Building decoration quality analysis system based on BIM technology

Abstract

The invention provides a building decoration quality analysis system based on a BIM technology, and belongs to the technical field of quality analysis, and the system comprises a collection module which is used for collecting related data of building decoration engineering to construct a standard three-dimensional BIM model; the association binding module is used for acquiring various data in the construction process in real time based on data acquisition equipment in the building decoration construction process, arranging the various data according to a preset format specification, and carrying out association binding on the data and corresponding building decoration units in the standard three-dimensional BIM model; and the quality analysis module is used for analyzing a deviation set of actual building decoration according to the predefined index of each building decoration unit, marking the corresponding building decoration unit to obtain single quality information and total quality information, generating a quality analysis report and outputting the quality analysis report. The problems of data splitting, fuzzy positioning and non-uniform standards in traditional quality control are solved, the efficiency and accuracy of quality analysis are improved, and a reliable basis is provided for quality tracing and rectification of decoration projects.

Description

Technical Field

[0001] This invention relates to the field of quality analysis technology, and in particular to a building decoration quality analysis system based on BIM technology. Background Technology

[0002] Building decoration projects are characterized by complex procedures, diverse materials, and close spatial connections, and their quality control directly affects the building's functionality and safety performance. Traditional decoration quality analysis mainly relies on manual inspections, paper records, and comparison with two-dimensional drawings, which has the following significant limitations:

[0003] Fragmented quality data: Data such as material parameters, geometric dimensions, and process records generated during construction are scattered and stored in different systems (such as Excel spreadsheets, supervision logs, and equipment terminals), lacking a unified data format and correlation mechanism, making it difficult to achieve full-chain traceability from design standards to construction process to quality results.

[0004] Insufficient spatial positioning accuracy: Traditional methods use two-dimensional drawings to mark quality problems, which cannot accurately correspond to specific decoration units in three-dimensional space (such as a certain area of ​​wall tiles or a certain section of ceiling joists), resulting in vague positioning deviations and easy to cause repeated construction or omissions during rectification.

[0005] Quality analysis lag: Manual inspection has a long cycle, and there is a time lag between deviation identification and report generation. It is difficult to provide real-time early warning of quality risks during construction. Rectification is often carried out only after problems have accumulated, which increases rework costs.

[0006] Missing Attribute Information Correlation: Decoration quality is strongly correlated with material type (such as fixed / non-fixed parts), manufacturer, installation process, etc., but traditional analysis has not established a dynamic correlation model between attribute information and quality indicators, making it impossible to formulate differentiated quality standards for different building decoration units (such as wall cladding and movable furniture).

[0007] Therefore, this invention proposes a building decoration quality analysis system based on BIM technology. Summary of the Invention

[0008] This invention provides a building decoration quality analysis system based on BIM technology, which integrates BIM technology with full-cycle construction data to achieve accurate analysis and efficient traceability of decoration quality. This solves the problems of data fragmentation, ambiguous positioning, and delayed analysis in traditional quality control, thereby improving the quality control level of building decoration projects.

[0009] This invention provides a building decoration quality analysis system based on BIM technology, comprising:

[0010] The collection module is used to collect relevant data on building decoration projects and use BIM modeling software to construct a standard three-dimensional BIM model according to theoretical building dimensions, decoration design requirements, and material parameters. During the modeling process, each building decoration unit is assigned a unique identification code, and the corresponding attribute information of the building decoration unit is entered. The attribute information is related to the building decoration type, the name, model, material, and manufacturer of the spatial components involved. The building decoration type includes the types of fixed spatial components and non-fixed spatial components involved in the corresponding building decoration unit, which can be a single type or a combination of types.

[0011] The association and binding module is used to collect various data in real time during the construction process based on data acquisition equipment, and to organize the data according to a preset format specification and associate and bind it with the corresponding building decoration unit in the standard three-dimensional BIM model.

[0012] The quality analysis module is used to analyze the deviation set of the actual building decoration based on the predefined indicators of each building decoration unit, and to annotate the corresponding building decoration unit to obtain individual quality information and total quality information, generate a quality analysis report and output it.

[0013] Preferably, the collection module includes:

[0014] The structure acquisition unit is used to acquire standard construction drawings of the target construction site from the relevant data, extract the three-dimensional spatial structure of the standard construction drawings, and coarsely divide the three-dimensional spatial structure according to regional characteristics to obtain several coarse spaces.

[0015] The subspace determination unit is used to find existing diagonal vertices based on the three-dimensional spatial boundary of the coarse space, and to determine the line intersection points and the single-cell subspace enclosed by the diagonal lines and the three-dimensional spatial boundary based on the diagonal lines formed by the diagonal vertices and the distribution of the diagonal lines in the coarse space.

[0016] The face acquisition unit is used to obtain the maximum cutting face of each single cell subspace and obtain the sub-intersection points of the maximum cutting face and the diagonal;

[0017] The complexity determination unit is used to determine the spatial complexity of the corresponding coarse space based on the distribution of the diagonals, the distribution of line intersections, the distribution of sub-intersections, and the boundary rules of the three-dimensional space boundary.

[0018] The coefficient determination unit is used to determine the dependency coefficient of the single cell space based on the boundary rules of the single cell space, the space occupancy of the single cell space, and the space complexity.

[0019] The radius calculation unit is used to regard single-cell subspaces with a coefficient greater than a preset coefficient as first subspaces, lock the spatial center point of the first subspace, determine the nearest and farthest distances between the spatial center point and the first subspace, the nearest distance between the spatial center point and the maximum cutting surface, and perform average calculations to obtain the initial radius.

[0020] The spatial division unit is used to obtain the cutting grid radius of the corresponding coarse space based on all initial radii, combined with the corresponding support coefficient and the radius weight determined by the relationship between the maximum cutting surface and the number of diagonal vertices involved in the corresponding single grid unit, and to divide the corresponding coarse space according to the cutting grid radius to obtain the building decoration unit.

[0021] Preferably, the association binding module includes:

[0022] A matrix construction unit is used to construct a multidimensional data mapping matrix, wherein the row vectors of the matrix correspond to each building decoration unit in the standard three-dimensional BIM model, and the column vectors correspond to preset data dimensions. The data dimensions include at least geometric parameter dimensions, time dimensions, material property dimensions, and process standard dimensions.

[0023] The mapping unit is used to perform spatiotemporal coordinate calibration on the collected data. Through the absolute coordinate system of the BIM model, the relative coordinate data obtained by different acquisition devices are uniformly mapped to the standard coordinate system.

[0024] Preferred options also include:

[0025] The index library construction module is used to construct a unit-spatiotemporal-viewpoint 3D image index library by selecting key process moments and multi-pose images of the corresponding building decoration unit from the construction data collection library based on the unique identification code of each building decoration unit in the standard 3D BIM model before performing spatiotemporal coordinate calibration on the collected data.

[0026] The function construction module is used to combine the intrinsic material parameters and construction environment data of the building decoration units in the standard 3D BIM model to construct the dynamic response surface function R(m,e,t)=f(M BIM E sensor ,T time ), and extract local feature vectors, where m is the material parameter vector, e is the environmental parameter vector, t is the time variable, and f(M) is the local feature vector. BIM E sensor ,T time M is a Gaussian process regression function trained using historical construction data; BIM E sensor T time These are the regression parameters based on m, e, and t, respectively.

[0027] The offset calculation module is used to extract the geometric contours of the spatial components involved in each multi-pose image in the 3D image index library through the edge detection algorithm and compare them with the theoretical contours of the standard 3D BIM model to calculate the contour offset coefficient.

[0028] The enhancement module is used to enhance the edge clarity and geometric regularity of spatial components in the corresponding pose image based on the theoretical geometric contour and contour offset coefficient of the standard 3D BIM model, and through the contour alignment attention mechanism.

[0029] The kernel determination module is used to determine the dynamically scaled convolution kernel based on the local feature vector and the actual feature vector generated by the actual environmental data and actual material parameters of the corresponding building decoration unit, so as to separate and enhance the inherent material texture of the spatial components in the corresponding building decoration unit.

[0030] The enhancement module is used to generate an intermediate enhanced image Tmid = G·Ie + (1-G)·It based on the geometric enhancement result and the texture enhancement result, where G is the fusion gate signal automatically generated based on the contour offset coefficient, Ie is the geometric enhancement result, and It is the texture enhancement result;

[0031] The index determination module is used to extract the frequency domain features and structural features of the intermediate enhanced image, compare them with the theoretical frequency domain and structural features of the standard 3D BIM model, calculate the enhancement deviation index, and continue to optimize the intermediate enhanced image based on the enhancement deviation index until the enhancement deviation index reaches the index standard.

