Construction progress evaluation method and system based on unmanned aerial vehicle visual collaboration, electronic equipment and storage medium

Abstract

The invention relates to the technical field of construction progress evaluation, in particular to a construction progress evaluation method and system based on unmanned aerial vehicle visual collaboration, electronic equipment and a storage medium, and the method comprises the steps: carrying out the image video data collection of a construction site based on a predetermined flight path and a time interval, and carrying out the geometric correction preprocessing; identifying construction equipment in the preprocessed image video data by using a target detection algorithm, and extracting related features of target construction equipment; constructing a construction progress evaluation model based on the construction equipment layout based on the graph neural network, and extracting an initial construction progress evaluation result according to the related features of the target construction equipment; and adjusting parameters of the construction progress evaluation model by using a zero-order feedback optimization algorithm based on the initial construction progress evaluation result and considering the uncertainty of the layout of the construction equipment. The objective of the invention is to solve the technical problem that the accuracy of construction progress evaluation is insufficient due to the uncertainty of the layout of construction equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of construction progress assessment, and in particular to a construction progress assessment method, system, electronic device and storage medium based on unmanned aerial vehicle visual collaboration. Background Art

[0002] In the field of construction, accurate construction progress assessment is extremely critical for project management, resource allocation and on-time delivery. With the rapid development of science and technology, drone technology has been increasingly widely used in construction progress assessment. The realization of construction progress assessment based on drones mainly relies on their powerful image acquisition and data transmission capabilities. Drones can collect all-round, high-resolution image and video data of the construction site according to the preset flight path and time interval. By analyzing the collected data, various types of construction equipment on the construction site, completed construction areas, and the stacking of construction materials can be identified.

[0003] At present, the common technical means of using drones for construction progress assessment in existing technologies mainly include image recognition technology and data comparison and analysis. First, the image video data collected by drones is processed using advanced target detection algorithms to identify target objects such as construction equipment and construction areas. Subsequently, the recognition results are compared and analyzed with the pre-set construction plan or historical data to infer the current construction progress. For example, the changes in the outline of the building, the position movement of the construction equipment, and the increase or decrease of construction materials in the images of different periods are compared to determine the progress of the construction progress. However, the layout of construction equipment at the construction site is not fixed, but is in dynamic change. For example, due to the adjustment of construction technology, the change of construction tasks or sudden equipment failures, construction equipment may be frequently moved, added or evacuated within the construction site. This uncertainty makes the construction progress assessment based on drone-collected data challenging. Traditional assessment methods often find it difficult to capture the impact of these equipment layout changes on the construction progress in real time and accurately, resulting in the inability of the assessment method to timely and accurately reflect such equipment layout changes in the construction progress assessment results, resulting in deviations between the assessment results and the actual construction progress, and unable to provide a reliable decision-making basis for construction project management. Summary of the invention

[0004] In order to solve the technical problem of insufficient accuracy of construction progress assessment caused by uncertainty in construction equipment layout, the present invention provides a construction progress assessment method, system, electronic device and storage medium based on UAV visual collaboration. The technical solutions adopted are as follows:

[0005] The technical solution of the first aspect of the present invention provides a construction progress assessment method based on UAV visual collaboration, the method comprising:

[0006] Capture image and video data of the construction site based on a predetermined flight path and time interval and perform geometric correction preprocessing;

[0007] Use the target detection algorithm to identify the construction equipment in the preprocessed image video data and extract the relevant features of the target construction equipment;

[0008] A construction progress assessment model based on construction equipment layout is built based on graph neural network, and the initial construction progress assessment results are extracted according to the relevant features of the target construction equipment;

[0009] Based on the initial construction progress assessment results and considering the uncertainty of construction equipment layout, the parameters of the construction progress assessment model are adjusted using a zero-order feedback optimization algorithm.

[0010] Furthermore, collecting image and video data of the construction site based on the predetermined flight path and time interval and performing geometric correction preprocessing includes:

[0011] Divide the construction site into areas according to the characteristics of the construction process and the functional differences of the construction areas;

[0012] Dynamically adjust the collection time interval within the target area based on the movement speed of construction equipment and the intensity of construction activities;

[0013] The image and video data are corrected according to the preset ground control points in the target area.

[0014] Furthermore, the target detection algorithm is used to identify the construction equipment in the preprocessed image video data, and the relevant features of the target construction equipment are extracted, including:

[0015] Obtain the rectified image and video data of each region and train the target detection model;

[0016] Extract the layout feature vector of construction equipment based on the trained object detection model;

[0017] The differences between the characteristic vectors of construction equipment layout at different times are calculated, and the parameters affecting the construction progress due to equipment layout changes are extracted.

[0018] Furthermore, the layout feature vector of the construction equipment is extracted based on the trained target detection model, including:

[0019] Extract location features based on the distance between construction equipment and key construction points;

[0020] Extracting operation status features according to the operation status changes of construction equipment at different time periods;

[0021] The correlation features are extracted according to the difference in the degree of correlation between construction equipment at different times.

[0022] Furthermore, a construction progress evaluation model based on construction equipment layout is constructed based on graph neural network, and the initial construction progress evaluation results are extracted according to the relevant characteristics of the target construction equipment, including:

[0023] The construction equipment is set as graph nodes, the edge connection relationship between nodes is determined based on the spatial distance and functional relevance, and the construction equipment layout graph structure is constructed and trained;

[0024] Input the current construction equipment layout diagram structure into the trained graph neural network model and output the initial construction progress assessment result;

[0025] According to the parameters affecting the construction progress caused by the changes in equipment layout, adjustment factors are extracted to optimize the initial construction progress assessment results.

