Automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm

Abstract

The present invention relates to the technical field of quality management, and discloses an automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithms; the system comprises preprocessing point cloud data and image data, identifying potential quality defects through a neural network recognition model, constructing a BIM model and a point cloud model of a building structure, performing a difference comparison between the point cloud model and the BIM model, assessing the quality grade of the building structure, and generating a quality assessment report for the engineering project; compared with the prior art, the present invention can ensure that the point cloud model and the BIM model can accurately and consistently represent all points of the building structure, and by performing a difference comparison between the point cloud model and the BIM model, the actual quality information of the building structure can be accurately and intuitively assessed and analyzed, and further the quality problems of the building structure in a real three-dimensional scene can be accurately assessed, with high timeliness and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of quality management, and more specifically, to an automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithms. Background Art

[0002] With the rapid development of the construction industry and the acceleration of digital transformation, the assessment of engineering project quality has become particularly important. Building Information Modeling (BIM) realizes the integrated management of building information by creating digital models of buildings. Combined with generative artificial intelligence (AIGC) algorithms, it can provide new ideas for engineering project quality assessment and achieve automatic assessment of engineering project quality.

[0003] The patent application with reference publication number CN117217543A discloses a quality assessment method for construction projects based on BIM, including obtaining construction project information, obtaining and screening similar reference BIM models of related projects and buildings based on big data; obtaining a construction project model by comparing and analyzing the obtained information data and similar BIM model data; simulating the construction project model according to different construction tools, labor, construction materials and other requirements, and outputting multiple simulation data; combining different output simulation data with construction technology assessment data, assessing the construction project from multiple perspectives such as construction safety and economy, and obtaining a risk coefficient; sorting and assessing according to the risk coefficient, screening qualified and optimized construction project models, and performing construction design based on the screened models; storing and filing qualified construction project model data;

[0004] The existing engineering project quality assessment system can identify the quality defects of the building structure at the overall level by collecting simulation data of the building structure and integrating and calculating with weights of different angles, thereby evaluating the quality of the engineering project at a macro level. For example, in the above-mentioned patent application, it uses the BIM model to simulate and screen standardized construction project models, and simulates according to different needs to conduct quality assessments on different construction project plans. Although it can achieve the quality assessment effect of the engineering project, it cannot accurately identify the specific types of quality defects in the building structure, and cannot ensure the precise alignment effect between the point cloud model and the BIM model, which leads to the subsequent quality assessment results of the building structure being inaccurate, and it is impossible to accurately identify and evaluate the quality defects at the detail points, thereby reducing the accuracy of the quality assessment.

[0005] In view of this, the present invention proposes an automatic assessment system for engineering project quality based on AIGC and BIM point cloud algorithm to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: an automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm, comprising:

[0007] The data acquisition module acquires the point cloud data and image data of the building structure in the engineering project and pre-processes the point cloud data and image data respectively;

[0008] The recognition and judgment module inputs the pre-processed point cloud data and image data into the pre-trained neural network recognition model to identify the potential quality defects of the building structure and determine whether to enter the quality assessment mode;

[0009] The first model module, if entering the quality assessment mode, obtains the basic design data of the building structure, which includes geometric data, attribute data and time data, constructs the BIM model of the building structure, and aligns the point cloud data with the BIM model;

[0010] The second model module builds a point cloud model of the building structure based on the point cloud data and aligns the point cloud model with the BIM model;

[0011] The quality assessment module compares the point cloud model with the BIM model to evaluate the quality level of the building structure, which includes the first-level defect level and the second-level defect level, and generates a quality assessment report for the project.

[0012] Furthermore, the preprocessing method of point cloud data includes:

[0013] All coordinate values ​​of the point cloud data are compared repeatedly in sequence, and the point cloud data with all repeated coordinate values ​​are recorded as repeated point clouds;

[0014] Arrange all duplicate point clouds in order of acquisition time, and delete the last duplicate point cloud;

[0015] Import the remaining point cloud data into Open3D for denoising and downsampling, calculate the normal vector of the remaining point cloud data through the PCL algorithm, and calculate the curvature of the remaining point cloud data through the numerical integration algorithm;

[0016] Point cloud data with normal vectors at the same vector level and curvature at the same curvature level are recorded as the same type of point clouds. Based on the same type of point clouds, the remaining point cloud data are divided into A point cloud sets, and the A point cloud sets are numbered in ascending order.

[0017] Furthermore, the image data preprocessing method includes:

[0018] Import the image data into the OpenCV library, and denoise the image data using the median filter method to obtain the denoised image data;

[0019] Convert the denoised image data to grayscale to generate a grayscale image, and mark the pixel value of each pixel in the grayscale image;

[0020] Pixels with the same pixel value are recorded as similar pixels. The boundary contours of similar pixels in the grayscale image are identified through the edge detection algorithm. After drawing lines along the boundary contours, the grayscale image is segmented into B adjacent feature areas.