[0032] Preferably, the convolution kernel determination module includes:

[0033] The difference construction unit is used to construct the global feature difference between the local feature vector and the actual feature vector. At the same time, it mines and compares the individual elements under the same element of the local feature vector and the actual feature vector to obtain the deep feature difference.

[0034] The difference fusion unit is used to generate a fusion vector based on the global feature differences and the deep feature differences, and input it into the lightweight graph convolutional decision network to output the three-dimensional dynamic parameters of the convolutional kernel.

[0035] Among them, the three-dimensional dynamic parameters include: the asymmetric convolution kernel size dynamically generated based on the ratio of material texture period to edge density; the weight allocation coefficients of the material texture branch and noise suppression branch output according to the conflict degree between the environmental interference vector and the intrinsic properties of BIM materials; and the parameter correlation degree and constraint parameter mutation of the current convolution kernel and the historical frame convolution kernel calculated in combination with the timestamp information of the construction process.

[0036] The verification unit is used to incorporate the construction process knowledge base to perform compliance verification on the dynamically generated convolution kernel parameters.

[0037] For painting processes, the anisotropic scaling direction of the convolution kernel is forced to align with the movement direction of the construction tool;

[0038] If it is a paving process, the response intensity of the convolution kernel in the joint area is limited according to the paving joint width parameters of the standard 3D BIM model.

[0039] Otherwise, keep the generated convolution kernel parameters unchanged.

[0040] Preferably, the quality analysis module includes:

[0041] The indicator determination unit is used to match the corresponding predefined indicator from the attribute-indicator lookup table according to the unit attributes of each building decoration unit, and regard it as the first indicator;

[0042] The deviation determination unit is used to extract the index deviation by performing feature comparison on the actual building decoration based on the index contents of each first index.

[0043] The set acquisition unit is used to obtain the deviation set based on the combination of deviations of all indicators.

[0044] Preferably, the quality analysis module includes:

[0045] The set mapping unit is used to map each deviation set to the building decoration unit corresponding to the standard 3D BIM model, forming a spatial heat map with topological relationships based on the spatial dimension, a deviation process stage based on the time dimension, and a deviation attribute vector based on the attribute dimension.

[0046] The network determines the unit, which is used to construct a knowledge graph with building decoration units as nodes and deviation associations as edges, and uses a graph neural network to determine the deviation propagation path, the responsible party for the deviation, and the probability of the current combination of deviations causing a quality accident.

[0047] The report generation unit is used to obtain the quality profile of the corresponding building decoration unit based on the determined results, obtain the individual quality information, and at the same time, to integrate and analyze all the individual quality information involved in each construction stage throughout the entire construction cycle to obtain the total quality information.

[0048] A quality analysis report is generated based on individual quality information and total quality information.

[0049] Preferably, the difference building unit includes:

[0050] A position locking subunit is used to lock the first position of a single element in a local feature vector and the second position of a single element in the actual feature vector from an element lookup table, wherein the element lookup table is sorted according to a standard step-decay order of the single elements.

[0051] The first determining subunit is used to determine the first stability coefficient of the corresponding individual element based on the first distance between the first position and the second position;

[0052] The second determining subunit is used to analyze the existence variable of each individual element, and to perform individual element elimination and correlation analysis on each actual feature vector in turn to obtain the influence coefficient and existence coefficient of the eliminated individual element, and match them with the double coefficient comparison table to obtain the second stability coefficient.

[0053] The variable mining subunit is used to find the matching variable that has the highest similarity to each existing variable, and to reconstruct a new vector based on all matching variables;

[0054] The defect acquisition subunit is used to analyze the first global comparison result between the local feature vector and the actual feature vector and the second global comparison result between the local feature vector and the new vector to obtain the first defect and the second defect.

[0055] The difference determination subunit is used to obtain deep feature differences based on the first stability coefficient, the second stability coefficient, and in combination with the first defect and the second defect.

[0056] Compared with the prior art, the beneficial effects of this application are:

[0057] By using BIM models to achieve refined management of building decoration units, and combining full-cycle data collection and correlation, quality deviations can be accurately analyzed and visualized reports can be generated. This solves the problems of data fragmentation, ambiguous positioning, and inconsistent standards in traditional quality control, improves the efficiency and accuracy of quality analysis, and provides a reliable basis for quality traceability and rectification of decoration projects.

[0058] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0059] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is a structural diagram of a building decoration quality analysis system based on BIM technology in an embodiment of the present invention. Detailed Implementation

[0062] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0063] This invention provides a building decoration quality analysis system based on BIM technology, such as... Figure 1 As shown, it includes:

[0064] The collection module is used to collect relevant data on building decoration projects and use BIM modeling software to construct a standard three-dimensional BIM model according to theoretical building dimensions, decoration design requirements, and material parameters. During the modeling process, each building decoration unit is assigned a unique identification code, and the corresponding attribute information of the building decoration unit is entered. The attribute information is related to the building decoration type, the name, model, material, and manufacturer of the spatial components involved. The building decoration type includes the types of fixed spatial components and non-fixed spatial components involved in the corresponding building decoration unit, which can be a single type or a combination of types.

[0065] The association and binding module is used to collect various data in real time during the construction process based on data acquisition equipment, and to organize the data according to a preset format specification and associate and bind it with the corresponding building decoration unit in the standard three-dimensional BIM model.

[0066] The quality analysis module is used to analyze the deviation set of the actual building decoration based on the predefined indicators of each building decoration unit, and to annotate the corresponding building decoration unit to obtain individual quality information and total quality information, generate a quality analysis report and output it.

[0067] In this embodiment, the relevant data refers to the basic data of the entire process of building decoration engineering design, construction, materials, etc., including construction drawings (building floor plan, decoration node details), material list (including model / material / manufacturer), specifications and standards (such as the "Building Decoration and Renovation Engineering Quality Acceptance Standard"), etc.

[0068] BIM modeling software is a professional tool used to build three-dimensional building information models. It can integrate data such as geometric dimensions and material properties. For example, Autodesk Revit (commonly used for architectural and interior modeling) and ArchiCAD (supports complex spatial shapes).

[0069] Theoretical building dimensions are the standard dimensions of building components or spaces specified in the design documents. They serve as the benchmark for construction and quality acceptance. For example, the theoretical dimensions of a living room bay are 3.6m and the theoretical value of a suspended ceiling height is 2.8m. Specifically, dimensional parameters are extracted from the construction drawings and entered into the BIM model as benchmark data.

[0070] Interior design requirements specify the style, craftsmanship, and effect of the renovation, including color matching and material splicing methods. For example, bathroom wall tiles must be laid with a herringbone pattern, and kitchen countertops must be resistant to acid and alkali corrosion. Specifically, the design requirements are converted into parametric constraints in the BIM model (such as "type attributes" in Revit). Material parameters are the physical and chemical properties of materials, which affect the quality and performance of the renovation.

[0071] A standard 3D BIM model is a 3D digital model built based on theoretical dimensions, design requirements, and material parameters. It contains geometric and attribute information of all elements of building decoration. For example, a residential decoration BIM model that includes components such as walls, floors, and furniture can intuitively show the spatial relationships of each part.

[0072] The building decoration unit is a reliable decoration management unit divided according to spatial characteristics and construction logic, which facilitates precise quality control.

[0073] The unique identification code is a unique code assigned to each building decoration unit for data association and quality traceability. For example, the code "ZH-201-L01" represents "Residential - Unit 201 - Living Room First Floor Tile Unit".

[0074] Attribute information is a data set that describes the characteristics of building decoration units and is associated with decoration type, material parameters, etc. For example, the attribute information of the master bedroom wardrobe unit includes: type is space fastener, material is E1 grade particleboard, manufacturer is XX custom furniture. Specifically, an attribute panel is added to each unit in the BIM model, and information is entered manually or in batches.

[0075] Building decoration types are categories classified according to the way spatial components are fixed, and are used to formulate differentiated quality standards.

[0076] Fixed space components are those that cannot be moved after installation, such as wall latex paint and ceiling joists.

[0077] Non-fixed space components are movable parts, such as sofas and dining tables.

[0078] Combination type is a unit that includes both fixed and non-fixed components, such as "kitchen wall cabinets (fixed) + under-counter cabinets (fixed) + pull-out baskets (non-fixed)".