[0026] Furthermore, based on the initial construction progress assessment results and considering the uncertainty of the construction equipment layout, the parameters of the construction progress assessment model are adjusted using the zero-order feedback optimization algorithm, including:

[0027] Extract key visual information based on image and video data, and calculate the estimated gradient values ​​at different times according to the objective function of the construction progress assessment model;

[0028] According to the estimated gradient values ​​of the objective function at different times, combined with the layout and collaboration relationship changes caused by the current device movement and the entry of new devices, the gradient update step size at the next moment is extracted;

[0029] The parameters of the construction progress evaluation model are iteratively updated according to the gradient update step size at the next moment.

[0030] Furthermore, the expression of the gradient update step size is:

[0031]

[0032] In the formula, a t It represents the gradient update compensation at time t, that is, the amplitude of parameter adjustment in the construction progress evaluation model; It represents the estimated value of the gradient of the objective function of the construction progress evaluation model at time t; It represents the estimated value of the gradient of the objective function of the construction progress evaluation model at time t-1; ΔC represents the change coefficient of the layout and collaboration relationship caused by the movement of equipment and the arrival of new equipment; λ represents the smoothing coefficient.

[0033] The technical solution of the second aspect of the present invention provides a construction progress assessment system based on UAV visual collaboration, using the construction progress assessment method based on UAV visual collaboration described in the technical solution of the first aspect of the present invention, and the system includes:

[0034] A data acquisition and preprocessing module, configured to acquire image and video data of the construction site based on a predetermined flight path and time interval and perform geometric correction preprocessing;

[0035] A target detection module configured to identify the construction equipment in the preprocessed image video data using a target detection algorithm and extract relevant features of the target construction equipment;

[0036] a construction progress assessment module, configured to construct a construction progress assessment model based on construction equipment layout based on a graph neural network, and extract an initial construction progress assessment result based on relevant features of the target construction equipment;

[0037] The construction progress assessment optimization module is configured to adjust the parameters of the construction progress assessment model using a zero-order feedback optimization algorithm based on the initial construction progress assessment results and taking into account the uncertainty of the construction equipment layout.

[0038] The technical solution of the third aspect of the present invention provides an electronic device, comprising: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can execute the steps of the construction progress assessment method based on drone visual collaboration described in the technical solution of the first aspect of the present invention.

[0039] The technical solution of the fourth aspect of the present invention provides a computer-readable storage medium, on which is stored a program for implementing a construction progress assessment method based on drone visual collaboration. The program for implementing a construction progress assessment method based on drone visual collaboration is executed by a processor to implement the steps of the construction progress assessment method based on drone visual collaboration described in the technical solution of the first aspect of the present invention.

[0040] The present invention has the following beneficial effects:

[0041] The construction progress assessment method based on UAV visual collaboration provided by the present invention can identify construction equipment and extract relevant features through a target detection algorithm, and can capture the dynamic changes of individual equipment; then a construction progress assessment model is constructed based on a graph neural network, and the graph neural network can set the construction equipment as graph nodes and construct edge connection relationships to reflect the overall logic and mutual influence of the construction equipment layout, and then obtain the initial construction progress assessment result based on the equipment layout characteristics; at the same time, the present invention takes into account the uncertainty of the construction equipment layout and uses a zero-order feedback optimization algorithm to adjust the model parameters; the zero-order feedback optimization algorithm extracts key visual information based on image video data to calculate gradient estimates, and can keenly perceive the impact trend of equipment layout changes on the evaluation results; by iteratively updating the parameters of the construction progress assessment model, the construction progress assessment model can adapt to the uncertainty of the construction equipment layout, and continuously optimize the evaluation results, thereby It effectively improves the accuracy of construction progress assessment when facing changes in equipment layout, and provides a more reliable decision-making basis for construction project management; secondly, the parameters of the construction progress assessment model are adjusted by using the zero-order feedback optimization algorithm, and the uncertainty of the construction equipment layout is considered to dynamically update the model parameters. It can continuously adjust and optimize itself according to the actual construction situation to adapt to complex and changeable construction scenes. Even if the accuracy of the initially constructed construction progress assessment model is insufficient, it can be dynamically adjusted to fit the actual construction progress. This method does not need to construct an accurate construction progress assessment model, but only relies on the information at different times obtained from the UAV visual data feedback to calculate the gradient estimate, avoiding complex display gradient calculations, reducing the amount of calculation, and is easy to deploy and implement at various construction sites. It can achieve effective evaluation of the construction progress when there is uncertainty in the construction process and the model is difficult to accurately construct. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A method flow chart of a construction progress assessment method based on UAV visual collaboration provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of the structure of a construction progress assessment system based on UAV visual collaboration provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a construction progress assessment method, system, electronic device and storage medium based on drone visual collaboration proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0047] The following is a detailed description of a construction progress assessment method, system, electronic device and storage medium based on UAV visual collaboration provided by the present invention in conjunction with the accompanying drawings.

