviation values ​​are automatically associated as key attributes with the corresponding construction zone objects in the digital twin scenario, which can be used to drive visual early warnings in the scenario or directly update the time series database to support more accurate trend prediction.

5. The method according to claim 1, characterized in that, The steps of identifying and extracting the material pile boundaries, calculating the volume of each pile relative to the reference ground using the spatial triangulation method, and converting the volume into material inventory based on the material density parameters specifically include: The three-dimensional real-world model is processed using point cloud semantic segmentation algorithms or three-dimensional instance segmentation networks to identify material piles belonging to different categories and extract the three-dimensional boundary point cloud of each pile. Based on the three-dimensional boundary point cloud, an irregular triangular mesh is used to construct its corresponding three-dimensional surface model, and the net stacking volume of each material stack is obtained by calculating the volume of the space enclosed between the three-dimensional surface model and the reference ground below. In the materials database, standard density parameters are preset for each type of material; Based on the calculated net stacking volume and the standard density parameter, the estimated material inventory of each material stack is automatically calculated, and the inventory value, material category, calculation time, and unique identifier are associated and stored in the material inventory record.

6. The method according to claim 1, characterized in that, The step of establishing a spatial attribution table by associating each construction zone with at least one material pile that is spatially closest to it in the 3D reality model based on the principle of spatial proximity specifically includes: For each construction zone identified in the 3D reality model, calculate the 3D spatial Euclidean distance from its geometric center point to the geometric center point of each material pile. Based on the calculated distance, a candidate set of material stacks is established for each construction zone, and sorted from near to far according to spatial distance; Associate the nearest material pile identifier with the construction zone identifier and record it in the spatial ownership table; The spatial affiliation table includes at least the following fields: construction zone identifier, list of associated material stack identifiers, and the most recent association update time.

7. The method according to claim 1, characterized in that, The step of calculating the actual consumption rate of the material stockpile associated with the construction zone by dividing the difference between the material inventory in two consecutive periods by the time interval specifically includes: Based on the spatial attribution table, identify at least one material pile that is associated with the target construction zone; Obtain the inventory records of the associated material stockpile in the current monitoring cycle and the previous monitoring cycle from the material inventory records; Calculate the change in the stock of the associated material pile over two monitoring periods, and combine this with the actual time interval between the two monitoring periods to calculate the actual consumption rate of the associated material pile per unit time. When a target construction zone is associated with multiple material stacks, the actual consumption rate of each associated stack is calculated separately, and the sum of the rates of all stacks is taken as the overall material consumption rate of the construction zone.

8. The method according to claim 1, characterized in that, The step of aligning the progress deviation value of each construction zone with the actual consumption rate of the associated material stockpile for that zone, and automatically issuing a warning signal for that zone when the same construction zone simultaneously meets the conditions of a progress deviation value exceeding a first threshold and an actual consumption rate of the associated material stockpile being lower than a second threshold, specifically includes: For each construction zone, the current progress deviation value is obtained from its attribute record, and the actual consumption rate of its associated material pile is obtained from the spatial attribution table and material consumption calculation record. Establish schedule-consumption correlation analysis rules: set a first threshold for schedule deviation and a second threshold for consumption rate; For each construction zone, determine whether it simultaneously meets the following two conditions: a) The progress deviation value exceeds the preset first threshold, indicating that the construction progress is significantly behind schedule; b) The actual consumption rate of the associated material pile is lower than the preset second threshold, indicating that the material supply or usage efficiency is insufficient; When the same construction zone simultaneously meets both conditions a) and b), an early warning signal is generated that includes the construction zone identifier, the current progress deviation value, the actual consumption rate, and information on the associated material stockpile. The warning signal is pushed to the digital twin scene, triggering a visual warning of highlighting, flashing, or color change of the model object corresponding to the construction zone.

9. The method according to claim 1, characterized in that, The step of automatically updating the spatial attribution table by re-executing the step of establishing associations based on the principle of closest spatial distance when a newly added or disappeared construction zone or material pile is detected in the three-dimensional real-scene model specifically includes: After the current cycle's 3D reality model is processed, it is compared with the previous cycle's model or construction zone and material stockpile list; Based on the comparison results, newly added construction zones, disappeared construction zones, newly added material piles, and disappeared material piles were identified. When a newly added or disappeared construction zone or material pile is detected, the relationship reconstruction process is triggered: based on the latest 3D real scene model, the spatial distance between all currently valid construction zones and all currently valid material piles is recalculated, and the relationship between construction zones and material piles is re-established according to the principle of closest spatial distance. Based on the reconstruction results, generate or update the spatial affiliation table, and record the timestamp and changes of this update.

10. An automated progress and material statistics system for infrastructure projects using drone imagery, applied to the method described in any one of claims 1 to 9, characterized in that, The system includes: The data acquisition and modeling module is used to control the drone to collect multi-angle images of the construction area at regular intervals along the planned route, and generate a three-dimensional real-scene model based on the images; The progress recognition module is used to register the current cycle's 3D real-world model with the previous cycle's 3D real-world model, identify the actual construction outline of each construction zone through deep learning semantic segmentation, compare it with the planned construction surface, and output the progress deviation value of each zone. The material statistics module is used to identify and extract the boundaries of material piles in the same 3D reality model, calculate the volume of each pile relative to the reference ground using the spatial triangulation method, and convert the volume into material inventory based on the material density parameter. The relationship table generation module is used to associate each construction zone with at least one material pile that is spatially closest to it in the three-dimensional real scene model based on the principle of spatial proximity, and generate a spatial ownership relationship table. The relationship table records the material pile identifier associated with each construction zone. The consumption rate calculation module is used to query the spatial attribution table, obtain the material piles associated with each construction zone, and calculate the actual consumption rate of the material piles associated with the construction zone based on the difference between the material inventory of two consecutive periods divided by the time interval. The early warning module is used to query the spatial affiliation table, align the progress deviation value of each construction zone with the actual consumption rate of the material pile associated with that zone, and automatically issue an early warning signal for that zone when the same construction zone simultaneously meets the conditions that the progress deviation value exceeds the first threshold and the actual consumption rate of the associated material pile is lower than the second threshold. The relationship table update module is used to automatically update the spatial affiliation relationship table when a newly added or disappeared construction zone or material pile is detected in the three-dimensional real scene model.

Images