[0079] Data acquisition equipment is a tool used to obtain actual data during the construction process, including geometric measurement equipment and environmental monitoring equipment, such as laser rangefinders (to measure wall flatness), high-definition cameras (to capture the effect of tile laying), and temperature and humidity sensors (to record the construction environment).

[0080] All data are actual information generated during the construction process, including geometric dimensions, actual material parameters, and process times. The preset format specifications are rules for establishing a unified data format to ensure that data from different sources can be integrated and correlated. For example, geometric data is specified to be recorded in "meters" and time data is recorded in the format of "year-month-day-hour-minute".

[0081] Association binding is to establish a mapping relationship between actual construction data and the corresponding building decoration unit in the BIM model, so as to realize the linkage between virtual model and physical construction. For example, the measured value of 2.5mm of the paint thickness on the east wall of the master bedroom is bound to the paint unit on the east wall of the master bedroom in the BIM model (code ZH-201-W01).

[0082] Predefined indicators are quality acceptance standards set for different building decoration units, which vary depending on the unit type. For example, for fixed parts (such as wall tiles): hollow rate ≤5%, joint height difference ≤0.5mm; for non-fixed parts (such as wardrobes): verticality deviation ≤1mm / m, cabinet door gap ≤2mm.

[0083] Actual architectural decoration refers to the physical finishing result after construction is completed, and it is the object of quality analysis.

[0084] Deviation set is the set of differences between the actual building decoration and the predefined indicators, reflecting quality problems. For example, the deviation set of the ceramic tile unit on the east wall of the master bedroom is: hollow rate 8% (3% exceeding the standard) and joint height difference 0.8mm (0.3mm exceeding the standard).

[0085] Annotation is the operation of visualizing the deviation set in the BIM model, which makes it easier to locate quality problems. For example, in the BIM model, hollow areas are marked with red boxes, and the hollow rate is marked as 8%. Specifically, the annotation function of BIM software is used to automatically associate deviation data to generate annotations.

[0086] Single quality information refers to the quality data of a single building decoration unit, reflecting the local quality status.

[0087] Total quality information is a summary of quality data for all building and decoration units, reflecting the overall project quality.

[0088] A quality analysis report is a document that integrates individual quality information, overall quality information, and the causes of deviations, and is used to guide corrective actions.

[0089] The beneficial effects of the above technical solution are: it enables refined management of building decoration units through BIM model, and, combined with full-cycle data collection and correlation, accurately analyzes quality deviations and generates visual reports, which solves the problems of data fragmentation, ambiguous positioning and inconsistent standards in traditional quality control, improves the efficiency and accuracy of quality analysis, and provides a reliable basis for quality traceability and rectification of decoration projects.

[0090] This invention provides a building decoration quality analysis system based on BIM technology, wherein the collection module includes:

[0091] The structure acquisition unit is used to acquire standard construction drawings of the target construction site from the relevant data, extract the three-dimensional spatial structure of the standard construction drawings, and coarsely divide the three-dimensional spatial structure according to regional characteristics to obtain several coarse spaces.

[0092] The subspace determination unit is used to find existing diagonal vertices based on the three-dimensional spatial boundary of the coarse space, and to determine the line intersection points and the single-cell subspace enclosed by the diagonal lines and the three-dimensional spatial boundary based on the diagonal lines formed by the diagonal vertices and the distribution of the diagonal lines in the coarse space.

[0093] The face acquisition unit is used to obtain the maximum cutting face of each single cell subspace and obtain the sub-intersection points of the maximum cutting face and the diagonal;

[0094] The complexity determination unit is used to determine the spatial complexity of the corresponding coarse space based on the distribution of the diagonals, the distribution of line intersections, the distribution of sub-intersections, and the boundary rules of the three-dimensional space boundary.

[0095] The coefficient determination unit is used to determine the dependency coefficient of the single cell space based on the boundary rules of the single cell space, the space occupancy of the single cell space, and the space complexity.

[0096] The radius calculation unit is used to regard single-cell subspaces with a coefficient greater than a preset coefficient as first subspaces, lock the spatial center point of the first subspace, determine the nearest and farthest distances between the spatial center point and the first subspace, the nearest distance between the spatial center point and the maximum cutting surface, and perform average calculations to obtain the initial radius.

[0097] The spatial division unit is used to obtain the cutting grid radius of the corresponding coarse space based on all initial radii, combined with the corresponding support coefficient and the radius weight determined by the relationship between the maximum cutting surface and the number of diagonal vertices involved in the corresponding single grid unit, and to divide the corresponding coarse space according to the cutting grid radius to obtain the building decoration unit.

[0098] In this embodiment, the target construction site is the physical space where the decoration construction is about to take place. It is the actual object of modeling and construction, such as the east side office on the 15th floor of an office building or apartment 302 in a residential building.

[0099] Standard construction drawings are construction reference drawings issued by the design unit that conform to the specifications and include information such as spatial dimensions and component layout.

[0100] The three-dimensional spatial structure is the three-dimensional structural information of the target site extracted from the construction drawings, including length, width, height and spatial relationships of components.

[0101] Regional characteristics are the unique attributes of a target site divided by function or structure, such as "kitchen area" (with stove and water pipes) or "load-bearing wall area" (cannot be demolished or altered).

[0102] Coarse division is an operation that performs preliminary segmentation of the three-dimensional spatial structure according to regional characteristics, resulting in a basic space with a large area or volume. For example, a 100-square-meter house can be roughly divided into "public areas" (living room + dining room, about 50 square meters) and "private areas" (bedroom + bathroom, about 50 square meters) according to function to obtain the coarse space.

[0103] The three-dimensional spatial boundary is the three-dimensional boundary that constitutes the coarse space (such as walls, floors, and ceilings). It is the physical limit of the spatial range. For example, the six faces of a cuboid living room (four walls, one floor, and one ceiling) constitute its three-dimensional spatial boundary.

[0104] Diagonal vertices are the vertices located at opposite ends of a diagonal line on the three-dimensional boundary of a coarse space. They are used to locate the internal structure of the space. For example, in a cuboid living room, the top left corner vertex and the bottom right corner vertex are a pair of diagonal vertices. A diagonal line is a straight line connecting a pair of diagonal vertices. It is used to analyze the distribution of the internal structure of the coarse space. For example, the straight line connecting the diagonal vertices of the living room mentioned above.

[0105] The intersection of lines is the point where multiple diagonals intersect in a coarse space, reflecting the complexity of the spatial structure.

[0106] A single-cell space is a small, enclosed space bounded by diagonals and three-dimensional spatial boundaries. It is the basic sub-unit of a larger space. For example, in a cuboid living room, the two diagonals and the walls enclose four congruent single-cell spaces.

[0107] The largest cutting plane is the plane that can divide a single cell space into two parts along a certain dimension (such as horizontal or vertical) and has the largest area.

[0108] A sub-intersection is the point where the largest cutting plane intersects with the diagonal line within a single cell subspace.

[0109] Space complexity is an indicator reflecting the complexity of the internal structure of a coarse space. It is jointly determined by the diagonal distribution, the number of line intersections, and the regularity of the boundary. Specifically, it is: KF = ε1·Dρ + ε2·Xρ + ε3·Zρ + ε4·Rg, where Dρ, Xρ, Zρ, and Rg are the diagonal distribution density, line intersection density, sub-intersection density, and boundary regularity, respectively, and ε1, ε2, ε3, and ε4 are the curves of the diagonal distribution density, line intersection density, sub-intersection density, and boundary regularity, respectively. Empirically, their values ​​are 0.2, 0.2, 0.3, and 0.3, respectively. Where Δmax is the maximum allowable error, such as 0.5m; Δ is the boundary fitting error, used to measure the average distance between the actual boundary profile and the theoretical regular profile. The diagonal distribution density is the ratio of the total length of the diagonals in the coarse space to the length of the diagonals of the coarse space volume; the line intersection density is the ratio of the total number of line intersections to the total number of nodes in the coarse space grid, the total number of nodes in the coarse space grid is calculated according to a 1m×1m×1m virtual grid; the sub-intersection density is the ratio of the total number of sub-intersections to the total number of single-cell subspaces.

[0110] Boundary regularity describes the regularity of the three-dimensional spatial boundary, such as whether it is a plane or whether there are concavities or convexities. Specifically, it involves obtaining boundary point cloud data through laser scanning and calculating the boundary fitting error (regular boundary error < 5 mm).