[0048] See also Figure 1 , which shows a method flow chart of a construction progress assessment method based on UAV visual collaboration provided by an embodiment of the present invention, the method comprising:

[0049] Step S100: Based on the predetermined flight path and time interval, image and video data of the construction site are collected and geometric correction preprocessing is performed; specifically, according to the scale, shape and key focus areas of the construction site, a predetermined flight path is formulated using professional drone flight path planning software; for example, for a large rectangular construction site, a main path can be set to fly around the boundary of the construction site, and cross-branch paths can be planned in the key internal construction areas (such as foundation pouring areas, tower crane operation areas, etc.) to ensure full coverage of the construction site and more intensive collection points for key areas; the flight altitude is determined according to the building height, equipment height and required field of view of the construction site, generally at a relatively safe altitude where clear images can be obtained, such as between 30-100 meters, and local drone flight airspace control regulations must be followed;

[0050] Step S100 specifically includes:

[0051] Step S110: Divide the construction site into areas according to the characteristics of the construction process and the functional differences of the construction areas; specifically, taking the main structure construction in the building construction as an example, in the main structure construction area, the main structure construction can be further divided into: concrete pouring sub-area, steel bar binding sub-area, formwork erection and dismantling sub-area and steel structure installation sub-area according to the construction process and function; wherein the concrete pouring sub-area involves the parking of concrete-related equipment and the scope of operation; the steel bar binding sub-area involves the equipment distribution of steel bar processing and lifting machinery; the formwork erection and dismantling sub-area is an area for storing, assembling and disassembling various types of building formwork; the steel structure installation sub-area is mainly the operation scope of steel component stacking, lifting and splicing and installation;

[0052] Step S120: dynamically adjusting the collection time interval within the target area based on the dynamic time interval adjustment mechanism of the horizontal component of the construction equipment moving speed and the construction activity intensity;

[0053] Step S120 specifically includes:

[0054] Step S121: Using the image and video data collected by the drone, the target detection and tracking algorithm is used to lock the key equipment in the target area, and the displacement of the key equipment in a unit time is extracted; specifically, the high-performance image acquisition equipment carried by the drone is used to collect image and video data of the target area according to the predetermined flight path and initial time interval; in the data processing stage, the YOLO series algorithm or the Faster R-CNN algorithm based on deep learning is used to analyze the collected data, and the key construction equipment such as tower cranes, concrete pumps, and cranes in the target area are accurately locked; by comparing the position information of the key equipment in the continuous frame images, combined with the timestamp of the image and the known image resolution and the actual geographic space ratio, the displacement of the key equipment in a unit time, including the direction and distance of the displacement, is calculated, and the displacement data is recorded for subsequent analysis;

[0055] Step S122: extracting the construction activity intensity level of the target area based on the construction personnel density and material transportation frequency in the target area; specifically, for the extraction of construction personnel density, image semantic segmentation technology can be used to separate the construction personnel in the drone collected images from background elements such as building structures, construction equipment, materials, etc., and count the number of construction personnel per unit area, which is used as an indicator of construction personnel density; the material transportation frequency can be calculated by performing statistical analysis on the number of appearances and time intervals of material transportation vehicles in the image video data, and calculating the average interval time. If the interval time is short and the number of times is large, it indicates that the material transportation frequency is high. Based on the construction personnel density and material transportation frequency information, the construction activity intensity level of the target area is divided into three levels of high, medium and low or more refined quantitative values ​​according to pre-set quantitative standards.

[0056] Step S123: dynamically adjust the collection time interval according to the displacement of key equipment in the target area per unit time and the intensity level of the construction activity; specifically, according to the construction characteristics and equipment operation rules in the target area, for example, in the concrete pouring sub-area, if the moving speed of the concrete delivery pump equipment exceeds a certain threshold and the quantified value of the construction activity intensity is greater than the set high-intensity threshold, it means that the construction progress in this area changes rapidly and more frequent data collection is required; on the contrary, if the equipment moves slowly and the intensity of the construction activity is low, it means that the construction state is stable and the collection time interval can be appropriately extended to;

[0057] In summary, step S120 can make the time interval for drone data collection closely match the actual construction progress by constructing a dynamic time interval adjustment mechanism; in the stage where the construction progress changes rapidly, such as frequent movement of equipment and large-scale investment of construction personnel and materials, the frequency of data collection can be increased, so as to obtain key construction change information in time, and provide rich and timely data support for subsequent accurate construction progress evaluation, avoiding misjudgment of construction progress due to data collection delays; and in periods when construction is relatively stable and progress changes slowly, appropriately extending the collection time interval can effectively reduce unnecessary data collection, reduce the burden of data transmission, storage and processing, and improve the operating efficiency and resource utilization of the entire construction progress evaluation system, ensuring that the system can operate efficiently and accurately under different construction rhythms, adapt to the complex and changeable construction site environment, and improve the accuracy and timeliness of construction progress evaluation.