[0021] Furthermore, potential quality defects include no defects, crack defects, deformation defects, and misalignment defects;

[0022] The training methods of the neural network recognition model include:

[0023] Collect multiple sets of point cloud data and image data in advance when potential quality defects are in the state of no defect, crack defect, deformation defect and dislocation defect;

[0024] Each set of point cloud data and image data is marked as a training feature, and the potential quality defects of each set of training features are annotated. No defects are converted to 0, crack defects are converted to 1, deformation defects are converted to 2, and misalignment defects are converted to 3. The annotated training features are divided into a training set and a test set. The training set is used to train the neural network recognition model, and the test set is used to test the neural network recognition model. An error threshold is preset. When the mean of the prediction errors of all training features in the test set is less than the error threshold, the neural network recognition model is output.

[0025] Furthermore, the method for determining whether to enter the quality assessment mode includes:

[0026] When the output of the neural network recognition model is 0, it is determined not to enter the quality assessment mode;

[0027] When the output of the neural network recognition model is 1, it is determined to enter the quality assessment mode;

[0028] When the output of the neural network recognition model is 2, it is determined to enter the quality assessment mode;

[0029] When the output of the neural network recognition model is 3, it is determined to enter the quality assessment mode.

[0030] Furthermore, the method of constructing the BIM model includes:

[0031] Through natural language processing technology, the key semantics of all data in the design and construction drawings are identified one by one, and the keywords of the key semantics are marked;

[0032] The key semantics with the keywords of geometry, attribute and time are recorded as geometry semantics, attribute semantics and time semantics respectively, and the data corresponding to the geometry semantics, attribute semantics and time semantics are recorded as geometry data, attribute data and time data respectively;

[0033] A model outline having three model positions is selected from the BIM library, and the three model positions are respectively recorded as a first model position, a second model position, and a third model position;

[0034] The geometric data, attribute data and time data are synchronously imported into the first model position, the second model position and the third model position of the model outline to construct a BIM model.

[0035] Furthermore, the method for constructing the point cloud model includes:

[0036] Bind the point cloud data and the corresponding normal vectors one by one to generate an input unit;

[0037] Import all input units in the A point cloud set into the Delaunay triangulation algorithm in ascending order, and convert the input units into triangular meshes to obtain C triangular meshes.

[0038] According to the points of the building structure corresponding to the triangular meshes, C triangular meshes are spliced ​​and combined in sequence from bottom to top to generate a splicing model;

[0039] The edges of two adjacent triangle meshes that touch each other are recorded as splicing edges, and the starting and ending points of all splicing edges in the splicing model are identified one by one through computer vision technology;

[0040] When the first and last points of the edges of the two triangular meshes corresponding to the splicing edge coincide, the splicing model is recorded as a point cloud model;

[0041] When the first and last points of the edges of two triangular meshes corresponding to the splicing edge do not coincide, the first and last points of the edges of the two triangular meshes are adjusted to be closer to each other until the first and last points coincide, thereby obtaining a point cloud model.

[0042] Further, the method of aligning the point cloud model with the BIM model includes:

[0043] The points in the point cloud model are preliminarily aligned with the points in the BIM model using the SIFT algorithm, and the points in the preliminarily aligned point cloud data are rotated and translated;

[0044] The point in the point cloud model is recorded as the first point, and the point in the BIM model is recorded as the second point. The second point with the smallest distance to the first point is marked from the BIM model through the KD tree data structure, and the first point and the second point with the smallest distance are recorded as alignment points.

[0045] The transformation value of the alignment point is calculated by the least squares method, and the points in the point cloud model are rotated and translated based on the transformation value until all the points in the point cloud model coincide with all the points in the BIM model in space.

[0046] Furthermore, the evaluation method of the primary defect level and the secondary defect level includes:

[0047] Mark the 3D coordinates of all points in the point cloud model and BIM model one by one, and obtain The first coordinate value and The second coordinate value;

[0048] Calculate one by one through the Euclidean distance formula The first coordinate value to The Euclidean distance of the second coordinate value is obtained Euclidean distance values, and The Euclidean distance values ​​are accumulated one by one and averaged to generate the model difference value;

[0049] The expression of the model difference value is:

[0050] ;

[0051] In the formula, is the model difference value, For the Euclidean distance values;

[0052] The model difference The difference threshold from the preset model Compare;

[0053] when Less than or equal to When the defect level is evaluated, the first defect level is evaluated;

[0054] when Greater than When the secondary defect level is evaluated.

[0055] Furthermore, the method for generating the quality assessment report includes:

[0056] A basic report sheet with two upper and lower report positions is selected from the report sheet library, and the two report positions are recorded as a defect report position and a grade report position from top to bottom;

[0057] Count the number of potential quality defects identified by the neural network recognition model, obtain the defect value, and establish a number of sub-units equal to the defect value in the defect reporting position;

[0058] The identified potential quality defects are imported one by one into the sub-units of the defect report position, and the quality grade is imported into the grade report position to generate a quality assessment report.