[0111] The reliability coefficient is an index that measures the reliability of a single cell space as a partitioning benchmark. It is calculated by comprehensively considering the boundary regularity, space occupancy, and space complexity of the single cell space, and its value ranges from 0 to 1.

[0112] YT=Er·Qr·(1-KF), where YT is the corresponding dependence coefficient; To represent the boundary regularity of the corresponding single-cell unit, Δi is the boundary fitting error of the corresponding single-cell space; Qr is the space utilization rate of the corresponding single-cell unit, and Where V0 is the volume of the fixed component within a single cell space; Vr is the corresponding volume of the single cell space.

[0113] Space occupancy refers to the degree to which a single cell is occupied by fixed components (such as columns or pipelines). For example, if a single cell has a column with a diameter of 30cm, the space occupancy rate is 15%; another cell has no components, so the occupancy rate is 0%.

[0114] The preset coefficient is the threshold for determining whether a single cell space can be used as a partitioning benchmark, and it is set to 0.6 based on engineering experience.

[0115] The first subspace is a single-cell space with a coefficient greater than the preset coefficient, and it is the core benchmark for dividing architectural decoration units.

[0116] The center point of space is the geometric center (midpoint of three-dimensional coordinates) of the first subspace.

[0117] The dependency relation (Re) measures the degree of correlation between the largest cutting plane and the number of intersections with the diagonals (the more intersections, the more significant the dividing effect of the cutting plane on space). For example, if a largest cutting plane intersects with 3 diagonals (3 points), the dependency relation weight is 0.3 (the weight increases with the number of intersections).

[0118] Where Ne is the total number of intersections between the maximum cutting surface and all diagonals in the coarse space; Nmax is the total number of diagonals in the coarse space; and Ae is the ratio of the area of ​​the maximum cutting surface to the total surface area of ​​the coarse space.

[0119] The radius weight is a weight value calculated by combining the initial radius, the dependency coefficient, and the dependency relationship. It is used to determine the final cutting radius, i.e., the cutting mesh radius. The cutting mesh radius is the mesh size used to divide the coarse space, and it is calculated by summing the radius weights of all first subspaces.

[0120] Standardize the initial radius r0 norm = r0 / rmax, where r0 is the initial radius and rmax is the maximum initial radius in the coarse space; r0 norm Standardized values;

[0121] Calculate the radius weight Br: Br = 0.4·YT + 0.3·Re + 0.3·r0 norm .

[0122] Calculate the cutting mesh radius: the sum of the products of each initial radius and its corresponding radius weight / the sum of all radius weights.

[0123] In this embodiment, the cutting grid radius determines the spatial scale of a single grid. That is, the side length of the grid cell must match the radius. Typically, the grid cell is a cube with a side length that is twice the cutting grid radius. Thus, the coarse space is cut into a grid of uniform size, which not only ensures the accuracy of spatial division but also preserves the topological relationship of the original space, providing a structured unit basis for subsequent quality inspection.

[0124] The beneficial effects of the above technical solution are: by refining the analysis of spatial structural characteristics and combining quantitative indicators such as spatial complexity and reliance coefficient, the scientific division of building decoration units is achieved, providing accurate unit benchmarks for subsequent construction data correlation and quality deviation analysis, and improving the refinement and efficiency of building decoration quality control.

[0125] This invention provides a building decoration quality analysis system based on BIM technology, wherein the association and binding module includes:

[0126] A matrix construction unit is used to construct a multidimensional data mapping matrix, wherein the row vectors of the matrix correspond to each building decoration unit in the standard three-dimensional BIM model, and the column vectors correspond to preset data dimensions. The data dimensions include at least geometric parameter dimensions, time dimensions, material property dimensions, and process standard dimensions.

[0127] The mapping unit is used to perform spatiotemporal coordinate calibration on the collected data. Through the absolute coordinate system of the BIM model, the relative coordinate data obtained by different acquisition devices are uniformly mapped to the standard coordinate system.

[0128] In this embodiment, the multidimensional data mapping matrix is, for example, a 3x4 matrix. The row vectors correspond to the living room floor tile unit, the bedroom wall unit, and the kitchen countertop unit, while the column vectors correspond to geometric parameters, time, material properties, and process standards. For example, the geometric parameter of the living room floor tile unit is a flatness deviation of 0.3mm. For example, the row vector of the bedroom wall unit (coded WS-001) includes its geometric parameters (verticality deviation of 0.5mm), time (construction on 2024-06-10), material properties (latex paint with 3000 scrub resistance cycles), and process standards (3 coats). For example, the time dimension of the column vector includes the construction timestamps of all units, such as the living room floor tile unit: 2024-06-05, and the bedroom wall unit: 2024-06-10.

[0129] The preset data dimensions are predefined and are core data categories used to describe the quality of building decoration, covering key factors that affect quality. The geometric parameter dimension describes the spatial dimensions and deviations of building decoration units, such as length deviation, verticality, and flatness. For example, the actual length of the living room floor tile is 2.002m (theoretical 2.0m, deviation +0.002m).

[0130] The time dimension records time-related information related to construction, such as the start time, completion time, and inspection time of each process.

[0131] Material property dimensions are parameters that reflect the physical or chemical properties of a material, such as moisture content, compressive strength, and corrosion resistance.

[0132] The process standard dimension refers to the operational specifications that must be followed during construction, such as the laying method (herringbone / straight seam), the number of coats, and the installation torque.

[0133] The collected data are actual construction process data obtained through data acquisition equipment, including geometric dimensions, environmental parameters, and process records.

[0134] Spatiotemporal coordinate calibration is a standardized process that unifies the spatial location and temporal information of collected data to ensure the consistency of data in time and space.

[0135] The absolute coordinate system of the BIM model is a preset global coordinate system in the BIM model. It uses a fixed origin (such as the bottom left corner of the building) as the reference and describes the spatial position with three-dimensional coordinates (x, y, z). It is the spatial reference of all data. For example, if the bottom left corner of the building is set as the origin (0, 0, 0), then the absolute coordinates of the central chandelier in the living room are (5.2, 3.8, 2.8) (unit: m).

[0136] Relative coordinate data is the location information recorded by the acquisition device based on its own coordinate system or temporary reference points. It is not associated with the absolute coordinates of BIM. The standard coordinate system is the absolute coordinate system of the BIM model. It is the target coordinate system for the final unified mapping of all data, ensuring that the spatial positions of data from different sources can be directly compared.

[0137] The beneficial effects of the above technical solution are: by constructing a multi-dimensional data mapping matrix, a structured association between building decoration units and multi-dimensional data is realized; combined with spatiotemporal coordinate calibration, the consistency of data from different sources in time and space is ensured, providing a standardized and comparable data foundation for subsequent quality deviation analysis, and significantly improving the accuracy and efficiency of building decoration quality analysis.

[0138] This invention provides a building decoration quality analysis system based on BIM technology, which also includes:

[0139] The index library construction module is used to construct a unit-spatiotemporal-viewpoint 3D image index library by selecting key process moments and multi-pose images of the corresponding building decoration unit from the construction data collection library based on the unique identification code of each building decoration unit in the standard 3D BIM model before performing spatiotemporal coordinate calibration on the collected data.

[0140] The function construction module is used to combine the intrinsic material parameters and construction environment data of the building decoration units in the standard 3D BIM model to construct the dynamic response surface function R(m,e,t)=f(M BIM E sensor ,T time ), and extract local feature vectors, where m is the material parameter vector, e is the environmental parameter vector, t is the time variable, and f(M) is the local feature vector. BIM E sensor ,T time M is a Gaussian process regression function trained using historical construction data; BIM E sensor T time These are the regression parameters based on m, e, and t, respectively.

[0141] The offset calculation module is used to extract the geometric contours of the spatial components involved in each multi-pose image in the 3D image index library through the edge detection algorithm and compare them with the theoretical contours of the standard 3D BIM model to calculate the contour offset coefficient.

[0142] The enhancement module is used to enhance the edge clarity and geometric regularity of spatial components in the corresponding pose image based on the theoretical geometric contour and contour offset coefficient of the standard 3D BIM model, and through the contour alignment attention mechanism.

[0143] The kernel determination module is used to determine the dynamically scaled convolution kernel based on the local feature vector and the actual feature vector generated by the actual environmental data and actual material parameters of the corresponding building decoration unit, so as to separate and enhance the inherent material texture of the spatial components in the corresponding building decoration unit.