[0058] Step S130: Correcting the image and video data according to the preset ground control points in the target area;

[0059] Step S130 specifically includes:

[0060] Step S131: ground control points are set at key locations in the target area; specifically, ground control points are set in a targeted manner according to the characteristics of different construction sub-areas, for example, in the concrete pouring sub-area, ground control points are set at the four corners of the pouring site; in the steel bar binding sub-area, ground control points are set at fixed markers of the steel bar processing shed; in the formwork erection and dismantling sub-area, ground control points are set at the nodes of the erected stable structural frame; in the steel structure installation sub-area, ground control points are arranged at the boundaries of the steel component stacking site;

[0061] Step S132: using an image feature recognition algorithm to identify the pixel coordinates of the ground control points in the target area, and based on the principle of photogrammetry, establishing a collinearity equation model between the pixel coordinates of the target area image and the geodetic coordinates, the collinearity equation model can describe the geometric relationship between the image point, the projection center and the corresponding ground point; then, by substituting the known geodetic coordinates of the ground control points and their corresponding pixel coordinates into the collinearity equation model, the least squares method can be used to solve the unknown parameters in the model, such as the internal orientation elements and external orientation elements of the camera, and the unknown parameters are used as correction parameters to perform geometric correction on the image video data, thereby eliminating the image deformation and coordinate deviation caused by factors such as the change of the UAV's flight attitude, terrain undulations and lens distortion, so that the image can accurately reflect the real position and shape information of the construction object in the target area;

[0062] Step S200: using a target detection algorithm to identify the construction equipment in the preprocessed image video data, and extracting relevant features of the target construction equipment;

[0063] Step S200 specifically includes:

[0064] Step S210: Obtain the corrected image and video data of each area and train the target detection model; specifically, first collect a large amount of corrected image and video data of each construction area, then annotate these image and video data to accurately mark the location and category information of key construction equipment in the target area; then use a deep learning framework to build and train the target detection model, and you can choose a classic target detection algorithm structure, such as Faster R-CNN, YOLO series, etc., to ensure that the model can accurately identify construction equipment in different scenarios.

[0065] Step S220: extracting the layout feature vector of the construction equipment based on the trained target detection model; specifically, the layout feature vector of the construction equipment includes: establishing the relationship between the equipment position and the key construction point, the operating status and the related features of the number of associated equipment; for the relationship between the equipment position and the key construction point, according to the distance between the construction equipment and the key construction point, the position feature is extracted by constructing the position influence function related to the distance from the equipment to the key construction point, and the influence function can be expressed as:

[0066]

[0067] Where, d i represents the distance from the i-th equipment to the key construction point; k represents the condition coefficient. The closer the equipment position is to the key construction point, the greater the weight of the impact on the construction progress;

[0068] For the operating status, the operating time of each construction equipment is determined by analyzing the time series of the image video data collected by the drone; for example, the changes in the hook movement of the tower crane, the rotation of the mixing drum of the concrete mixer truck and other features in the continuous frames are observed to determine the start time and duration of the equipment's operation. An influence function is constructed based on the principle that the longer the operating time, the greater the impact of the change in its operating status on the construction progress. This function can be customized according to the equipment type and construction process requirements. For example, for some key lifting equipment, an exponential growth function can be used, that is, as the operating time increases, the weight of the impact of its operating status change on the progress increases rapidly. By substituting the operating time of the equipment into the corresponding influence function, the operating status characteristic value of each equipment is obtained;

[0069] For the association feature, the difference between the characteristic vectors of the construction equipment layout at different times is calculated, and then the influence function is constructed according to the change of the equipment association relationship with more associated equipment. The association feature can be expressed as:

[0070]

[0071] In the formula, m i represents the associated features of the i-th device; R ij It indicates the degree of association between the ith device and the jth device. Its value range is usually set in the interval [0,1]. The closer the value is to 1, the closer the association between the devices is. For example, in the concrete pouring sub-area, the material transportation between the concrete pump and the concrete mixer truck is closely connected, and the value between them will be relatively high; and the closer the value is, the weaker the association is. t indicates the time; f rel (n i ) represents the number of device data associated with the i-th device n i The association influence function is used to dynamically adjust the weight of the impact of association relationship changes on the construction progress according to the number of other equipment associated with the equipment;

[0072] Step S230: Calculate the differences between the characteristic vectors of the construction equipment layout at different times, and extract the parameters affecting the equipment layout changes on the construction progress; specifically, record the layout characteristic vectors of the construction equipment in the target area at different times; then construct an equipment layout change matrix, in which each row of the matrix represents the change in the layout characteristic vector of a device at different times. Then, perform singular value decomposition on the equipment layout change matrix, which can decompose the matrix into the product of three matrices, where the singular values ​​reflect the important characteristic information of the matrix. Take the maximum singular value obtained by the decomposition as the parameter affecting the equipment layout changes on the construction progress. This parameter can comprehensively reflect the overall degree of change of the construction equipment layout at different times and its potential impact on the construction progress, and provide an important basis for the construction and adjustment of the subsequent construction progress evaluation model;

[0073] In summary, step S200 can comprehensively and deeply mine the information related to the construction progress in the drone image video data by systematically training the target detection model and extracting the layout feature vector of the construction equipment. The layout feature vector constructed based on the position influence function, the operating status influence function and the association influence function can accurately quantify the influence weight of the characteristics of the construction equipment in terms of position, operating status and association relationship and their changes on the construction progress, avoiding the limitations of the simple qualitative analysis of the construction equipment information in the traditional evaluation method; by calculating the influence parameters obtained by the singular value decomposition of the equipment layout change matrix, it is possible to grasp the dynamic changes of the construction equipment layout as a whole, providing an objective and accurate quantitative indicator for the construction progress evaluation.