[0059] Technical effects and advantages of the automatic project quality assessment system based on AIGC and BIM point cloud algorithm of the present invention:

[0060] The present invention can effectively eliminate duplicate point cloud data and effectively divide the features in the image by preprocessing the point cloud data and the image data, thereby achieving an effective classification effect of the point cloud data and the image data, facilitating the subsequent direct use of the point cloud data and the image data, and combining with the neural network recognition model to preliminarily identify the defect types corresponding to the point cloud data and the image data of the building structure, providing a prerequisite for the subsequent specific quality assessment operation of the building structure, and at the same time, by constructing a BIM model and a point cloud model, and aligning the point cloud model with the BIM model, it can be ensured that the point cloud model and the BIM model can accurately and consistently represent all points of the building structure, providing a reasonable, accurate and effective assessment basis for the subsequent automatic quality assessment of the building structure, and by comparing the difference between the point cloud model and the BIM model, the actual quality information of the building structure can be accurately and intuitively evaluated and analyzed, achieving the effect of artificial intelligence generation of the quality assessment results of the building structure in the engineering project, avoiding the inefficiency caused by manual evaluation and calculation of massive data one by one, and then accurately evaluating the quality problems of the building structure in the real three-dimensional scene, with high timeliness and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of a module of an automatic project quality assessment system based on AIGC and BIM point cloud algorithm provided in the first embodiment of the present invention;

[0062] Figure 2 A flow chart of an automatic project quality assessment method based on AIGC and BIM point cloud algorithm provided in the second embodiment of the present invention;

[0063] Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention;

[0064] Figure 4 A schematic diagram of the structure of a computer-readable storage medium provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0065] 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.

[0066] Example 1: Please refer to Figure 1 As shown, the automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm described in this embodiment includes:

[0067] The data acquisition module acquires the point cloud data and image data of the building structure in the engineering project and pre-processes the point cloud data and image data respectively;

[0068] Point cloud data refers to the three-dimensional coordinate information corresponding to points at different angles and positions in the building structure within the project, which can represent the three-dimensional coordinates of all points in the building structure;

[0069] Image data refers to the image information of the facades and key nodes in the building structure at different positions, angles and lighting conditions within the project, which can represent the image information of the facades and key nodes in the building structure;

[0070] Due to the large overall area and volume of the building structure, the building structure will contain a large number of points, and each point corresponds to a point cloud data. Therefore, the amount of point cloud data of the building structure is also very large. When obtaining the point cloud data of the building structure, it is necessary to use a point cloud scanning device to perform a comprehensive scan of the building structure and then obtain it; specific point cloud scanning equipment includes but is not limited to handheld point cloud scanning equipment, backpack point cloud scanning equipment, stand-type scanning equipment, etc., as long as it can meet the use requirements; when obtaining image data, it is necessary to use an unmanned aerial device equipped with a high-definition camera to capture images of the key nodes and facades pre-marked on the building structure along a pre-planned flight path, so as to obtain a large amount of image data of the facades and key nodes in different positions, different angles and under different lighting conditions.

[0071] After the point cloud data and image data are acquired, since the number of point cloud data and image data is large and the types are complex, there may be some duplication, errors and other point cloud data and image data in the point cloud data and image data. In order to ensure the high quality of the point cloud data and image data, it is necessary to perform preprocessing operations on the acquired point cloud data and image data;

[0072] The preprocessing methods of point cloud data include:

[0073] All coordinate values ​​of the point cloud data are compared repeatedly in sequence, and the point cloud data with all repeated coordinate values ​​are recorded as repeated point clouds; point cloud data is used to represent the three-dimensional coordinates of different points in the building structure, so each point cloud data contains the coordinate values ​​of the X-axis, Y-axis and Z-axis, and the coordinate values ​​of the X-axis, Y-axis and Z-axis are summarized to form the final coordinates of the point;

[0074] Arrange all duplicate point clouds in order of acquisition time, and delete the last duplicate point cloud;

[0075] Import the remaining point cloud data into Open3D for denoising and downsampling, and calculate the normal vector of the remaining point cloud data through the PCL algorithm, and calculate the curvature of the remaining point cloud data through the numerical integration algorithm; Noise removal is an important step in point cloud preprocessing, which can reduce outliers or error points in point cloud data, and downsampling can effectively reduce the amount of point cloud data, thereby reducing processing time and computing load;

[0076] Point cloud data with normal vectors at the same vector level and curvature at the same curvature level are recorded as the same type of point cloud. Based on the same type of point cloud, the remaining point cloud data is divided into A point cloud sets, and the A point cloud sets are numbered in ascending order. The vector level is a preset range between the minimum and maximum values ​​of a normal vector, and the curvature level is a preset range between the minimum and maximum values ​​of a curvature, so that the point cloud data of normal vectors and curvatures of different sizes can be distinguished based on the basis, thereby achieving the effective distinction effect of a large amount of point cloud data.

[0077] It should be noted that the normal vectors and curvatures of the point cloud data in the same point cloud set are kept relatively constant or within a small range of variation, so that the point cloud data in the same point cloud set can represent the points of the building structures of the same type or in similar positions.