[0144] The enhancement module is used to generate an intermediate enhanced image Tmid = G·Ie + (1-G)·It based on the geometric enhancement result and the texture enhancement result, where G is the fusion gate signal automatically generated based on the contour offset coefficient, Ie is the geometric enhancement result, and It is the texture enhancement result;

[0145] The index determination module is used to extract the frequency domain features and structural features of the intermediate enhanced image, compare them with the theoretical frequency domain and structural features of the standard 3D BIM model, calculate the enhancement deviation index, and continue to optimize the intermediate enhanced image based on the enhancement deviation index until the enhancement deviation index reaches the index standard.

[0146] In this embodiment, the construction data acquisition library is a database that stores construction images and sensor data (such as MongoDB, which records camera images and timestamps of wall construction).

[0147] Key process moments are the core nodes of the entire construction cycle (such as "completion of wall putty application" and "tile laying acceptance"), which are marked through the process management system.

[0148] Multi-pose images are multi-view images of the same unit (such as upward, downward, and 45° side views of a wall, automatically acquired by a mobile pan-tilt camera).

[0149] A 3D image index library based on unit-spatiotemporal-viewpoint: This index structure associates unit ID, process time, and shooting viewpoint (e.g., built using Elasticsearch, supporting fast retrieval by unit ID + time range + viewpoint). Specifically:

[0150] Export the unit code list from the BIM model and connect it to the API of the construction data acquisition library;

[0151] Filter the corresponding images by unit code and extract the timestamp (accurate to the moment) and shooting angle (analyzed from gyroscope data);

[0152] Build an index library that supports combined queries of cell, time interval, and viewpoint (e.g., querying overhead images of WALL-1F-001 taken between 09:00 and 12:00 on June 1, 2024).

[0153] The intrinsic parameters of a material are its inherent properties (such as the texture period of a ceramic tile being 0.1m and the gloss of wood being 85%, extracted from the manufacturer's parameter table).

[0154] The construction environment data consists of real-time environmental parameters (such as humidity 60%RH and light intensity 1500 lux, collected by temperature and humidity sensors and light meters).

[0155] Dynamic response surface function R(m,e,t): A mathematical model describing the material-environment-time coupling relationship (such as Gaussian process regression, fitting the visual performance of tile texture on the 10th day at 60%+ humidity).

[0156] Local feature vectors are key features extracted from the response surface (such as the frequency domain energy peak of the texture and the reflectivity gradient, which are obtained by dimensionality reduction through principal component analysis).

[0157] The Gaussian process regression function is a regression model trained on historical construction data (such as 500 sets of tile + humidity + time → texture change data) to predict the characteristics of materials in specific environments and at specific times.

[0158] Collect historical construction materials, environment, time, and corresponding image features to construct a training dataset;

[0159] The model is trained using Python's GaussianProcessRegressor, with input (m,e,t) and output predicted features.

[0160] For the current cell, substitute the real-time m, e, t to calculate the local feature vector (e.g., texture frequency 0.05Hz, reflectivity 0.7).

[0161] Edge detection algorithms are algorithms that extract the contours of spatial objects in an image (such as OpenCV's Canny algorithm, which detects the edges of walls and backgrounds).

[0162] Spatial components are physical components of building decoration (such as bedroom wooden doors and kitchen tiled walls, corresponding to "door families" and "wall families" in the BIM model).

[0163] The theoretical profile is the design profile of spatial components in the BIM model (e.g., the theoretical boundary of a wooden door is x = 5.2~6.0m, y = 2.0~2.4m, exported as a JSON coordinate set).

[0164] The profile offset coefficient is the degree of deviation between the actual profile and the theoretical profile (coefficient = average offset distance / design tolerance; for wooden doors, the average offset is 3mm and the tolerance is 5mm, so the coefficient is 0.6).

[0165] Contour alignment attention mechanism is an attention module that focuses on regions with contour deviation (such as the attention layer of Transformer, which assigns high weights to edges with large offsets).

[0166] Geometric regularity refers to the degree to which the outline of a spatial component matches the theoretical outline (e.g., after reinforcement, a broken line on a wall edge becomes a straight line, conforming to the rectangular design in BIM). Specifically:

[0167] Construct a deep learning model (such as U-Net + attention module), inputting the original image and contour offset coefficients;

[0168] The attention module adjusts the weights based on the offset coefficient (areas with larger offsets have higher weights) to enhance edge sharpness;

[0169] Output geometrically enhanced images (e.g., sharper edges on wooden doors, conforming to the theoretical rectangular outline).

[0170] Dynamically scaled convolution kernels are convolution kernels whose size is adaptively adjusted according to the feature vector (e.g., a 7×7 kernel is used when the texture is complex, and a 3×3 kernel is used when it is simple).

[0171] The inherent texture of a material is its own texture (such as the natural texture of wood), and it needs to be separated from stains and reflections (which interfere with the texture).

[0172] The weights generated by the fusion gate signal G based on the contour offset coefficient (e.g., when the offset coefficient is 0.6, G = 0.7, preserving more of the geometric enhancement results).

[0173] The geometric enhancement result Ie is an image with sharp edges (such as straight wall edges that fit the BIM profile).

[0174] The result of texture enhancement is an image with clear texture (such as natural wood texture without reflective interference).

[0175] Frequency domain characteristics are the frequency distribution of an image after Fourier transform (e.g., the frequency domain peak of the tile texture in BIM theory is 0.1Hz, and the deviation after enhancement is 0.02Hz).

[0176] Structural features are the geometric structure of the image (such as the edge direction described by HOG features; in BIM theory, the wall surface has vertical edges, with a 10° deviation after enhancement).

[0177] The enhancement deviation index is the difference between actual and theoretical characteristics, D: D = w1Dp + w2Dj, where Dp is the standardized coefficient of the frequency domain characteristics, Dj is the standardized coefficient of the structural characteristics, and w1 and w2 are the weights, with values ​​of 0.6 and 0.4 respectively.

[0178] In this embodiment, the index standard is a preset deviation threshold (e.g., 0.05, below which the enhancement is considered to have met the standard).

[0179] The beneficial effects of the above technical solution are as follows: By guiding BIM theoretical features, fusing multimodal data, and using a dynamic enhancement network, it solves the problems of large environmental interference, difficulty in correcting geometric deviations, and inaccurate texture extraction in traditional image enhancement. From multi-view image screening to dual-branch enhancement of geometry and texture, and then to deviation closed-loop optimization, it achieves geometric alignment BIM design of building decoration images, restoration of material authenticity in texture, and controllable quantification of deviations, providing high-precision and traceable image data support for subsequent quality analysis.

[0180] This invention provides a building decoration quality analysis system based on BIM technology, wherein the convolution kernel determination module includes:

[0181] The difference construction unit is used to construct the global feature difference between the local feature vector and the actual feature vector. At the same time, it mines and compares the individual elements under the same element of the local feature vector and the actual feature vector to obtain the deep feature difference.

[0182] The difference fusion unit is used to generate a fusion vector based on the global feature differences and the deep feature differences, and input it into the lightweight graph convolutional decision network to output the three-dimensional dynamic parameters of the convolutional kernel.

[0183] Among them, the three-dimensional dynamic parameters include: the asymmetric convolution kernel size dynamically generated based on the ratio of material texture period to edge density; the weight allocation coefficients of the material texture branch and noise suppression branch output according to the conflict degree between the environmental interference vector and the intrinsic properties of BIM materials; and the parameter correlation degree and constraint parameter mutation of the current convolution kernel and the historical frame convolution kernel calculated in combination with the timestamp information of the construction process.

[0184] The verification unit is used to incorporate the construction process knowledge base to perform compliance verification on the dynamically generated convolution kernel parameters.

[0185] For painting processes, the anisotropic scaling direction of the convolution kernel is forced to align with the movement direction of the construction tool;

[0186] If it is a paving process, the response intensity of the convolution kernel in the joint area is limited according to the paving joint width parameters of the standard 3D BIM model.

[0187] Otherwise, keep the generated convolution kernel parameters unchanged.

[0188] In this embodiment, the local feature vector is the intrinsic theoretical features of the material extracted from the BIM model (such as the tile texture period of 0.1 mm and gloss of 85%, calculated through the dynamic response surface function).

[0189] The actual feature vector is the material characteristics actually collected after construction (such as tile texture period of 0.12mm and gloss of 82%, obtained through image texture analysis and sensor data).

[0190] Global feature difference is the overall difference between two vectors.

[0191] Individual elements under the same element are subdivided features in the same dimension of the feature vector (such as the fundamental frequency, harmonic components, fluctuation amplitude, etc. under the texture period dimension).