[0074] Step S300: constructing a construction progress assessment model based on construction equipment layout based on the graph neural network, and extracting initial construction progress assessment results based on relevant features of the target construction equipment;

[0075] Step S300 specifically includes:

[0076] Step S310: Set the construction equipment as graph nodes, determine the edge connection relationship between nodes based on spatial distance and functional relevance, build the construction equipment layout graph structure and train it; specifically, use the construction equipment in the target area as nodes in the graph neural network, and the attributes of each node include the layout feature vector information extracted from step S200, that is, the relationship characteristics between the equipment location and key construction points, the operating status characteristics, and the number of related equipment related characteristics; similarly, use concrete pouring as the target area to determine the edge connection relationship:

[0077] For spatial distance: a spatial distance threshold is set according to the actual size of the concrete pouring sub-area and the equipment operating range. For example, the concrete pouring sub-area is a rectangular area with a length of 50 meters and a width of 30 meters. For equipment such as concrete mixers and conveying pumps that require material transmission, the spatial distance threshold can be set to 20 meters. If the actual spatial distance between two equipment is less than the threshold, it is considered that they are closely connected in space and it is possible to establish an edge connection.

[0078] Regarding functional relevance: the functional relevance between equipment is determined based on construction process knowledge. During the concrete pouring process, there is a clear material supply relationship between the concrete mixer truck and the concrete delivery pump, and this functional relevance is relatively strong; although there is no direct material transmission between the vibrator and the mixer truck, they both serve the overall process of concrete pouring, and there is a certain indirect functional relevance, but the degree of relevance is relatively weak; for equipment pairs with strong functional relevance, a higher functional relevance weight is assigned; for equipment pairs with indirect functional relevance, a lower functional relevance weight is assigned;

[0079] Step S320: Input the current construction equipment layout graph structure into the trained graph neural network model, and output the initial construction progress assessment result; specifically, input the concrete pouring sub-area construction equipment layout graph structure obtained by the drone at the current moment and processed by steps S100 and S200 into the trained graph neural network model. The model will perform information propagation and feature fusion through network structures such as graph convolution layers and fully connected layers according to the attributes of the graph nodes (equipment layout feature vectors) and the edge connection relationships between the nodes (including spatial distance and functional correlation information); finally, a numerical value representing the construction progress is output in the output layer, which can be expressed as p i , its value range is set between [0,1], where 0 represents 0% construction progress and 1 represents 100% construction progress. This value is the initial construction progress assessment result;

[0080] Step S330: extract adjustment factors according to the parameters affecting the construction progress due to the equipment layout change, and optimize the initial construction progress evaluation result to obtain the optimized construction progress evaluation result; specifically, the parameters affecting the construction progress due to the equipment layout change (i.e., the maximum singular value of the layout change matrix) are obtained from step S230, and the adjustment factors can be expressed as:

[0081]

[0082] In the formula, a represents the adjustment factor; I represents the maximum singular value of the layout change matrix; when the influencing parameter I is small, the adjustment factor is close to 1, indicating that the equipment layout changes slightly and the adjustment range of the initial construction progress assessment result is small; when the influencing parameter I is large, the adjustment factor will increase with the increase of I, and the growth rate will gradually slow down, which means that when the equipment layout changes significantly, the initial construction progress assessment result will be adjusted to a large extent, but the adjustment range will not be increased indefinitely to avoid over-correction;

[0083] In summary, step S300 fully utilizes the spatial distance and functional correlation information between construction equipment by constructing a graph neural network model based on the layout of construction equipment, and effectively integrates the multi-dimensional layout feature vector of construction equipment into the model, so that the model can deeply understand the collaborative relationship between equipment during the construction process and its impact on the construction progress from the perspective of the overall layout, avoiding the limitations of traditional evaluation methods that only consider a single device or simple linear combination of equipment information in isolation. By dynamically adjusting factors to optimize the initial construction progress evaluation results, it can adapt to changes in the layout of construction equipment in a timely manner, improve the accuracy and timeliness of construction progress evaluation, and provide more reliable progress evaluation information for construction project management, whether in a stable construction period or a period of frequent changes in equipment layout.

[0084] Step S400: Based on the initial construction progress assessment result and taking into account the uncertainty of the construction equipment layout, the parameters of the construction progress assessment model are adjusted using a zero-order feedback optimization algorithm.

[0085] Step S400 specifically includes:

[0086] Step S410: extracting key visual information based on the image video data, and calculating the gradient estimation values ​​at different times according to the objective function of the construction progress assessment model;

[0087] Step S410 includes:

[0088] Step S411: extracting the equipment position change amount and equipment operation status change characteristics and the regional material accumulation and consumption conditions from the image video data of the target area;

[0089] Step S412: According to the objective function of the construction progress assessment model, the finite difference method is used to calculate the estimated gradient value. The objective function of the construction progress assessment model is Jθ, where θ is a model parameter, and the finite difference method is used to calculate the estimated gradient value. =, for the model parameter θ(t), its estimated gradient value The calculation formula is:

[0090]

[0091] Among them, e j is the jth unit vector, ε represents the disturbance value; when calculating the objective function, the extracted key visual information is taken into consideration, and the objective function is assumed to be proportional to the device position change J p (θ,{Δp i}), Operation status changes J s (θ,{Δsa i}) and the parts related to material conditions are J m (θ, Δh, ΔA), then the objective function can be expressed as:

[0092] Jθ=w p J p (θ,{Δp i})+w s J s (θ,{Δsa i})+w m J m (θ, Δh, ΔA)

[0093] Among them, w p ,w s ,w m Both represent weight coefficients, indicating the influence weight of different features on the objective function, and w p +w s +wm = 1; by calculating the objective function in different parts affected by the parameter θ j The changes after the disturbance are used to obtain the estimated gradient according to the finite difference formula.

