UAV monitoring construction quality early warning system and method based on big data

Abstract

The present invention relates to the technical field of construction quality early warning, and discloses a construction quality early warning system and method for unmanned aerial vehicle monitoring based on big data. First, the feature points of the initial construction site image are extracted to establish a three-dimensional point cloud model of the construction site; secondly, the principal component comprehensive index of the marked points in the three-dimensional model of the construction site is extracted based on the principal component analysis method, and the marked points are screened according to the contribution rate of the principal component comprehensive index to obtain the construction quality early warning position; the construction quality early warning index is assigned at the construction quality early warning position, and the construction quality early warning index is processed using the gray correlation method to obtain the construction quality early warning index weight; finally, based on the BP neural network model, the construction quality early warning index weight is input into the BP neural network model, and the warning level is output to realize the construction quality early warning. The present invention realizes the construction quality early warning by extracting the construction quality early warning index, and the result is accurate and objective.

Description

Technical Field

[0001] The present invention relates to the technical field of construction quality early warning, and specifically to a construction quality early warning system and method for unmanned aerial vehicle monitoring based on big data. Background Art

[0002] Chinese patent CN118298610A discloses an intelligent early warning system for safety at construction sites. The system specifically includes an information acquisition module, which can acquire equipment operation information, road construction information and geological information within the monitoring range; a road segmentation module, which analyzes the segment length according to the road construction information within the monitoring range, and divides the construction roads within the monitoring range according to the results of the segment length; a direction state analysis module, which analyzes the state of the segmented road according to the construction road division results and the road construction information, and analyzes the direction state of the construction road according to the results of the segmented road state analysis; a quality state analysis module, which analyzes the quality state of the segmented road according to the construction road division results and the road construction information; an edge state analysis module, which analyzes the edge state of the construction road according to the geological information and the road construction information; and a safety early warning module, which provides users with safety early warnings according to the results of the construction road direction state analysis, the results of the segmented road quality state analysis and the results of the construction road edge state analysis. This invention only monitors road construction and has a small scope of application.

[0003] Traditional construction quality early warning methods have disadvantages such as slow speed and low accuracy. After using drones to monitor construction quality, traditional monitoring methods are easily affected by unstable factors such as wind speed, construction obstructions and inability to work continuously. At the same time, high-tech technologies such as neural networks and machine vision are not used, the monitoring range is small and the early warning accuracy is low. Summary of the invention

[0004] In response to the problems in the related technology, the present invention provides a construction quality early warning system and method for drone monitoring based on big data to overcome the above-mentioned technical problems existing in the existing related technology.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a construction quality early warning method for drone monitoring based on big data, comprising the following steps:

[0007] S1, obtaining an initial construction site image, extracting feature points of the initial construction site image, obtaining feature points of the initial construction site image, then extracting an ultra-high density point cloud of the initial construction site image, constructing a three-dimensional point cloud model, filtering the three-dimensional point cloud model to obtain a three-dimensional model of the construction site, and then performing error evaluation on the three-dimensional model of the construction site;

[0008] S2, selecting marking points in the three-dimensional model of the construction site, extracting the principal component comprehensive index of the marking points based on the principal component analysis method, and deleting the marking points that do not meet the requirements according to the contribution rate of the principal component comprehensive index to obtain the construction quality warning position;

[0009] S3, assigning values ​​to the construction quality warning positions to obtain construction quality warning indicators, obtaining a comprehensive weight matrix and a gray correlation coefficient matrix based on a gray correlation method, normalizing the comprehensive weight matrix and the gray correlation coefficient matrix to obtain construction quality warning indicator weights;

[0010] S4. Train the BP neural network to obtain a BP neural network model, input the construction quality early warning indicator weights into the BP neural network model, output the warning level, and complete the construction quality early warning.

[0011] The invention obtains an initial construction site image, and uses the SIFT feature point extraction algorithm to construct a source point cloud set and a target point cloud set, extracts the feature points of the initial construction site image, and then uses a multi-view dense matching algorithm to form an ultra-high density point cloud. The ultra-high density point cloud is filtered using a cloth model to remove non-ground point clouds to obtain a three-dimensional model of the construction site; the three-dimensional model of the construction site is constructed by the point cloud model, and obstructions are effectively removed, and the error of the constructed model is small; secondly, marking points are selected in the three-dimensional model of the construction site, and the principal component comprehensive index of the marking points is extracted, and then the contribution of the principal component comprehensive index is calculated to obtain the construction quality warning position; through the contribution positioning and information related to the construction quality, interference and errors can be reduced, which is convenient for subsequent data processing; then the construction quality warning position is assigned a value to obtain a construction quality warning index, and the correlation of the construction quality warning index is obtained based on the grey correlation analysis, and the weight of the construction quality warning index is obtained after assigning weights; the method effectively establishes the association between the indicators and is easy to understand; finally, the BP neural network is trained using the construction quality warning index weight to obtain a BP neural network model, and the warning level is output to achieve construction quality warning, and the neural network converges quickly and has a high recognition accuracy.