[0192] The difference in depth features is a comparison of subdivided features (e.g., theoretical fundamental frequency 10Hz vs. actual fundamental frequency 12Hz, calculated frequency deviation 2Hz; theoretical fluctuation amplitude 0.05mm vs. actual 0.08mm, amplitude deviation 0.03mm).

[0193] Specifically: Global difference calculation: Use Python's numpy.l inalg.norm to calculate the Euclidean distance between vectors, or cosine similarity to measure directional differences.

[0194] In-depth difference mining:

[0195] For each feature dimension (such as texture period), it is decomposed into subbands (fundamental frequency, harmonics) through wavelet transform;

[0196] Calculate the deviation of subband characteristics (e.g., after decomposing scipy.signal.wavelet, compare the mean and variance of each subband);

[0197] The subband biases are summarized to form a depth feature difference vector (e.g., [2Hz, 0.03mm, ...]).

[0198] The fusion vector is a vector that concatenates global feature differences (e.g., 3.001) and deep feature differences (e.g., [2Hz, 0.03mm]) (e.g., [3.001, 2, 0.03], integrating macroscopic and microscopic differences).

[0199] Lightweight Graph Convolutional Decision Networks are simplified Graph Convolutional Networks (GCNs) that reduce parameters and computation (e.g., retaining only 1 layer of graph convolution and 2 layers of fully connected layers to handle the structured relationships of fused vectors).

[0200] Three-dimensional dynamic parameters:

[0201] Asymmetric convolution kernel size: Based on the material texture period (0.1mm) / edge density (5 kernels / mm) = 0.02, a 7×5 asymmetric kernel (major axis along the texture direction) is generated.

[0202] Branch weight allocation coefficients: Environmental interference vector (e.g., humidity 80%RH, higher than the intrinsic 60%RH of BIM, conflict degree 0.6), allocate material texture branch weight 0.3 and noise suppression branch weight 0.7.

[0203] Parameter correlation and constraints: The construction process timestamp (e.g., painting on the 10th day) is correlated with the historical frame (the 8th day, kernel size 5×3), and the correlation is calculated to be 0.8. The current kernel size fluctuation is constrained to be ≤20% (to avoid sudden changes).

[0204] Lightweight GCN implementation:

[0205] Based on PyTorch's torch_geometric library, we construct a 2-layer graph convolution (input dimension 3, output dimension 16) + fully connected layer (output dimension 9, corresponding to 3 sub-parameters of the 3-dimensional parameters);

[0206] Input the fusion vector and output three-dimensional dynamic parameters (e.g., [7,5,0.3,0.7,0.8,0.2], corresponding to size, weight, correlation degree and mutation threshold, respectively).

[0207] The construction process knowledge base is a knowledge graph that stores the rules of the process (e.g., painting → painting process, tool movement direction is vertical; paving → tile laying, joint width is 2mm).

[0208] Painting processes involve applying latex paint to walls, and the tools used (rollers) are mostly moved vertically or horizontally.

[0209] The anisotropic scaling direction is the direction of the long axis of the convolution kernel (e.g., when painting, the long axis is forced to be vertical to match the trajectory of the roller).

[0210] For paving processes such as laying floor tiles, the BIM model presets a joint width of 2mm.

[0211] The response intensity of the seam region is the weight of the convolution kernel at the seam (e.g., limited to 0.5 to reduce misjudgment of seam width deviation).

[0212] Specifically: The knowledge base is constructed using Neo4j to build a knowledge graph, with nodes representing process types (painting, paving) and edges representing constraint rules (direction, seam width);

[0213] Input rules: (Painting {Tool Direction: Vertical}) - [Constraints] -> (Core Direction {Axis: Y}), (Paving {Gap Width: 2mm}) - [Constraints] -> (Gap Response {Weight: 0.5}).

[0214] Parameter verification: Identify the current process (e.g., extract latex paint application date from construction log 2024-06-10 and determine it as a painting process);

[0215] Matching knowledge base rules: If it is a coating, force the core major axis to be along the y-axis (vertical); if it is a pavement, read the BIM joint width (2mm) and set the joint response weight to 0.5;

[0216] Use if-else logic to correct the three-dimensional dynamic parameters (e.g., original kernel size 7×5 → corrected to 7×5, major axis along the y-axis).

[0217] The beneficial effects of the above technical solution are as follows: By combining global and deep feature dual-dimensional analysis, lightweight graph convolution decision-making, and hard constraints of process rules, it solves the problems of traditional dynamic convolution kernels relying solely on data, ignoring construction logic, and experiencing uncontrolled parameter mutations. From the microscopic mining of feature differences to the precise embedding of process knowledge, it achieves deep adaptation of convolution kernel parameters to material intrinsics, environmental interference, and construction procedures, making texture enhancement more suitable for architectural decoration engineering scenarios and providing a dual guarantee of data intelligence and process compliance for high-quality image enhancement.

[0218] This invention provides a building decoration quality analysis system based on BIM technology, wherein the quality analysis module includes:

[0219] The indicator determination unit is used to match the corresponding predefined indicator from the attribute-indicator lookup table according to the unit attributes of each building decoration unit, and regard it as the first indicator;

[0220] The deviation determination unit is used to extract the index deviation by performing feature comparison on the actual building decoration based on the index contents of each first index.

[0221] The set acquisition unit is used to obtain the deviation set based on the combination of deviations of all indicators.

[0222] In this embodiment, the unit attribute is an inherent feature of the building decoration unit, covering dimensions such as type (fixed / non-fixed), material (tile / wood / latex paint), and location (living room wall / bedroom floor). For example, the unit attribute of the kitchen wall is a fixed component, glazed tile, beige, and located 1.2m above the ground.

[0223] The attribute-index lookup table is a pre-established rule library for mapping unit attributes to quality acceptance indicators. It is customized according to national standards and design requirements. For example, in the lookup table, the corresponding indicators for fasteners-glazed tiles-kitchen walls are: flatness deviation ≤ 0.5mm (geometric index); hollow rate ≤ 5% (process index); stain resistance level ≥ 4 (material attribute index).

[0224] The predefined index (first index) is a specific quality acceptance standard matched from the reference table. It is the benchmark for deviation calculation. For example, the first index of the kitchen wall tile unit is [flatness ≤ 0.5mm, hollow rate ≤ 5%, stain resistance ≥ 4].

[0225] The content of each index includes the rules such as the detection method, threshold range, and data format for each predefined index. For example, the flatness deviation includes: detection method: measurement with a 2m straightedge and feeler gauge, and the maximum deviation value is taken; threshold: ≤0.5mm; unit: millimeters (mm).

[0226] Actual architectural decoration refers to the physical finished product of construction (such as the already laid kitchen wall tiles or the installed bedroom wooden doors).

[0227] Feature comparison extraction involves comparing the actual architectural and decorative features (geometric, technical, and material properties) with predefined indicators one by one to extract the differences. For example, when using a 2m straightedge to inspect the wall, the actual maximum flatness deviation is 0.8mm. Compared with the indicator ≤0.5mm, the deviation is extracted as +0.3mm. When tapping with a hollow hammer, three hollow spots are found, accounting for 8%. Compared with the indicator ≤5%, the deviation is extracted as +3%.

[0228] Index deviation is the difference between the actual value and the standard value of a single predefined index, including numerical deviation, percentage deviation, etc. For example, flatness deviation: 0.8mm-0.5mm=+0.3mm (numerical deviation);

[0229] Hollow rate deviation: 8% - 5% = +3% (percentage deviation).

[0230] The deviation set is the collection of all index deviations of the same building decoration unit, reflecting the overall quality defects of the unit. For example, the deviation set of the kitchen wall tile unit is {flatness deviation +0.3mm, hollow rate deviation +3%, stain resistance grade deviation -1 level} (actual stain resistance grade 3, standard ≥ grade 4).

[0231] The beneficial effects of the above technical solution are as follows: By accurately mapping attributes to indicators, quantifying deviations through feature comparison, and aggregating multi-dimensional defects, it solves the problems of arbitrary indicator matching, coarse deviation calculation, and fragmented defect descriptions in traditional quality analysis. With the help of predefined rules and automated comparison, it achieves standardized and traceable construction of deviation sets, providing a precise and structured data foundation for subsequent quality diagnosis (such as deviation propagation analysis and responsibility tracing), enabling building decoration quality analysis to move from experience-based judgment to data-driven analysis.