[0094] Step S420: extracting the gradient update step size at the next moment according to the estimated gradient values ​​of the objective function at different moments, combined with the layout and collaboration relationship changes caused by the movement of the device at the current moment and the entry of the new device;

[0095] Step S420 specifically includes:

[0096] Step S421: First, analyze the impact of equipment movement on the layout and collaborative relationship, calculate the moving speed of each construction equipment, and combine its importance weight to the construction progress in the target area to obtain the corresponding impact coefficient of each equipment movement; then, for the new equipment entering the site, determine the impact coefficient of the new equipment entering the site based on the potential collaborative correlation between the new equipment and the existing equipment and its own initial influence coefficient; finally, summarize the impact coefficients of all equipment movements and new equipment entering the site, and obtain the total layout and collaborative relationship change coefficient caused by the equipment movement and the new equipment entering the site at the current moment, so as to quantify the degree of dynamic change in the construction scene;

[0097] Equipment movement influence coefficient calculation For each construction equipment, the moving speed v is calculated using the equipment position change information obtained by image analysis. i , and its importance weight in the target area is (For example, the delivery pump has a higher weight due to its key role in concrete delivery, followed by the mixer truck). B i Indicates the equipment movement influence coefficient;

[0098] Calculation of the impact coefficient of new equipment entering the site: If new equipment enters the site, determine the potential collaboration correlation γ between the new equipment type k and the existing equipment based on the construction process knowledge k (determined according to the construction process, such as the degree of collaboration between the new vibrating equipment and the delivery pump and mixer truck), and the initial influence coefficient of the new equipment is η k (can be set according to equipment performance parameters), then the impact coefficient of the new equipment entering the site is δ k =γ k η k ; Summarize the impact coefficients of all equipment movement and new equipment entry, and get the change coefficient of the overall layout and collaborative relationship: ΔC = ∑ i B i +∑ k δ k , in order to quantify the degree of dynamic changes in the construction scene;

[0099] Step S422: Based on the estimated gradient values ​​of the objective function at the previous moment and the current moment, combined with the layout and collaboration relationship change coefficients, the gradient update step size at the next moment is calculated. The size of the step size will be adaptively adjusted according to the dynamic changes of the construction scene to ensure that the subsequent model parameter updates can keep up with the construction changes without being too large and unstable;

[0100] Let the current time be t, and the estimated gradient value of the previous time be The estimated gradient at the current moment is Gradient update step length a t The calculation formula is:

[0101]

[0102] In the formula, a t It represents the gradient update compensation at time t, that is, the amplitude of parameter adjustment in the construction progress evaluation model; It represents the estimated value of the gradient of the objective function of the construction progress evaluation model at time t; represents the gradient estimate of the objective function of the construction progress assessment model at time t-1; ΔC represents the change coefficient of the layout and collaborative relationship caused by equipment movement and the entry of new equipment; λ represents the smoothing coefficient, which is used to prevent the denominator from being zero or too small, resulting in a large step size; this formula dynamically adjusts the gradient update step size according to the change in the gradient estimate and the change coefficient of the layout and collaborative relationship. When the layout and collaborative relationship change greatly, the step size will be reduced accordingly to update the model parameters more robustly; otherwise, the step size will be appropriately increased to speed up the model convergence and ensure that the model parameter update can adapt to changes in the construction scene.

[0103] Step S430: iteratively updating the parameters of the construction progress assessment model according to the gradient update step size at the next moment;

[0104] Step S430 specifically includes:

[0105] Step S431: Based on the construction progress assessment model parameters at the current moment, according to the calculated gradient update step size at the next moment, the key parameters in the model are updated using the gradient descent method and other update rules, so that the model is adjusted in a direction that better fits the actual construction progress. Suppose the current model parameters are θ(t), and the calculated gradient update step size a is used. t , use the gradient descent method to update the model parameters:

[0106]

[0107] In the formula, θ(t+1) represents the parameters of the construction progress assessment model at time t+1; for the construction progress assessment model built based on graph neural network, the specific updated parameters include node feature weight parameters (such as adjusting the importance of equipment location features, operating status features, and related equipment quantity related features in assessing the construction progress), convolution kernel parameters of the graph convolution layer (optimizing the information transmission and aggregation method between nodes to adapt to the equipment layout and construction dynamic changes in the target area), and connection weights and bias parameters of the fully connected layer, adjusting the combination effect of different feature dimensions after passing through the fully connected layer, so that the model output is more in line with the actual construction progress;