[0012] Preferably, the S1 comprises the following steps:

[0013] S11, selecting phased points at the construction site, and using a drone to shoot construction site images at the construction site according to designated route information, wherein the designated route information includes the drone's flight altitude, the drone's camera angle, and the shooting overlap, to obtain initial construction site images, forming an initial construction site image set; establishing a three-dimensional coordinate system for the initial construction site images in the initial construction site image set, setting a source point cloud set and a target point cloud set, and the variable-scale Gaussian function of the source point cloud set is: , the variable scale Gaussian function of the target point cloud is , the curvature function of the source point cloud is , the curvature function of the target point cloud is , then the source point cloud scale space and the target point cloud scale space The calculation formula is as follows:

[0014] ; ;

[0015] Constructing Gaussian difference functions that satisfy the source point cloud scale space and the target point cloud scale space, respectively obtaining a first Gaussian difference function and a second Gaussian difference function; selecting sampling points on the initial construction site image in the initial construction site image set, calculating the first Gaussian difference function value and the second Gaussian difference function value of the sampling point in the scale space, if the first Gaussian difference function value and the second Gaussian difference function value in the current scale space and the adjacent scale space are both the maximum or the minimum, then the sampling point is recorded as the extreme point in the scale space;

[0016] Performing Taylor fitting on the first Gaussian difference function and the second Gaussian difference function in the scale space to obtain a Taylor fitting formula for the extreme point in the scale space, making the first-order derivative of the Taylor fitting formula for the extreme point in the scale space 0, obtaining an extreme point offset, adding the extreme point offset and the coordinates of the extreme point in the scale space to obtain a scale space feature point; finding the scale space feature point corresponding to the initial construction site image in the initial construction site image set, and then extracting other feature points based on the pyramid structure to form a feature point set of the initial construction site image;

[0017] S12, set the coordinates of the feature points in the initial construction site image feature point set to ( a,b,c ), the coordinates of the photography center are , the direction cosine of the feature point is , , , , , , , and , the collinear equation is established as follows:

[0018] ,

[0019] ;

[0020] in, A and B represents the collinear equation, d Indicates focal length;

[0021] According to the collinear equation, an ultra-high density point cloud of the initial construction site image in the initial construction site image set is obtained, and the specific steps are as follows:

[0022] S121, using the multi-view clustering method to segment the initial construction site image, taking the weights corresponding to the feature points in the feature point set of the initial construction site image as the image edge value, and setting the edge threshold to , will be greater than The image edge value is taken as a cluster, the efficiency of the image edge value that is not in the cluster is calculated, and then the image edge value corresponding to the maximum efficiency is added to the cluster; S211 is repeated until all feature points of the initial construction site image are clustered to obtain the initial construction site image cluster;

[0023] S122, performing initial feature matching according to the initial construction site image cluster to obtain a feature point matching set, performing expansion processing on the feature point matching set to obtain a three-dimensional point cloud set, and filtering the three-dimensional point cloud set to obtain an ultra-high density point cloud;

[0024] S13, constructing a three-dimensional point cloud model using an ultra-high density point cloud, assuming that there is a cloth model in the three-dimensional point cloud model and that the cloth model and the three-dimensional point cloud model fit the ground; using the cloth model to process non-ground point clouds in the three-dimensional point cloud model to obtain a three-dimensional point cloud model of the construction site, the specific steps are as follows:

[0025] S131, selecting action point clouds on the cloth model to form an action point cloud set, flipping the action point clouds in the action point cloud set, allowing the cloth model to cover the action point cloud set under the action of gravity and elastic force, and performing a first filtering on non-ground point clouds in the three-dimensional point cloud model according to the shape of the action point clouds covering the action point cloud set;

[0026] S132: The cloth model is vertically offset, and the action point cloud in the action point cloud set is moved below the ground point cloud, and displacement correction is performed under the action of gravity and elastic force. The distance threshold is set to , calculate the distance between the action point cloud and the ground point cloud in the action point cloud set, recorded as the judgment distance, when the judgment distance is greater than , the action point cloud in the action point cloud set corresponding to the determination distance is recorded as a non-ground point cloud, otherwise the action point cloud in the action point cloud set corresponding to the determination distance is recorded as a ground point cloud; the non-ground point cloud is deleted to complete the second filtering and obtain the three-dimensional point cloud model of the construction site;

[0027] S14, converting the three-dimensional point cloud model of the construction site into a three-dimensional model of the construction site, and setting detection points, and the root mean square errors of the detection points in three dimensions are , and , the first The three-dimensional coordinates of the detection points are , and , GPS coordinate locator measured the first The three-dimensional coordinates of the detection points are , and , then the root mean square error calculation formula of the detection point in three-dimensional direction is as follows:

[0028] , , ;

[0029] Set the error threshold to , when the root mean square error of the detection point in the three-dimensional direction is less than , the three-dimensional model of the construction site meets the requirements, otherwise the three-dimensional model of the construction site is reconstructed until the root mean square error of the three-dimensional direction of the detection point is less than .