[0078] When preprocessing image data, it is necessary to keep the image data clear, complete and unique, and be able to reflect the feature points in each image data, so as to facilitate the subsequent intuitive, accurate and rapid understanding and judgment of the image data;

[0079] The image data preprocessing methods include:

[0080] Import the image data into the OpenCV library, and denoise the image data using the median filter method to obtain the denoised image data; the median filter method is a nonlinear filtering method that achieves denoising by replacing extreme noise points with the median of neighboring pixels. The median filter method is a conventional denoising method;

[0081] Convert the denoised image data to grayscale to generate a grayscale image, and mark the pixel value of each pixel in the grayscale image;

[0082] Pixels with the same pixel value level are recorded as similar pixels. The boundary contours of similar pixels in the grayscale image are identified by the edge detection algorithm. After drawing lines along the boundary contours, the grayscale image is segmented into B adjacent feature regions. The pixel level is used to preset the range between the minimum and maximum values ​​of the pixel values ​​of pixels in the same pixel level, which can provide an accurate basis for distinguishing similar pixels and ensure the accuracy of subsequent feature region segmentation.

[0083] It should be noted that the image information contained in each feature area in the image data corresponds to the image information of the same or similar building structure, so that the same or similar image information in the image data can be uniformly classified, facilitating the subsequent accurate identification of the image data.

[0084] The recognition and judgment module inputs the pre-processed point cloud data and image data into the pre-trained neural network recognition model to identify the potential quality defects of the building structure and determine whether to enter the quality assessment mode;

[0085] After obtaining the pre-processed point cloud data and image data, the pre-processed point cloud data and image data can be imported into the pre-trained neural network recognition model, so that the neural network recognition model can identify the potential quality defects corresponding to the point cloud data and image data according to the input point cloud data and image data, thereby providing a judgment basis for the subsequent quality assessment of the building structure;

[0086] Potential quality defects refer to the defects existing in the building structure corresponding to the different point cloud data and image data of the building structure, so as to represent the specific types of defects of the building structure. Specifically, potential quality defects include no defects, crack defects, deformation defects and dislocation defects;

[0087] No defects means that there are no defects at all points of the building structure, crack defects mean that cracks exist at points of the building structure, deformation defects mean that deformations exist at points of the building structure, and misalignment defects mean that misalignment exists at points of the building structure. Potential quality defects are obtained by collecting a large amount of historical point cloud data and image data corresponding to no defects, crack defects, deformation defects and misalignment defects.

[0088] The training methods of the neural network recognition model include:

[0089] Collect multiple sets of point cloud data and image data in advance when potential quality defects are in the state of no defect, crack defect, deformation defect and dislocation defect;

[0090] Mark each set of point cloud data and image data as training features, annotate the potential quality defects of each set of training features, including no defects, crack defects, deformation defects and misalignment defects, convert no defects, crack defects, deformation defects and misalignment defects into digital annotations, for example, convert no defects into 0, convert crack defects into 1, convert deformation defects into 2, and convert misalignment defects into 3; divide the annotated training features into a training set and a test set; use 70% of the training features as the training set and 30% of the training features as the test set; use the training set to train the neural network recognition model, and use the test set to test the neural network recognition model; preset an error threshold, and when the mean of the prediction errors of all training features in the test set is less than the error threshold, output the neural network recognition model;

[0091] Specifically, the neural network recognition model adopts any one of a support vector machine model or a random forest model; the preset error threshold is pre-set according to the actual accuracy required by the neural network recognition model;

[0092] The calculation formula for the prediction error is:

[0093] ;

[0094] In the formula, is the prediction error, is the group number of the training feature; For the The predicted state value corresponding to the group training features, For the The actual state value corresponding to the group training features.

[0095] By inputting the pre-processed point cloud data and image data into the neural network recognition model, the potential quality defects corresponding to the point cloud data and image data can be identified, and then according to the identified potential quality defects, it is determined whether to enter the quality assessment mode. The quality assessment mode is a mode used to accurately evaluate the specific and detailed quality level and information of the building structure;

[0096] The methods for determining whether to enter the quality assessment mode include:

[0097] When the output of the neural network recognition model is 0, it means that there is no quality defect in the point of the building structure at this time, and it is determined not to enter the quality assessment mode;

[0098] When the output of the neural network recognition model is 1, it means that there are cracks in the points of the building structure at this time, and it is determined to enter the quality assessment mode;

[0099] When the output of the neural network recognition model is 2, it means that there is deformation in the points of the building structure at this time, and it is determined to enter the quality assessment mode;

[0100] When the output of the neural network recognition model is 3, it means that there is a misalignment phenomenon in the points of the building structure at this time, and it is determined to enter the quality assessment mode.