[0232] This invention provides a building decoration quality analysis system based on BIM technology, wherein the quality analysis module includes:

[0233] The set mapping unit is used to map each deviation set to the building decoration unit corresponding to the standard 3D BIM model, forming a spatial heat map with topological relationships based on the spatial dimension, a deviation process stage based on the time dimension, and a deviation attribute vector based on the attribute dimension.

[0234] The network determines the unit, which is used to construct a knowledge graph with building decoration units as nodes and deviation associations as edges, and uses a graph neural network to determine the deviation propagation path, the responsible party for the deviation, and the probability of the current combination of deviations causing a quality accident.

[0235] The report generation unit is used to obtain the quality profile of the corresponding building decoration unit based on the determined results, obtain the individual quality information, and at the same time, to integrate and analyze all the individual quality information involved in each construction stage throughout the entire construction cycle to obtain the total quality information.

[0236] A quality analysis report is generated based on individual quality information and total quality information.

[0237] In this embodiment, a spatial heatmap with topological relationships is generated along the spatial dimension:

[0238] Topological relationships: Spatial connectivity of building decoration units (e.g., "bedroom wall unit" and "baseboard unit" are adjacent).

[0239] Spatial heatmap: In the BIM model, unit deviations are marked with a color gradient (red → yellow → green corresponding to deviations from large to small), and topologically related units are connected with lines (such as visualization of deviation propagation between adjacent units).

[0240] The deviation process stage in the time dimension refers to the construction process in which the deviation occurs (e.g., "tile hollow deviation" originates from the "laying process"), and is associated with the construction timestamp (e.g., "tile laying completed at 14:00 on June 5, 2024").

[0241] The deviation attribute vector of the attribute dimension is to convert the deviation set into a numerical vector (such as flatness +0.2mm→0.2, gap width -0.5mm→-0.5), and aggregate the multi-index deviation features of the unit. For example, the deviation attribute vector of the bedroom wooden door unit is [0.3 (verticality), -0.2 (gap)].

[0242] In this embodiment, the building decoration unit is a node. Each unit (such as the paint unit on the east wall of the master bedroom) is used as a knowledge graph node, and its attributes include deviation attribute vector, spatial coordinates, etc.

[0243] Deviation correlation is defined as the causal / influence relationship between nodes (e.g., uneven base layer → paint dripping, edge attribute is correlation strength).

[0244] Knowledge graphs: using graph databases (such as Neo4j) to store the relationships between units, deviations, and processes, for example:

[0245] Nodes: (Unit{id:WALL-M-001, Deviation Vector:[0.3,-0.4]}), (Deviation{type:Drop,value:+2mm}), (Process{name:Paint Application,time:2024-06-10});

[0246] Edge: (Unit)-[HAS_DEVIATION]->(Deviation), (Process)-[CAUSES]->(Deviation).

[0247] The propagation path of deviation is to identify the causal chain (such as deviation in base leveling (process A) → paint dripping (deviation B)).

[0248] The party responsible for the deviation is the construction personnel / equipment associated with the deviation (e.g., paint dripping is caused by worker Zhang San (60% responsibility) and the roller equipment (40% responsibility)).

[0249] The probability of quality incidents is based on historical data (e.g., when there is dripping and insufficient thickness, the probability of falling off increases by 50%), and predicts the risk of the current combination of deviations (e.g., when the dripping deviation is 2mm, the probability of falling off is 60%).

[0250] A quality profile is a multi-dimensional quality overview of a single building decoration unit, integrating spatial heat maps, time processes, deviation vectors, and risk predictions. For example, the quality profile of the master bedroom's east wall paint unit includes:

[0251] Space layer: red heat map (fall deviation +2mm) + topologically related ceiling units (indirect deviation, highlighted in yellow);

[0252] Time layer: Paint application process marked in red (high risk);

[0253] Risk layer: 60% probability of detachment, responsible worker Zhang San (60%).

[0254] Single quality information is the quality profile data of a single unit (such as the profile of the unit on the east wall of the master bedroom mentioned above).

[0255] Each construction stage in the entire construction cycle is a time segment of construction (such as "base treatment, main material installation, and final acceptance"), and each stage contains multiple units.

[0256] Overall quality information: Quality aggregation across the entire lifecycle and all units (e.g., an overall pass rate of 80% in the base treatment stage and a 40% risk of detachment during the main material installation stage).

[0257] A quality analysis report is a structured document that integrates individual / total quality information and includes:

[0258] Spatial dimension: Deviation distribution heatmap (BIM model screenshot + topological relationships);

[0259] Time dimension: Process-deviation correlation curve (such as an Echarts line chart, showing the percentage of deviation for each process);

[0260] Risk dimensions: Ranking of accident probability (e.g., detachment risk of 60% ranks first, cracking risk of 20% ranks second);

[0261] Recommendations for improvement: Solutions based on knowledge graph reasoning (such as providing process training to worker Zhang San or replacing the roller equipment);

[0262] Data notarization: Deviation hash of blockchain notarization (ensuring immutability).

[0263] The beneficial effects of the above technical solution are as follows: By mapping and connecting spatial, temporal, and attribute data through BIM multi-dimensional mapping, using knowledge graphs and GNNs to mine correlations and risks, and aggregating comprehensive information through quality profiling, a smart and traceable report is ultimately generated. From precise deviation location to intelligent diagnosis of responsibility and risk, and then to dynamic control of quality throughout the entire lifecycle, a full-chain, intelligent solution is provided for engineering quality management.

[0264] This invention provides a building decoration quality analysis system based on BIM technology, wherein the difference construction unit includes:

[0265] A position locking subunit is used to lock the first position of a single element in a local feature vector and the second position of a single element in the actual feature vector from an element lookup table, wherein the element lookup table is sorted according to a standard step-decay order of the single elements.

[0266] The first determining subunit is used to determine the first stability coefficient of the corresponding individual element based on the first distance between the first position and the second position;

[0267] The second determining subunit is used to analyze the existence variable of each individual element, and to perform individual element elimination and correlation analysis on each actual feature vector in turn to obtain the influence coefficient and existence coefficient of the eliminated individual element, and match them with the double coefficient comparison table to obtain the second stability coefficient.

[0268] The variable mining subunit is used to find the matching variable that has the highest similarity to each existing variable, and to reconstruct a new vector based on all matching variables;

[0269] The defect acquisition subunit is used to analyze the first global comparison result between the local feature vector and the actual feature vector and the second global comparison result between the local feature vector and the new vector to obtain the first defect and the second defect.

[0270] The difference determination subunit is used to obtain deep feature differences based on the first stability coefficient, the second stability coefficient, and in combination with the first defect and the second defect.

[0271] In this embodiment, the element reference table is a pre-constructed standard element decay sequence table, arranged in descending order of the importance of the elements to the quality (e.g., flatness deviation > hollow rate > color uniformity). For example, in the wall tile reference table, the elements are sorted as follows: flatness deviation (weight 0.4), hollow rate (weight 0.3), joint width (weight 0.2), and color difference (weight 0.1).

[0272] A single element is a specific element in the feature vector.

[0273] First position / Second position: The ranking of a single element in the element reference table (e.g., if "flatness" ranks 1st in the reference table, then the first position = 1; if "void rate" ranks 2nd, then the second position = 2).

[0274] The first distance is the difference between the positions of individual elements in the theoretical and actual vectors (e.g., if the first position of "flatness" is 1 and the second position is 1, then the first distance is 0).

[0275] The first stability coefficient is an index reflecting the positional stability of an element, and its formula is S1 = e -∝·|d| , where d is the first distance, and ∝ is the attenuation coefficient, with a value of 0.5.

[0276] The variable is the actual detection value of a single element (e.g., "empty drum rate 5%)".

[0277] Elimination and correlation analysis involves removing one element at a time and then analyzing the impact of the remaining elements on the overall quality. For example, if removing "hollow rate" causes the overall quality score to increase from 75 to 80, it means that the impact coefficient of "hollow rate" is 5.

[0278] The influence coefficient is the change in quality score after removing an element (e.g., the influence coefficient of "hollow rate" = 5).

[0279] The existence coefficient is the degree of deviation between the actual value and the standard value of an element (e.g., "the standard for the hollow rate is ≤3%, while the actual rate is 5%", the existence coefficient = 5% / 3% ≈ 1.67).

[0280] The dual-coefficient lookup table is a predefined mapping table of (influence coefficient, existence coefficient) → stability coefficient. For example, in the lookup table, (influence coefficient = 5, existence coefficient = 1.67) corresponds to a second stability coefficient of 0.7.