[0030] The invention obtains the initial construction site image, uses the SIFT feature point extraction algorithm to extract the feature points of the initial construction site image, and then uses the multi-view dense matching algorithm to form an ultra-high density point cloud. The ultra-high density point cloud is filtered using a cloth model to remove non-ground point clouds, effectively clear obstructions, and the constructed model has a small error, thereby obtaining a three-dimensional model of the construction site.

[0031] Preferably, S2 comprises the following steps:

[0032] S21. The drone flies in the three-dimensional model of the construction site and locates the marking points, which are the locations of the construction site selected according to the expert survey method to measure the construction quality, and are recorded as the marking point set. ,in Indicates i Marked points, set the feature dimension of the marked points in the marked point set to j， Generate marker matrix as follows:

[0033] ;

[0034] in, Indicates i The feature dimension is j The marking point;

[0035] Set the normalized random vector of the marker point set to ,in express The corresponding standardized random vector value, the calculation formula of the principal component comprehensive index is as follows:

[0036] ;

[0037] in, represents the comprehensive index of principal components;

[0038] The specific steps for obtaining the principal component comprehensive index are as follows:

[0039] S211, calculating the principal component comprehensive index Variance ,in represents the relationship coefficient matrix, express The transposition of the principal component comprehensive index The variance of the relationship coefficient matrix is ​​obtained according to the Lagrange multiplier method to form a characteristic root set. ,in Indicates characteristic roots; select the first characteristic root in the characteristic root set; Characteristic root , the eigenvector corresponding to the largest eigenroot in the eigenroot set is , when the eigenvector satisfy When the principal component comprehensive index ,in Represents the feature vector otherwise, reselect the characteristic roots in the characteristic root set until the characteristic root is satisfied. ;

[0040] S212, make the principal component comprehensive index And the principal component comprehensive index The covariance is 0, repeat S211 to obtain the principal component comprehensive index , and obtain the comprehensive indexes of all principal components in turn;

[0041] S22, set The characteristic root corresponding to the principal component comprehensive index is , then The calculation formula for the contribution rate of the principal component comprehensive index is as follows:

[0042] ;

[0043] in, represents the contribution rate of the main component comprehensive index, Indicates The characteristic roots corresponding to the principal component comprehensive indicators;

[0044] Get the principal component comprehensive index contribution rate set ,in Indicates iThe contribution rate of the principal component comprehensive index is set Indicates a comprehensive evaluation. Represents the principal component comprehensive index set composed of the principal component comprehensive index, Represents the transposition of the principal component comprehensive index contribution rate set, and the comprehensive evaluation calculation formula is as follows:

[0045] ;

[0046] Set the evaluation threshold to , retain the comprehensive evaluation greater than Corresponding principal component comprehensive index, delete the comprehensive evaluation less than or equal to The corresponding principal component comprehensive index is obtained to obtain the screened principal component comprehensive index, and the marked points corresponding to the screened principal component comprehensive index are selected and recorded as the construction quality warning positions.

[0047] The invention selects marking points in the three-dimensional model of the construction site, calculates the characteristic roots by constructing a relationship coefficient matrix, uses the principal component analysis method to extract the principal component comprehensive index of the marking points, and then calculates the contribution of the principal component comprehensive index. The principal component comprehensive index is screened by a threshold to obtain the construction quality warning position, which can extract information related to the construction quality and overcome the influence of other factors.

[0048] Preferably, S3 comprises the following steps:

[0049] S31, assigning values ​​to the construction quality warning positions according to the construction quality risk degree based on the expert experience method, and recording the assigned values ​​as construction quality warning indicators to obtain a construction quality warning indicator matrix as follows:

[0050] ;

[0051] in Indicates m The first construction quality warning location n Secondary assignment;

[0052] Select the maximum assigned value and make a difference with the construction quality warning indicator in the construction quality warning indicator matrix to obtain a difference matrix; calculate the maximum range of the difference in the difference matrix and minimum range , select The first construction quality warning location The next assignment is , calculate the quotient of the maximum range, recorded as the resolution error ; Set the resolution factor to ,when hour, ,when hour, ; According to the resolution coefficient, the subjective weights of the construction quality warning indicators in the construction quality warning indicator matrix are assigned to obtain a subjective weight matrix;

[0053] S32, calculating the average value and standard deviation of the construction quality warning indicators in the construction quality warning indicator matrix, and obtaining the standard deviation coefficient of the construction quality warning indicators in the construction quality warning indicator matrix according to the average value and standard deviation, and the calculation formula is as follows:

[0054] ;

[0055] in, represents the standard deviation coefficient, represents the average value of the construction quality early warning indicators in the construction quality early warning indicator matrix, Represents the standard deviation of the construction quality early warning indicator in the construction quality early warning indicator matrix;

[0056] The standard deviation coefficient is normalized, and the normalization result is the objective weight of the construction quality early warning indicator in the construction quality early warning indicator matrix, so as to obtain an objective weight matrix;

[0057] S33, combining the subjective weight matrix and the objective weight matrix and assigning weights to obtain a comprehensive weight matrix, and then calculating the grey correlation coefficient of the construction quality early warning index matrix, the calculation formula is as follows:

[0058] ;

[0059] in, Indicates The first construction quality warning location Grey relational coefficient of secondary assignment;

[0060] A grey correlation coefficient matrix is ​​obtained, and the grey correlation index of the construction quality early warning index is obtained by combining the comprehensive weight matrix. The grey correlation index is normalized, and the normalization result is the construction quality early warning index weight.