[0101] The first model module obtains the basic design data of the building structure, constructs the BIM model of the building structure, and aligns the point cloud data with the BIM model;

[0102] Basic design data refers to the relevant design parameters of the building structure on the initial construction design drawings, which can represent the type, size and other data of the specific structure of the standard design of the building structure, and provide a data basis for the subsequent construction of the BIM model;

[0103] Basic design data includes geometric data, attribute data and time data; geometric data specifically refers to the shape, size and spatial relationship of buildings in the building structure, including but not limited to the geometric size and shape of components such as walls, beams, columns, roofs, etc., providing geometric size support for components in the BIM model; attribute data refers to manufacturer information such as the physical properties, performance parameters, number of components, etc. of building materials in the building structure; time data refers to time-related information such as the construction progress, maintenance plan, and service life of the building in the building structure.

[0104] Once the basic design data is obtained, the BIM model corresponding to the building structure can be constructed based on the basic design data;

[0105] The construction methods of BIM model include:

[0106] Through natural language processing technology, the key semantics of all data in the design and construction drawings are identified one by one, and the keywords of the key semantics are marked; the key semantics is used to simplify the data in the design and construction drawings, and the keywords are the smallest units that make up the key semantics and can accurately express the meaning of the key semantics;

[0107] The key semantics with the keywords of geometry, attribute and time are recorded as geometry semantics, attribute semantics and time semantics respectively, and the data corresponding to the geometry semantics, attribute semantics and time semantics are recorded as geometry data, attribute data and time data respectively;

[0108] A model outline with three model positions is selected from the BIM library, and the three model positions are recorded as the first model position, the second model position and the third model position respectively; the model outline is the basic structure of the BIM model, and the model position is the position in the model outline used to import geometric data, attribute data and time data, so as to ensure the accurate construction effect of the BIM model;

[0109] The geometric data, attribute data and time data are synchronously imported into the first model position, the second model position and the third model position of the model outline to construct a BIM model.

[0110] When the BIM model is constructed, the BIM model is constructed on the basis of the basic design data, so that the points in the BIM model and the points of the building structure corresponding to the acquired point cloud data do not necessarily maintain an accurate correspondence in the same space. In order to improve the matching degree between the point cloud data and the BIM model and achieve an accurate match between the points of the point cloud data and the points of the BIM model, it is necessary to accurately match the point cloud data with the BIM model;

[0111] When aligning point cloud data with a BIM model, it is necessary to first import the point cloud data and the BIM model into the BIM software, keep the point cloud data and the BIM model consistent in position and direction in the same space, and set a reference object in the same space. At this time, the reference object is used as the reference object, and the corresponding points of the point cloud data and the BIM model are matched one by one with the help of the alignment tool, thereby achieving a fast and accurate alignment effect between the point cloud data and the BIM model. The above-mentioned BIM software and alignment tools are both commonly used auxiliary tools in the prior art.

[0112] The second model module builds a point cloud model of the building structure based on the point cloud data and aligns the point cloud model with the BIM model;

[0113] Point cloud model refers to the representation of the actual points of the building structure in three-dimensional space, which is constructed based on the point cloud data of the building structure. It can accurately represent the three-dimensional coordinates of each point corresponding to the point cloud data, thereby achieving a true simulation effect of the three-dimensional space of the building structure.

[0114] The point cloud model construction methods include:

[0115] Bind the point cloud data and the corresponding normal vectors one by one to generate an input unit;

[0116] All input units in the A point cloud set are synchronously imported into the Delaunay triangulation algorithm in the order of numbering from small to large, and the input units are converted into triangular meshes to obtain C triangular meshes; the Delaunay triangulation algorithm is a commonly used algorithm for converting point cloud data into triangular meshes. The goal of the Delaunay triangulation algorithm is to divide a given point cloud into a number of non-intersecting triangles that satisfy a specific property;

[0117] According to the points of the building structure corresponding to the triangular meshes, C triangular meshes are spliced ​​and combined in sequence from bottom to top to generate a splicing model;

[0118] The edges of two adjacent triangular meshes that touch each other are recorded as splicing edges, and the starting and ending points of all the splicing edges in the splicing model are identified one by one through computer vision technology; the starting and ending points are the two end points of the edge of the triangular mesh, the end point of the edge of the triangular mesh located at the lower position is recorded as the starting point, and the end point of the edge of the triangular mesh located at the upper position is recorded as the ending point, and the starting and ending points are used as the basis for subsequent judgment on whether the splicing edges overlap;

[0119] When the first and last points of the edges of the two triangular meshes corresponding to the splicing edge coincide, there is no phenomenon that the splicing edge is not closed, and the splicing model is recorded as a point cloud model;

[0120] When the first and last points of the edges of two triangular meshes corresponding to the splicing edge do not coincide, the first and last points of the edges of the two triangular meshes are adjusted to be closer to each other until the first and last points coincide, thereby obtaining a point cloud model.