[0108] Specifically, what is iteratively updated are the key parameters in the construction progress assessment model that will affect the final output of the construction progress assessment results. The following is a detailed explanation using the construction progress assessment model built based on the graph neural network as an example: when the construction equipment is abstracted as a graph node, each node contains information such as the relationship characteristics between the equipment location and the key construction point, the operating status characteristics, and the related characteristics of the number of associated equipment (refer to the node attribute assignment in step S310). In the process of information transmission and feature fusion, the graph neural network will assign different weight parameters to these node features. By iteratively updating these weight parameters through the gradient update step, the model can better make a reasonable assessment of the construction progress according to the changes in various aspects of the equipment at different times; on the other hand, the graph convolution layer in the graph neural network is the core structure for realizing information transmission and aggregation between nodes. Its convolution kernel parameters determine how to fuse the node's own features and the features of neighboring nodes to update the node's representation. Different convolution kernel parameters will cause the node to obtain different feature representations after each layer of graph convolution operation. The construction progress evaluation result that affects the final output is iteratively updated by the convolution kernel parameters according to the gradient update step size, so that the model can learn a feature fusion method that is more suitable for, for example, the equipment layout of the concrete pouring sub-area and the dynamic changes of the construction, so as to more accurately mine the information related to the construction progress from the equipment layout graph structure; on the other hand, in the subsequent fully connected layer of the graph neural network, the connection weights between the neurons and the bias parameters corresponding to each neuron will further nonlinearly transform and integrate the features output from the graph convolution layer, and finally generate an evaluation value of the construction progress (such as the initial construction progress evaluation result output in step S320). The connection weights and bias parameters of these fully connected layers are iteratively updated using the gradient update step size, which can adjust the combination effect of different feature dimensions after passing through the fully connected layer, so that the construction progress evaluation result output by the model is more in line with the actual construction progress of the concrete pouring sub-area, and can adapt to the dynamic changes in construction caused by factors such as changes in equipment layout and the entry of new equipment;

[0109] Step S432: After completing a parameter update, use the pre-divided validation set data to evaluate the performance of the updated model and calculate the mean square error evaluation index. If the evaluation index shows that the model performance has not improved, it indicates that the parameter update may be too large or in the wrong direction. In this case, the gradient update step size a should be appropriately reduced. t , and re-update the parameters; if the evaluation index continues to optimize, it means that the update is effective, then continue to iterate and update according to the above steps until the evaluation index converges to the set threshold (the mean square error is less than 0.01); through continuous iterative updates, the construction progress evaluation model can be continuously optimized according to factors such as changes in equipment layout, operating status, and material conditions in the concrete pouring sub-area, thereby improving the accuracy and adaptability of the construction progress evaluation.

[0110] Step S400 realizes the adaptability of the model to the uncertainty of construction equipment layout by extracting key visual information based on image video data and adjusting the parameters of the construction progress assessment model using a zero-order feedback optimization algorithm; in the process of calculating the gradient estimate, key visual information such as equipment position changes, operating status changes, and material conditions are integrated into the objective function, so that the model parameter adjustment is closely aligned with the actual construction scene changes, avoiding the problem of the model being out of touch with the actual construction in the traditional method; the adaptive calculation of the gradient update step can reasonably control the model parameter update amplitude according to the dynamic changes of the construction scene, ensuring that the model can respond to changes in the construction progress in a timely manner and preventing the model from being unstable due to drastic changes in the construction scene; by iteratively updating the model parameters, especially optimizing the key parameters in the graph neural network, the model can continuously learn and adapt to, for example, the dynamic changes in the equipment layout in the concrete pouring sub-area, the changes in the operating status, and the use of materials, thereby continuously improving the accuracy and timeliness of the construction progress assessment.

[0111] In summary, the construction progress assessment method based on UAV visual collaboration uses the target detection algorithm to identify construction equipment and extract relevant features, accurately captures the dynamic changes of individual equipment, and lays the foundation for subsequent evaluation; the evaluation model constructed based on graph neural network, by setting construction equipment as graph nodes and constructing edge connection relationships, effectively reflects the overall logic and mutual influence of construction equipment layout, and then obtains the initial evaluation results; in particular, the uncertainty of construction equipment layout is considered, and the model parameters are adjusted using the zero-order feedback optimization algorithm. The algorithm can extract key visual information based on image video data to calculate gradient estimates, and keenly perceive the impact trend of equipment layout changes on evaluation results. By iteratively updating model parameters, the model can adapt to this uncertainty, continuously optimize the evaluation results, and significantly improve the accuracy of evaluation when facing equipment layout changes, providing a reliable decision-making basis for construction project management. At the same time, this method does not require the construction of an accurate model. It relies on the information at different times fed back by the UAV visual data to calculate the gradient estimate, avoids complex explicit gradient calculations, reduces the amount of calculation, and is easy to deploy and implement at various construction sites. Even if there is uncertainty in the construction process and it is difficult to accurately build a model, it is still possible to effectively evaluate the construction progress, which effectively ensures that the construction progress evaluation work is carried out efficiently and accurately in complex and changeable construction scenarios.

[0112] See also Figure 2 As shown, the technical solution of the second aspect of the present invention provides a construction progress assessment system based on UAV visual collaboration, using the construction progress assessment method based on UAV visual collaboration described in the technical solution of the first aspect of the present invention, and the system includes:

[0113] A data acquisition and preprocessing module, configured to acquire image and video data of the construction site based on a predetermined flight path and time interval and perform geometric correction preprocessing;

[0114] A target detection module configured to identify the construction equipment in the preprocessed image video data using a target detection algorithm and extract relevant features of the target construction equipment;

[0115] a construction progress assessment module, configured to construct a construction progress assessment model based on construction equipment layout based on a graph neural network, and extract an initial construction progress assessment result based on relevant features of the target construction equipment;

[0116] The construction progress assessment optimization module is configured to adjust the parameters of the construction progress assessment model using a zero-order feedback optimization algorithm based on the initial construction progress assessment results and taking into account the uncertainty of the construction equipment layout.

[0117] The technical solution of the third aspect of the present invention provides an electronic device, comprising: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can execute the steps of the construction progress assessment method based on drone visual collaboration described in the technical solution of the first aspect of the present invention.