[0061] The invention obtains construction quality warning indicators by assigning values ​​to construction quality warning positions, obtains a comprehensive weight matrix based on grey correlation analysis, and then obtains construction quality warning indicator weights based on the grey correlation coefficient and the comprehensive weight matrix. This method effectively establishes associations between indicators and facilitates subsequent neural network processing.

[0062] Preferably, S4 comprises the following steps:

[0063] S41, the UAV flies in the three-dimensional model of the construction site, relocates the new marking points, generates new construction quality warning indicator weights, and divides the new construction quality warning indicator weights into a weight training set and a weight test set; sets the number of input layer nodes of the BP neural network to the number of new construction quality warning indicators, the output layer nodes to 1, and the learning rate to , introduce the momentum factor to optimize the weight, set the current number of iterations to g, the maximum number of iterations to G, input the weight training set into the BP neural network, and when the current number of iterations reaches the maximum number of iterations, the trained BP neural network is obtained; then input the weight test set into the trained BP neural network, and set the accuracy threshold to , when the output result accuracy is less than When , the BP neural network model is obtained, otherwise the weights are adjusted until the output result accuracy is less than ;

[0064] S42. Input the weights of the construction quality early warning indicators into the BP neural network model. The BP neural network model outputs a result. When the output is 1, it indicates that the warning level is safe, and the construction quality early warning is a very small risk; when the output is 2, it indicates that the warning level is low, and the construction quality early warning is a small risk; when the output is 3, it indicates that the warning level is medium, and the construction quality early warning is a medium risk; when the output is 4, it indicates that the warning level is high, and the construction quality early warning is a high risk; when the output is 5, it indicates that the warning level is special, and the construction quality early warning is a very high risk; the construction quality early warning is completed.

[0065] The invention trains and tests the BP neural network using the weights of construction quality early warning indicators. After meeting the accuracy requirements, a BP neural network model is obtained. The output numbers of the BP neural network model represent the warning level, and the risk level of the construction quality early warning is obtained to achieve construction quality early warning. The BP neural network has a wide range of applications, fast recognition speed and high accuracy.

[0066] The present invention also discloses a system for a construction quality early warning method of drone monitoring based on big data, which specifically includes: a construction site three-dimensional model building module, a construction quality early warning position extraction module, a construction quality early warning index weight calculation module and a neural network construction quality early warning module;

[0067] The construction site three-dimensional model building module is used to build a construction site three-dimensional model according to the construction site image;

[0068] The construction quality warning position extraction module is used to extract construction quality warning positions that meet the requirements based on the principal component analysis method;

[0069] The construction quality early warning indicator weight calculation module is used for the construction quality early warning indicator, and the construction quality early warning indicator weight is calculated based on the grey correlation method;

[0070] The neural network construction quality early warning module is used to train the BP neural network, obtain the BP neural network model, output the warning level, and realize the construction quality early warning.

[0071] The present invention has the following beneficial effects:

[0072] 1. The invention uses the SIFT feature point extraction algorithm to extract the feature points of the initial construction site image, and then uses the multi-view dense matching algorithm to form an ultra-high density point cloud. The cloth model is used to filter the ultra-high density point cloud, which can remove non-ground point clouds and effectively remove obstructions, making the model error smaller.

[0073] 2. The invention extracts the principal component comprehensive index of the marking point based on the principal component analysis method, calculates the contribution of the principal component comprehensive index, and screens the principal component comprehensive index through a threshold to obtain the construction quality warning position, which can effectively extract information related to the construction quality and overcome the influence of other factors.

[0074] 3. The invention obtains the weights of construction quality warning indicators based on the grey correlation analysis method through the construction quality warning indicators. This method effectively establishes the correlation between indicators, extracts the weights of indicators with more relevant relationships, reduces the amount of calculation, and facilitates subsequent processing.

[0075] 4. The invention obtains a BP neural network model by training the BP neural network, and outputs a number representing the warning level to achieve construction quality early warning. The BP neural network has fast recognition speed and high accuracy.

[0076] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0078] Figure 1 A schematic diagram of the process of early warning of construction quality monitored by drones by the early warning system for construction quality monitored by drones based on big data provided by the present invention. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0081] Example 1

[0082] This implementation discloses a construction quality early warning method based on big data using drone monitoring, which specifically includes the following contents:

[0083] S1, obtaining an initial construction site image, extracting feature points of the initial construction site image, obtaining feature points of the initial construction site image, then extracting an ultra-high density point cloud of the initial construction site image, constructing a three-dimensional point cloud model, filtering the three-dimensional point cloud model to obtain a three-dimensional model of the construction site, and then performing error evaluation on the three-dimensional model of the construction site;

[0084] The S1 comprises the following steps:

[0085] S11, selecting phased points at the construction site, and using a drone to shoot construction site images at the construction site according to designated route information, wherein the designated route information includes the drone's flight altitude, the drone's camera angle, and the shooting overlap, to obtain initial construction site images, forming an initial construction site image set; establishing a three-dimensional coordinate system for the initial construction site images in the initial construction site image set, setting a source point cloud set and a target point cloud set, and the variable-scale Gaussian function of the source point cloud set is: , the variable scale Gaussian function of the target point cloud is , the curvature function of the source point cloud is , the curvature function of the target point cloud is , then the source point cloud scale space and the target point cloud scale space The calculation formula is as follows:

[0086] ; ;

[0087] Constructing Gaussian difference functions that satisfy the source point cloud scale space and the target point cloud scale space, respectively obtaining a first Gaussian difference function and a second Gaussian difference function; selecting sampling points on the initial construction site image in the initial construction site image set, calculating the first Gaussian difference function value and the second Gaussian difference function value of the sampling point in the scale space, if the first Gaussian difference function value and the second Gaussian difference function value in the current scale space and the adjacent scale space are both the maximum or the minimum, then the sampling point is recorded as the extreme point in the scale space;

[0088] Performing Taylor fitting on the first Gaussian difference function and the second Gaussian difference function in the scale space to obtain a Taylor fitting formula for the extreme point in the scale space, making the first-order derivative of the Taylor fitting formula for the extreme point in the scale space 0, obtaining an extreme point offset, adding the extreme point offset and the coordinates of the extreme point in the scale space to obtain a scale space feature point; finding the scale space feature point corresponding to the initial construction site image in the initial construction site image set, and then extracting other feature points based on the pyramid structure to form a feature point set of the initial construction site image;

[0089] S12, set the coordinates of the feature points in the initial construction site image feature point set to ( a,b,c ), the coordinates of the photography center are , the direction cosine of the feature point is , , , , , , , and , the collinear equation is established as follows:

[0090] ,

[0091] ;

[0092] in, A and B represents the collinear equation, d Indicates focal length;

[0093] According to the collinear equation, an ultra-high density point cloud of the initial construction site image in the initial construction site image set is obtained, and the specific steps are as follows:

[0094] S121, using the multi-view clustering method to segment the initial construction site image, taking the weights corresponding to the feature points in the feature point set of the initial construction site image as the image edge value, and setting the edge threshold to , will be greater than The image edge value is taken as a cluster, the efficiency of the image edge value that is not in the cluster is calculated, and then the image edge value corresponding to the maximum efficiency is added to the cluster; S211 is repeated until all feature points of the initial construction site image are clustered to obtain the initial construction site image cluster;

[0095] S122, performing initial feature matching according to the initial construction site image cluster to obtain a feature point matching set, performing expansion processing on the feature point matching set to obtain a three-dimensional point cloud set, and filtering the three-dimensional point cloud set to obtain an ultra-high density point cloud;

[0096] S13, constructing a three-dimensional point cloud model using an ultra-high density point cloud, assuming that there is a cloth model in the three-dimensional point cloud model and that the cloth model and the three-dimensional point cloud model fit the ground; using the cloth model to process non-ground point clouds in the three-dimensional point cloud model to obtain a three-dimensional point cloud model of the construction site, the specific steps are as follows:

[0097] S131, selecting action point clouds on the cloth model to form an action point cloud set, flipping the action point clouds in the action point cloud set, allowing the cloth model to cover the action point cloud set under the action of gravity and elastic force, and performing a first filtering on non-ground point clouds in the three-dimensional point cloud model according to the shape of the action point clouds covering the action point cloud set;

[0098] S132: The cloth model is vertically offset, and the action point cloud in the action point cloud set is moved below the ground point cloud, and displacement correction is performed under the action of gravity and elastic force. The distance threshold is set to , calculate the distance between the action point cloud and the ground point cloud in the action point cloud set, recorded as the judgment distance, when the judgment distance is greater than , the action point cloud in the action point cloud set corresponding to the determination distance is recorded as a non-ground point cloud, otherwise the action point cloud in the action point cloud set corresponding to the determination distance is recorded as a ground point cloud; the non-ground point cloud is deleted to complete the second filtering and obtain the three-dimensional point cloud model of the construction site;

[0099] S14, converting the three-dimensional point cloud model of the construction site into a three-dimensional model of the construction site, and setting detection points, and the root mean square errors of the detection points in three dimensions are , and , the first The three-dimensional coordinates of the detection points are , and , GPS coordinate locator measured the first The three-dimensional coordinates of the detection points are , and , then the root mean square error calculation formula of the detection point in three-dimensional direction is as follows:

[0100] , , ;

[0101] Set the error threshold to , when the root mean square error of the detection point in the three-dimensional direction is less than , the three-dimensional model of the construction site meets the requirements, otherwise the three-dimensional model of the construction site is reconstructed until the root mean square error of the three-dimensional direction of the detection point is less than ;

[0102] S2, selecting marking points in the three-dimensional model of the construction site, extracting the principal component comprehensive index of the marking points based on the principal component analysis method, and deleting the marking points that do not meet the requirements according to the contribution rate of the principal component comprehensive index to obtain the construction quality warning position;