[0121] After obtaining the point cloud model, it is necessary to align the point cloud model with the BIM model one by one to ensure that the constructed point cloud model can accurately represent all points in space with the BIM model, and ensure that the point cloud model and the BIM model can truly and accurately represent the points on the building structure in a unified manner;

[0122] Methods for aligning point cloud models with BIM models include:

[0123] The points in the point cloud model are preliminarily aligned with the points in the BIM model through the SIFT algorithm, and the points in the preliminarily aligned point cloud data are rotated and translated; the SIFT algorithm is the scale-invariant feature transform algorithm, which is a common method for describing local features of images in the field of image processing and computer vision;

[0124] The point in the point cloud model is recorded as the first point, and the point in the BIM model is recorded as the second point. The second point with the smallest distance to the first point is marked from the BIM model through the KD tree data structure, and the first point and the second point with the smallest distance are recorded as alignment points.

[0125] The transformation value of the alignment point is calculated by the least squares method, and the points in the point cloud model are rotated and translated based on the transformation value until all the points in the point cloud model coincide with all the points in the BIM model in space.

[0126] Through the above steps of spatial alignment of the point cloud model and the BIM model, the point cloud model and the BIM model can be effectively and accurately aligned to ensure that the point cloud model and the BIM model can represent all points of the building structure in the same way, and provide a reasonable, accurate and effective evaluation basis for the subsequent automatic evaluation of the quality of the building structure.

[0127] The quality assessment module compares the point cloud model with the BIM model, evaluates the quality level of the building structure, and generates a quality assessment report for the project;

[0128] Difference comparison refers to the comparison of the differences between the points of the actual building structure represented by the point cloud model and the points of the standard building structure represented by the BIM model. That is, the differences between the actual building structure and the standard building structure can be compared, and the quality level of the building structure can be evaluated by comparing the differences.

[0129] The quality grade is used to indicate the quality of the building structure. Specifically, the quality grade includes the first defect grade and the second defect grade. The first defect grade means that the defect of the building structure is relatively light, and the quality defect of the building structure is slightly poor. The second defect grade means that the defect of the building structure is relatively heavy, and the quality defect of the building structure is relatively serious.

[0130] The evaluation methods for primary defect level and secondary defect level include:

[0131] Mark the 3D coordinates of all points in the point cloud model and BIM model one by one, and obtain The first coordinate value and The second coordinate value;

[0132] Calculate one by one through the Euclidean distance formula The first coordinate value to The Euclidean distance of the second coordinate value is obtained Euclidean distance values, and The Euclidean distance values ​​are accumulated one by one and then averaged to generate the model difference value; Euclidean distance is a commonly used distance definition, which is the real distance between two points in multidimensional space, so that the similarity between all points in the point cloud model and the BIM model can be expressed. When the Euclidean distance value is larger, it means that the similarity between the point cloud model and the BIM model is smaller, and the difference between the two is greater;

[0133] The expression of the model difference value is:

[0134] ;

[0135] In the formula, is the model difference value, For the Euclidean distance values;

[0136] The model difference The difference threshold from the preset model Comparison; The preset model difference threshold refers to the maximum value of the model difference value corresponding to the first-level defect level when the quality level of the building structure is the first-level defect level, which can determine the size of the actual model difference value, thereby providing a basis for the accurate evaluation of the first-level defect level and the second-level defect level;

[0137] when Less than or equal to When , it means that the difference between the point cloud model and the BIM model is small, and the actual building structure has a minor defect, and a first-level defect level is assessed;

[0138] when Greater than When , it means that the difference between the point cloud model and the BIM model is large, and the actual building structure has serious defects, then a secondary defect level is assessed.

[0139] The quality assessment report is a report used to express the quality results of the actual building structure in the project, and can provide intuitive and accurate quality information of the project to the management department and the assessors, and effectively display the output information of the automatic quality assessment system;

[0140] The method for generating a quality assessment report includes:

[0141] A basic report sheet with two report positions arranged up and down is selected from the report sheet library, and the two report positions are recorded as defect report positions and grade report positions from top to bottom; the basic report sheet is the basis for constructing the quality assessment report sheet, and the report position is the smallest unit for constructing the basic report sheet, and provides accurate position restrictions for importing potential quality defects and quality grades;

[0142] Count the number of potential quality defects identified by the neural network recognition model, obtain the defect value, and establish a number of sub-units equal to the defect value in the defect reporting position;

[0143] The identified potential quality defects are imported one by one into the sub-units of the defect report position, and the quality grade is imported into the grade report position to generate a quality assessment report.

[0144] The quality assessment report can help management departments and evaluators to intuitively and clearly view and understand the actual quality and defects of the project, thereby helping the automatic quality assessment system to achieve the effect of artificial intelligence generated content (AIGC). There is no need for manual calculation of large amounts of data for the project, which greatly improves the work efficiency of the automatic quality assessment system.