[0118] The technical solution of the fourth aspect of the present invention provides a computer-readable storage medium, on which is stored a program for implementing a construction progress assessment method based on drone visual collaboration. The program for implementing a construction progress assessment method based on drone visual collaboration is executed by a processor to implement the steps of the construction progress assessment method based on drone visual collaboration described in the technical solution of the first aspect of the present invention.

[0119] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0120] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. The construction progress assessment method based on UAV visual collaboration is characterized by: The method comprises: Capture image and video data of the construction site based on a predetermined flight path and time interval and perform geometric correction preprocessing; Use the target detection algorithm to identify the construction equipment in the preprocessed image video data and extract the relevant features of the target construction equipment; A construction progress assessment model based on construction equipment layout is built based on graph neural network, and the initial construction progress assessment results are extracted according to the relevant features of the target construction equipment; Based on the initial construction progress assessment results and considering the uncertainty of construction equipment layout, the parameters of the construction progress assessment model are adjusted using a zero-order feedback optimization algorithm.

2. The construction progress assessment method based on UAV visual collaboration as claimed in claim 1 is characterized in that: The image and video data acquisition and geometric correction preprocessing of the construction site based on the predetermined flight path and time interval include: Divide the construction site into areas according to the characteristics of the construction process and the functional differences of the construction areas; Dynamically adjust the collection time interval within the target area based on the movement speed of construction equipment and the intensity of construction activities; The image and video data are corrected according to the preset ground control points in the target area.

3. The construction progress assessment method based on UAV visual collaboration as claimed in claim 1, characterized in that: The target detection algorithm is used to identify the construction equipment in the preprocessed image video data, and the relevant features of the target construction equipment are extracted, including: Obtain the rectified image and video data of each region and train the target detection model; Extract the layout feature vector of construction equipment based on the trained object detection model; The differences between the characteristic vectors of construction equipment layout at different times are calculated, and the parameters affecting the construction progress due to equipment layout changes are extracted.

4. The construction progress assessment method based on UAV visual collaboration as claimed in claim 3 is characterized in that: The layout feature vectors of construction equipment extracted based on the trained target detection model include: Extract location features based on the distance between construction equipment and key construction points; Extracting operation status features according to the operation status changes of construction equipment at different time periods; The correlation features are extracted according to the difference in the degree of correlation between construction equipment at different times.

5. The construction progress assessment method based on UAV visual collaboration as claimed in claim 1, characterized in that: A construction progress assessment model based on construction equipment layout is constructed based on graph neural network, and the initial construction progress assessment results are extracted according to the relevant characteristics of the target construction equipment, including: The construction equipment is set as graph nodes, the edge connection relationship between nodes is determined based on the spatial distance and functional relevance, and the construction equipment layout graph structure is constructed and trained; Input the current construction equipment layout diagram structure into the trained graph neural network model and output the initial construction progress assessment result; According to the parameters affecting the construction progress caused by the changes in equipment layout, adjustment factors are extracted to optimize the initial construction progress assessment results.

6. The construction progress assessment method based on UAV visual collaboration according to any one of claims 1 to 5, characterized in that: Based on the initial construction progress assessment results and considering the uncertainty of construction equipment layout, the parameters of the construction progress assessment model are adjusted using the zero-order feedback optimization algorithm, including: Extract key visual information based on image and video data, and calculate the estimated gradient values ​​at different times according to the objective function of the construction progress assessment model; According to the estimated gradient values ​​of the objective function at different times, combined with the layout and collaboration relationship changes caused by the current device movement and the entry of new devices, the gradient update step size at the next moment is extracted; The parameters of the construction progress evaluation model are iteratively updated according to the gradient update step size at the next moment.

7. The construction progress assessment method based on UAV visual collaboration as claimed in claim 6 is characterized in that: The expression of gradient update step size is: In the formula, a t It represents the gradient update compensation at time t, that is, the amplitude of parameter adjustment in the construction progress evaluation model; It represents the estimated value of the gradient of the objective function of the construction progress evaluation model at time t; It represents the estimated value of the gradient of the objective function of the construction progress evaluation model at time t-1; ΔC represents the change coefficient of the layout and collaboration relationship caused by the movement of equipment and the arrival of new equipment; λ represents the smoothing coefficient.

8. The construction progress assessment system based on UAV visual collaboration is characterized by: The construction progress assessment method based on UAV visual collaboration according to any one of claims 1 to 7 is adopted, and the system comprises: A data acquisition and preprocessing module, configured to acquire image and video data of the construction site based on a predetermined flight path and time interval and perform geometric correction preprocessing; A target detection module configured to identify the construction equipment in the preprocessed image video data using a target detection algorithm and extract relevant features of the target construction equipment; a construction progress assessment module, configured to construct a construction progress assessment model based on construction equipment layout based on a graph neural network, and extract an initial construction progress assessment result based on relevant features of the target construction equipment; The construction progress assessment optimization module is configured to adjust the parameters of the construction progress assessment model using a zero-order feedback optimization algorithm based on the initial construction progress assessment results and taking into account the uncertainty of the construction equipment layout.

9. An electronic device, characterized in that: The electronic device includes: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can execute the steps of the construction progress assessment method based on drone visual collaboration as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program for implementing a construction progress assessment method based on drone visual collaboration, and the program for implementing a construction progress assessment method based on drone visual collaboration is executed by a processor to implement the steps of the construction progress assessment method based on drone visual collaboration as described in any one of claims 1 to 7.

Images