[0103] The S2 comprises the following steps:

[0104] S21. The drone flies in the three-dimensional model of the construction site and locates the marking points, which are the locations of the construction site selected according to the expert survey method to measure the construction quality, and are recorded as the marking point set. ,in Indicates i Marked points, set the feature dimension of the marked points in the marked point set to j， Generate marker matrix as follows:

[0105] ;

[0106] in, Indicates i The feature dimension is j The marking point;

[0107] Set the normalized random vector of the marker point set to ,in express The corresponding standardized random vector value, the calculation formula of the principal component comprehensive index is as follows:

[0108] ;

[0109] in, represents the comprehensive index of principal components;

[0110] The specific steps for obtaining the principal component comprehensive index are as follows:

[0111] S211, calculating the principal component comprehensive index Variance ,in represents the relationship coefficient matrix, express The transposition of the principal component comprehensive index The variance of the relationship coefficient matrix is ​​obtained according to the Lagrange multiplier method to form a characteristic root set. ,in Indicates characteristic roots; select the first characteristic root in the characteristic root set; Characteristic root , the eigenvector corresponding to the largest eigenroot in the eigenroot set is , when the eigenvector satisfy When the principal component comprehensive index ,in Represents the feature vector otherwise, reselect the characteristic roots in the characteristic root set until the characteristic root is satisfied. ;

[0112] S212, make the principal component comprehensive index And the principal component comprehensive index The covariance is 0, repeat S211 to obtain the principal component comprehensive index , and obtain the comprehensive indexes of all principal components in turn;

[0113] S22, set The characteristic root corresponding to the principal component comprehensive index is , then The calculation formula for the contribution rate of the principal component comprehensive index is as follows:

[0114] ;

[0115] in, represents the contribution rate of the main component comprehensive index, Indicates The characteristic roots corresponding to the principal component comprehensive indicators;

[0116] Get the principal component comprehensive index contribution rate set ,in Indicates i The contribution rate of the principal component comprehensive index is set Indicates a comprehensive evaluation. Represents the principal component comprehensive index set composed of the principal component comprehensive index, Represents the transposition of the principal component comprehensive index contribution rate set, and the comprehensive evaluation calculation formula is as follows:

[0117] ;

[0118] Set the evaluation threshold to , retain the comprehensive evaluation greater than Corresponding principal component comprehensive index, delete the comprehensive evaluation less than or equal to The corresponding principal component comprehensive index is obtained to obtain the screened principal component comprehensive index, and the marked point corresponding to the screened principal component comprehensive index is selected and recorded as the construction quality warning position;

[0119] S3, assigning values ​​to the construction quality warning positions to obtain construction quality warning indicators, obtaining a comprehensive weight matrix and a gray correlation coefficient matrix based on a gray correlation method, normalizing the comprehensive weight matrix and the gray correlation coefficient matrix to obtain construction quality warning indicator weights;

[0120] The S3 comprises the following steps:

[0121] S31, assigning values ​​to the construction quality warning positions according to the construction quality risk degree based on the expert experience method, and recording the assigned values ​​as construction quality warning indicators to obtain a construction quality warning indicator matrix as follows:

[0122] ;

[0123] in Indicates m The first construction quality warning location n Secondary assignment;

[0124] Select the maximum assigned value and make a difference with the construction quality warning indicator in the construction quality warning indicator matrix to obtain a difference matrix; calculate the maximum range of the difference in the difference matrix and minimum range , select The first construction quality warning location The next assignment is , calculate the quotient of the maximum range, recorded as the resolution error ; Set the resolution factor to ,when hour, ,when hour, ; According to the resolution coefficient, the subjective weights of the construction quality warning indicators in the construction quality warning indicator matrix are assigned to obtain a subjective weight matrix;

[0125] S32, calculating the average value and standard deviation of the construction quality warning indicators in the construction quality warning indicator matrix, and obtaining the standard deviation coefficient of the construction quality warning indicators in the construction quality warning indicator matrix according to the average value and standard deviation, and the calculation formula is as follows:

[0126] ;

[0127] in, represents the standard deviation coefficient, represents the average value of the construction quality early warning indicators in the construction quality early warning indicator matrix, Represents the standard deviation of the construction quality early warning indicator in the construction quality early warning indicator matrix;

[0128] The standard deviation coefficient is normalized, and the normalization result is the objective weight of the construction quality early warning indicator in the construction quality early warning indicator matrix, so as to obtain an objective weight matrix;

[0129] S33, combining the subjective weight matrix and the objective weight matrix and assigning weights to obtain a comprehensive weight matrix, and then calculating the grey correlation coefficient of the construction quality early warning index matrix, the calculation formula is as follows:

[0130] ;

[0131] in, Indicates The first construction quality warning location Grey relational coefficient of secondary assignment;

[0132] A grey correlation coefficient matrix is ​​obtained, and a grey correlation index of the construction quality early warning index is obtained by combining the comprehensive weight matrix, and the grey correlation index is normalized, and the normalization result is the weight of the construction quality early warning index;