[0145] In this embodiment, by preprocessing the point cloud data and the image data, the duplicate point cloud data can be effectively eliminated, and the features in the image can be effectively divided, so as to achieve the effective classification effect of the point cloud data and the image data, facilitate the subsequent direct use of the point cloud data and the image data, and combine the neural network recognition model to preliminarily identify the defect types corresponding to the point cloud data and the image data of the building structure, so as to provide a prerequisite for the subsequent specific quality assessment operation of the building structure. At the same time, by constructing the BIM model and the point cloud model, and aligning the point cloud model with the BIM model, it can be ensured that the point cloud model and the BIM model can accurately and consistently represent all points of the building structure, and provide a reasonable, accurate and effective assessment basis for the subsequent automatic quality assessment of the building structure. By comparing the difference between the point cloud model and the BIM model, the actual quality information of the building structure can be accurately and intuitively evaluated and analyzed, and the effect of artificial intelligence generation of the quality assessment results of the building structure in the engineering project is achieved, avoiding the inefficiency caused by manual evaluation and calculation of massive data one by one, and then the quality problems of the building structure in the real three-dimensional scene can be accurately evaluated, with high timeliness and accuracy.

[0146] Example 2: Please refer to Figure 2 As shown, the part not described in detail in this embodiment is described in the first embodiment, and an automatic evaluation method for engineering project quality based on AIGC and BIM point cloud algorithm is provided, which is implemented by an automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm, including:

[0147] S1: Obtain point cloud data and image data of building structures in engineering projects, and pre-process the point cloud data and image data respectively;

[0148] S2: Input the pre-processed point cloud data and image data into the pre-trained neural network recognition model to identify potential quality defects of the building structure and determine whether to enter the quality assessment mode;

[0149] S3: If the quality assessment mode is entered, the basic design data of the building structure is obtained, which includes geometric data, attribute data and time data, and a BIM model of the building structure is constructed, and the point cloud data is registered with the BIM model;

[0150] S4: Based on the point cloud data, a point cloud model of the building structure is constructed and the point cloud model is aligned with the BIM model;

[0151] S5: Compare the point cloud model with the BIM model to evaluate the quality level of the building structure, which includes the first-level defect level and the second-level defect level, and generate a quality assessment report for the project.

[0152] Example 3: Please refer to Figure 3 As shown, this embodiment discloses an electronic device, including a processor and a memory;

[0153] Wherein, the memory stores a computer program that can be called by the processor;

[0154] The processor executes the automatic assessment method for engineering project quality based on AIGC and BIM point cloud algorithm by calling the computer program stored in the memory.

[0155] Since the electronic device introduced in this embodiment is an electronic device used to implement the automatic assessment method for engineering project quality based on AIGC and BIM point cloud algorithms in Example 2 of this application, based on the automatic assessment method for engineering project quality based on AIGC and BIM point cloud algorithms introduced in the embodiment of this application, a person skilled in the art can understand the specific implementation method of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as a person skilled in the art implements the electronic device used by the automatic assessment method for engineering project quality based on AIGC and BIM point cloud algorithms in the embodiment of this application, it belongs to the scope of protection of this application.

[0156] Example 4: Please refer to Figure 4 As shown, this embodiment discloses a computer-readable storage medium on which a rewritable computer program is stored;

[0157] When the computer program is run, the automatic evaluation method for engineering project quality based on AIGC and BIM point cloud algorithm is implemented.

[0158] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. The automatic quality assessment system of engineering projects based on AIGC and BIM point cloud algorithm is characterized by: include: The data acquisition module acquires the point cloud data and image data of the building structure in the engineering project and pre-processes the point cloud data and image data respectively; The recognition and judgment module inputs the pre-processed point cloud data and image data into the pre-trained neural network recognition model to identify the potential quality defects of the building structure and determine whether to enter the quality assessment mode; The first model module, if entering the quality assessment mode, obtains the basic design data of the building structure, which includes geometric data, attribute data and time data, constructs the BIM model of the building structure, and aligns the point cloud data with the BIM model; The second model module builds a point cloud model of the building structure based on the point cloud data and aligns the point cloud model with the BIM model; The quality assessment module compares the point cloud model with the BIM model to evaluate the quality level of the building structure, which includes the first-level defect level and the second-level defect level, and generates a quality assessment report for the project.

2. The automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm according to claim 1 is characterized in that: The preprocessing methods of point cloud data include: All coordinate values ​​of the point cloud data are compared repeatedly in sequence, and the point cloud data with all repeated coordinate values ​​are recorded as repeated point clouds; Arrange all duplicate point clouds in order of acquisition time, and delete the last duplicate point cloud; Import the remaining point cloud data into Open3D for denoising and downsampling, calculate the normal vector of the remaining point cloud data through the PCL algorithm, and calculate the curvature of the remaining point cloud data through the numerical integration algorithm; The point cloud data whose normal vectors are at the same vector level and whose curvatures are at the same curvature level are recorded as the same type of point clouds. Based on the same type of point clouds, the remaining point cloud data are divided into A point cloud sets, and the A point cloud sets are numbered in ascending order.

3. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 2 is characterized in that: The image data preprocessing methods include: Import the image data into the OpenCV library, and denoise the image data using the median filter method to obtain the denoised image data; Convert the denoised image data to grayscale to generate a grayscale image, and mark the pixel value of each pixel in the grayscale image; Pixels with the same pixel value are recorded as similar pixels. The boundary contours of similar pixels in the grayscale image are identified through the edge detection algorithm. After drawing lines along the boundary contours, the grayscale image is segmented into B adjacent feature areas.

4. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 3 is characterized in that: Potential quality defects include no defects, crack defects, deformation defects and misalignment defects; The training methods of the neural network recognition model include: Collect multiple sets of point cloud data and image data in advance when potential quality defects are in the state of no defect, crack defect, deformation defect and dislocation defect; Each set of point cloud data and image data is marked as a training feature, and the potential quality defects of each set of training features are annotated. No defects are converted to 0, crack defects are converted to 1, deformation defects are converted to 2, and misalignment defects are converted to 3. The annotated training features are divided into a training set and a test set. The training set is used to train the neural network recognition model, and the test set is used to test the neural network recognition model. An error threshold is preset. When the mean of the prediction errors of all training features in the test set is less than the error threshold, the neural network recognition model is output.

5. The automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm according to claim 4 is characterized in that: The methods for determining whether to enter the quality assessment mode include: When the output of the neural network recognition model is 0, it is determined not to enter the quality assessment mode; When the output of the neural network recognition model is 1, it is determined to enter the quality assessment mode; When the output of the neural network recognition model is 2, it is determined to enter the quality assessment mode; When the output of the neural network recognition model is 3, it is determined to enter the quality assessment mode.

6. The automatic evaluation system for engineering project quality based on AIGC and BIM point cloud algorithm according to claim 5 is characterized in that: The construction methods of BIM model include: Through natural language processing technology, the key semantics of all data in the design and construction drawings are identified one by one, and the keywords of the key semantics are marked; The key semantics with the keywords of geometry, attribute and time are recorded as geometry semantics, attribute semantics and time semantics respectively, and the data corresponding to the geometry semantics, attribute semantics and time semantics are recorded as geometry data, attribute data and time data respectively; A model outline having three model positions is selected from the BIM library, and the three model positions are respectively recorded as a first model position, a second model position, and a third model position; The geometric data, attribute data and time data are synchronously imported into the first model position, the second model position and the third model position of the model outline to construct a BIM model.

7. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 6 is characterized in that: The point cloud model construction methods include: Bind the point cloud data and the corresponding normal vectors one by one to generate an input unit; Import all input units in the A point cloud set into the Delaunay triangulation algorithm in ascending order, and convert the input units into triangular meshes to obtain C triangular meshes. According to the points of the building structure corresponding to the triangular meshes, C triangular meshes are spliced ​​and combined in sequence from bottom to top to generate a splicing model; The edges of two adjacent triangle meshes that touch each other are recorded as splicing edges, and the starting and ending points of all splicing edges in the splicing model are identified one by one through computer vision technology; When the first and last points of the edges of the two triangular meshes corresponding to the splicing edge coincide, the splicing model is recorded as a point cloud model; When the first and last points of the edges of two triangular meshes corresponding to the splicing edge do not coincide, the first and last points of the edges of the two triangular meshes are adjusted to be closer to each other until the first and last points coincide, thereby obtaining a point cloud model.

8. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 7 is characterized in that: Methods for aligning point cloud models with BIM models include: The points in the point cloud model are preliminarily aligned with the points in the BIM model using the SIFT algorithm, and the points in the preliminarily aligned point cloud data are rotated and translated; The point in the point cloud model is recorded as the first point, and the point in the BIM model is recorded as the second point. The second point with the smallest distance to the first point is marked from the BIM model through the KD tree data structure, and the first point and the second point with the smallest distance are recorded as alignment points. The transformation value of the alignment point is calculated by the least squares method, and the points in the point cloud model are rotated and translated based on the transformation value until all the points in the point cloud model coincide with all the points in the BIM model in space.

9. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 8 is characterized in that: The evaluation methods for primary defect level and secondary defect level include: Mark the 3D coordinates of all points in the point cloud model and BIM model one by one, and obtain The first coordinate value and The second coordinate value; Calculate one by one through the Euclidean distance formula The first coordinate value to The Euclidean distance of the second coordinate value is obtained Euclidean distance values, and The Euclidean distance values ​​are accumulated one by one and averaged to generate the model difference value; The expression of the model difference value is: ； In the formula, is the model difference value, For the Euclidean distance values; The model difference The difference threshold from the preset model Compare; when Less than or equal to When the defect level is evaluated, the first defect level is evaluated; when Greater than When the secondary defect level is evaluated.

10. The automatic quality assessment system for engineering projects based on AIGC and BIM point cloud algorithm according to claim 9 is characterized in that: The method for generating a quality assessment report includes: A basic report sheet with two upper and lower report positions is selected from the report sheet library, and the two report positions are recorded as a defect report position and a grade report position from top to bottom; Count the number of potential quality defects identified by the neural network recognition model, obtain the defect value, and establish a number of sub-units equal to the defect value in the defect reporting position; The identified potential quality defects are imported one by one into the sub-units of the defect report position, and the quality grade is imported into the grade report position to generate a quality assessment report.

Images