[0133] S4, training the BP neural network to obtain a BP neural network model, inputting the construction quality early warning indicator weight into the BP neural network model, outputting the warning level, and completing the construction quality early warning;

[0134] The S4 comprises the following steps:

[0135] S41, the UAV flies in the three-dimensional model of the construction site, relocates the new marking points, generates new construction quality warning indicator weights, and divides the new construction quality warning indicator weights into a weight training set and a weight test set; sets the number of input layer nodes of the BP neural network to the number of new construction quality warning indicators, the output layer nodes to 1, and the learning rate to , introduce the momentum factor to optimize the weight, set the current number of iterations to g, the maximum number of iterations to G, input the weight training set into the BP neural network, and when the current number of iterations reaches the maximum number of iterations, the trained BP neural network is obtained; then input the weight test set into the trained BP neural network, and set the accuracy threshold to , when the output result accuracy is less than When , the BP neural network model is obtained, otherwise the weights are adjusted until the output result accuracy is less than ;

[0136] S42. Input the weights of the construction quality early warning indicators into the BP neural network model. The BP neural network model outputs a result. When the output is 1, it indicates that the warning level is safe, and the construction quality early warning is a very small risk; when the output is 2, it indicates that the warning level is low, and the construction quality early warning is a small risk; when the output is 3, it indicates that the warning level is medium, and the construction quality early warning is a medium risk; when the output is 4, it indicates that the warning level is high, and the construction quality early warning is a high risk; when the output is 5, it indicates that the warning level is special, and the construction quality early warning is a very high risk; the construction quality early warning is completed.

[0137] Example 2

[0138] The present invention also discloses a system for a construction quality early warning method of drone monitoring based on big data, which specifically includes: a construction site three-dimensional model building module, a construction quality early warning position extraction module, a construction quality early warning index weight calculation module and a neural network construction quality early warning module;

[0139] The construction site three-dimensional model building module is used to build a construction site three-dimensional model according to the construction site image;

[0140] The construction quality warning position extraction module is used to extract construction quality warning positions that meet the requirements based on the principal component analysis method;

[0141] The construction quality early warning indicator weight calculation module is used for the construction quality early warning indicator, and the construction quality early warning indicator weight is calculated based on the grey correlation method;

[0142] The neural network construction quality early warning module is used to train the BP neural network, obtain the BP neural network model, output the warning level, and realize the construction quality early warning.

[0143] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0144] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. A construction quality early warning method based on drone monitoring based on big data, characterized in that: The steps include: S1, obtaining an initial construction site image, extracting feature points of the initial construction site image, obtaining feature points of the initial construction site image, then extracting an ultra-high density point cloud of the initial construction site image, constructing a three-dimensional point cloud model, filtering the three-dimensional point cloud model to obtain a three-dimensional model of the construction site, and then performing error evaluation on the three-dimensional model of the construction site; S11, taking an initial construction site image by a drone at the construction site according to a specified route, constructing a source point cloud set and a target point cloud set in the initial construction site image, and obtaining feature points of the initial construction site image by selecting extreme value points of the Gaussian function of the source point cloud set and the target point cloud set; S12, based on the feature points of the initial construction site image, obtaining an ultra-high density point cloud of the initial construction site image; S13, constructing a three-dimensional point cloud model using the ultra-high density point cloud, and processing the non-ground point cloud in the three-dimensional point cloud model using the cloth model to obtain a three-dimensional point cloud model of the construction site; S14, converting the three-dimensional point cloud model of the construction site into a three-dimensional model of the construction site, and implementing error evaluation of the three-dimensional model of the construction site by calculating the root mean square error of the three-dimensional direction in the three-dimensional model of the construction site; S2, selecting marking points in the three-dimensional model of the construction site, extracting the principal component comprehensive index of the marking points based on the principal component analysis method, and deleting the marking points that do not meet the requirements according to the contribution rate of the principal component comprehensive index to obtain the construction quality warning position; S21. The drone selects marking points in the three-dimensional model of the construction site, generates a marking point matrix, establishes a relationship between the marking point matrix and the principal component comprehensive index, and calculates the principal component comprehensive index using the principal component analysis method; S22, calculating the contribution rate of the main component comprehensive index, and making a comprehensive evaluation on the contribution rate of the main component comprehensive index; Set the evaluation threshold to , retain the comprehensive evaluation results greater than Corresponding principal component comprehensive index, delete the comprehensive evaluation results less than or equal to The corresponding principal component comprehensive index is obtained to obtain the screened principal component comprehensive index, and the marked point corresponding to the screened principal component comprehensive index is recorded as the construction quality warning position; S3, assigning values ​​to the construction quality warning positions to obtain construction quality warning indicators, obtaining a comprehensive weight matrix and a gray correlation coefficient matrix based on a gray correlation method, normalizing the comprehensive weight matrix and the gray correlation coefficient matrix to obtain construction quality warning indicator weights; S4. Train the BP neural network to obtain a BP neural network model, input the construction quality early warning indicator weights into the BP neural network model, output the warning level, and complete the construction quality early warning.

2. The method for early warning of construction quality monitored by drones based on big data according to claim 1 is characterized in that: The S12 comprises the following steps: S121, using a multi-view clustering method to segment the initial construction site image, performing clustering processing on all feature points of the initial construction site image, and obtaining an initial construction site image cluster; S122, performing initial feature matching according to the initial construction site image cluster to obtain a feature point matching set, performing expansion processing on the feature point matching set to obtain a three-dimensional point cloud set, and filtering the three-dimensional point cloud set to obtain an ultra-high density point cloud.

3. The method for early warning of construction quality monitored by drones based on big data according to claim 2 is characterized in that: The S13 comprises the following steps: S131, constructing a three-dimensional point cloud model using an ultra-high density point cloud, setting a cloth model, selecting an action point cloud on the cloth model, and flipping the action point cloud, so that the cloth model covers the action point cloud under the action of gravity and elastic force, and performing a first filtering on the non-ground point cloud in the three-dimensional point cloud model according to the shape of the covered action point cloud; S132, the cloth model is vertically offset, and displacement correction is performed under the action of gravity and elastic force, the distance between the action point cloud and the ground point cloud is calculated, the non-ground point cloud is filtered by distance, the second filtering is completed, and the three-dimensional point cloud model of the construction site is obtained.

4. The method for early warning of construction quality monitored by drones based on big data according to claim 3 is characterized in that: The method of extracting the principal component comprehensive index of the marking point based on the principal component analysis method comprises the following steps: S211, calculating the variance of the principal component comprehensive index to maximize the variance of the principal component comprehensive index, obtaining a characteristic root set according to the Lagrange multiplier method, selecting a characteristic root from the characteristic root set that satisfies the maximum variance of the principal component comprehensive index, and obtaining the first principal component comprehensive index; S212. Set the covariance of the second principal component comprehensive index and the first principal component comprehensive index to 0, repeat S211 to obtain the second principal component comprehensive index, and obtain all principal component comprehensive indexes in sequence.

5. The method for early warning of construction quality monitored by unmanned aerial vehicles based on big data according to claim 4 is characterized in that: The S3 comprises the following steps: S31, assigning values ​​to the construction quality warning positions, recording them as construction quality warning indicators, obtaining a construction quality warning indicator matrix, and assigning subjective weights to the construction quality warning indicators in the construction quality warning indicator matrix, obtaining a subjective weight matrix; S32, calculating the standard deviation coefficient of the construction quality warning indicator in the construction quality warning indicator matrix, normalizing the standard deviation coefficient, and obtaining the objective weight matrix by obtaining the normalization result, that is, the objective weight of the construction quality warning indicator in the construction quality warning indicator matrix; S33. Combining the subjective weight matrix and the objective weight matrix and assigning weights to obtain a comprehensive weight matrix and a grey correlation coefficient matrix, normalizing the comprehensive weight matrix and the grey correlation coefficient, and the normalization result is the construction quality early warning indicator weight.

6. The method for early warning of construction quality monitored by drones based on big data according to claim 5 is characterized in that: The S4 comprises the following steps: S41, the UAV flies in the three-dimensional model of the construction site, relocates the new marking points, generates new construction quality warning indicator weights, and divides the new construction quality warning indicator weights into a weight training set and a weight test set; sets the number of input layer nodes of the BP neural network to the number of new construction quality warning indicators, the output layer nodes to 1, and the learning rate to , introduce the momentum factor to optimize the weight, set the current number of iterations to g, the maximum number of iterations to G, input the weight training set into the BP neural network, and when the current number of iterations reaches the maximum number of iterations, the trained BP neural network is obtained; then input the weight test set into the trained BP neural network, and set the accuracy threshold to , when the output result accuracy is less than When , the BP neural network model is obtained, otherwise the weights are adjusted until the output result accuracy is less than ; S42. Input the weights of the construction quality early warning indicators into the BP neural network model. The BP neural network model outputs a result. When the output is 1, it indicates that the warning level is safe, and the construction quality early warning is very small risk; when the output is 2, it indicates that the warning level is low, and the construction quality early warning is small risk; when the output is 3, it indicates that the warning level is medium, and the construction quality early warning is medium risk; when the output is 4, it indicates that the warning level is high, and the construction quality early warning is high risk; when the output is 5, it indicates that the warning level is special, and the construction quality early warning is very high risk; the construction quality early warning is completed.

7. A system for implementing the big data-based UAV monitoring construction quality early warning method as described in any one of claims 1 to 6, characterized in that: Specifically include: Construction site 3D model building module, construction quality warning location extraction module, construction quality warning index weight calculation module and neural network construction quality warning module; The construction site three-dimensional model building module is used to build a construction site three-dimensional model according to the construction site image; The construction quality warning position extraction module is used to extract construction quality warning positions that meet the requirements based on the principal component analysis method; The construction quality early warning indicator weight calculation module is used for the construction quality early warning indicator, and the construction quality early warning indicator weight is calculated based on the grey correlation method; The neural network construction quality early warning module is used to train the BP neural network, obtain the BP neural network model, output the warning level, and realize the construction quality early warning.

Images