[0281] The theoretical variable that has the highest similarity to the existing variable.

[0282] Reconstructing a new vector involves replacing elements in the actual feature vector with matching variables to generate a new vector (e.g., [flatness 0.8mm, hollow rate 5%, seam 1.5mm, color = 1.0]).

[0283] The first global comparison result is the overall difference between the local feature vector and the actual feature vector.

[0284] The second global comparison result is the overall difference between the local feature vector and the new vector.

[0285] First Defect / Second Defect: Quality defects derived from the first / second comparison results respectively (e.g., "First Defect: Large overall deviation; Second Defect: Color close to standard but seam exceeds standard").

[0286] The depth feature difference is the final difference index that combines the stability coefficient and defects, and the formula is: Where S1 and S2 are the quantification values ​​of the first defect and the second defect, and F1 and F2 are the corresponding first stability coefficient and second stability coefficient;

[0287] The beneficial effects of the above technical solution are as follows: by combining standard decay sorting, element position stability analysis, elimination correlation experiments, and variable reconstruction comparison, it breaks through the limitations of traditional feature difference analysis that only focuses on numerical deviations, and deeply explores the structural stability and interactive effects between elements. From the dynamic changes in element position (first stability coefficient) to the impact assessment after elimination (second stability coefficient), and then to the hierarchical diagnosis of defects, it ultimately achieves a deep analysis of quality differences.

[0288] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A building decoration quality analysis system based on BIM technology, characterized in that, include: The collection module is used to collect relevant data on building decoration projects and use BIM modeling software to construct a standard three-dimensional BIM model according to theoretical building dimensions, decoration design requirements, and material parameters. During the modeling process, each building decoration unit is assigned a unique identification code, and the corresponding attribute information of the building decoration unit is entered. The attribute information is related to the building decoration type, the name, model, material, and manufacturer of the spatial components involved. The building decoration type includes the types of fixed spatial components and non-fixed spatial components involved in the corresponding building decoration unit, which can be a single type or a combination of types. The association and binding module is used to collect various data in real time during the construction process based on data acquisition equipment, and to organize the data according to a preset format specification and associate and bind it with the corresponding building decoration unit in the standard three-dimensional BIM model. The quality analysis module is used to analyze the deviation set of the actual building decoration based on the predefined indicators of each building decoration unit, and to annotate the corresponding building decoration unit to obtain individual quality information and total quality information, generate a quality analysis report and output it.

2. The building decoration quality analysis system based on BIM technology according to claim 1, characterized in that, The collection module includes: The structure acquisition unit is used to acquire standard construction drawings of the target construction site from the relevant data, extract the three-dimensional spatial structure of the standard construction drawings, and coarsely divide the three-dimensional spatial structure according to regional characteristics to obtain several coarse spaces. The subspace determination unit is used to find existing diagonal vertices based on the three-dimensional spatial boundary of the coarse space, and to determine the line intersection points and the single-cell subspace enclosed by the diagonal lines and the three-dimensional spatial boundary based on the diagonal lines formed by the diagonal vertices and the distribution of the diagonal lines in the coarse space. The face acquisition unit is used to obtain the maximum cutting face of each single cell subspace and obtain the sub-intersection points of the maximum cutting face and the diagonal; The complexity determination unit is used to determine the spatial complexity of the corresponding coarse space based on the distribution of the diagonals, the distribution of line intersections, the distribution of sub-intersections, and the boundary rules of the three-dimensional space boundary. The coefficient determination unit is used to determine the dependency coefficient of the single cell space based on the boundary rules of the single cell space, the space occupancy of the single cell space, and the space complexity. The radius calculation unit is used to regard single-cell subspaces with a coefficient greater than a preset coefficient as first subspaces, lock the spatial center point of the first subspace, determine the nearest and farthest distances between the spatial center point and the first subspace, the nearest distance between the spatial center point and the maximum cutting surface, and perform average calculations to obtain the initial radius. The spatial division unit is used to obtain the cutting grid radius of the corresponding coarse space based on all initial radii, combined with the corresponding support coefficient and the radius weight determined by the relationship between the maximum cutting surface and the number of diagonal vertices involved in the corresponding single grid unit, and to divide the corresponding coarse space according to the cutting grid radius to obtain the building decoration unit.

3. The building decoration quality analysis system based on BIM technology according to claim 1, characterized in that, The association binding module includes: A matrix construction unit is used to construct a multidimensional data mapping matrix, wherein the row vectors of the matrix correspond to each building decoration unit in the standard three-dimensional BIM model, and the column vectors correspond to preset data dimensions. The data dimensions include at least geometric parameter dimensions, time dimensions, material property dimensions, and process standard dimensions. The mapping unit is used to perform spatiotemporal coordinate calibration on the collected data. Through the absolute coordinate system of the BIM model, the relative coordinate data obtained by different acquisition devices are uniformly mapped to the standard coordinate system.

4. The building decoration quality analysis system based on BIM technology according to claim 3, characterized in that, Also includes: The index library construction module is used to construct a unit-spatiotemporal-viewpoint 3D image index library by selecting key process moments and multi-pose images of the corresponding building decoration unit from the construction data collection library based on the unique identification code of each building decoration unit in the standard 3D BIM model before performing spatiotemporal coordinate calibration on the collected data. The function construction module is used to combine the intrinsic material parameters and construction environment data of the building decoration units in the standard 3D BIM model to construct the dynamic response surface function R(m,e,t)=f(M BIM E sensor ,T time ), and extract local feature vectors, where m is the material parameter vector, e is the environmental parameter vector, t is the time variable, and f(M) is the local feature vector. BIM E sensor ,T time M is a Gaussian process regression function trained using historical construction data; BIM E sensor T time These are the regression parameters based on m, e, and t, respectively. The offset calculation module is used to extract the geometric contours of the spatial components involved in each multi-pose image in the 3D image index library through the edge detection algorithm and compare them with the theoretical contours of the standard 3D BIM model to calculate the contour offset coefficient. The enhancement module is used to enhance the edge clarity and geometric regularity of spatial components in the corresponding pose image based on the theoretical geometric contour and contour offset coefficient of the standard 3D BIM model, and through the contour alignment attention mechanism. The kernel determination module is used to determine the dynamically scaled convolution kernel based on the local feature vector and the actual feature vector generated by the actual environmental data and actual material parameters of the corresponding building decoration unit, so as to separate and enhance the inherent material texture of the spatial components in the corresponding building decoration unit. The enhancement module is used to generate an intermediate enhanced image Tmid = G·Ie + (1-G)·It based on the geometric enhancement result and the texture enhancement result, where G is the fusion gate signal automatically generated based on the contour offset coefficient, Ie is the geometric enhancement result, and It is the texture enhancement result; The index determination module is used to extract the frequency domain features and structural features of the intermediate enhanced image, compare them with the theoretical frequency domain and structural features of the standard 3D BIM model, calculate the enhancement deviation index, and continue to optimize the intermediate enhanced image based on the enhancement deviation index until the enhancement deviation index reaches the index standard.

5. The building decoration quality analysis system based on BIM technology according to claim 4, characterized in that, The convolution kernel determination module includes: The difference construction unit is used to construct the global feature difference between the local feature vector and the actual feature vector. At the same time, it mines and compares the individual elements under the same element of the local feature vector and the actual feature vector to obtain the deep feature difference. The difference fusion unit is used to generate a fusion vector based on the global feature differences and the deep feature differences, and input it into the lightweight graph convolutional decision network to output the three-dimensional dynamic parameters of the convolutional kernel. Among them, the three-dimensional dynamic parameters include: the asymmetric convolution kernel size dynamically generated based on the ratio of material texture period to edge density; the weight allocation coefficients of the material texture branch and noise suppression branch output according to the conflict degree between the environmental interference vector and the intrinsic properties of BIM materials; and the parameter correlation degree and constraint parameter mutation of the current convolution kernel and the historical frame convolution kernel calculated in combination with the timestamp information of the construction process. The verification unit is used to incorporate the construction process knowledge base to perform compliance verification on the dynamically generated convolution kernel parameters. For painting processes, the anisotropic scaling direction of the convolution kernel is forced to align with the movement direction of the construction tool; If it is a paving process, the response intensity of the convolution kernel in the joint area is limited according to the paving joint width parameters of the standard 3D BIM model. Otherwise, keep the generated convolution kernel parameters unchanged.

6. The building decoration quality analysis system based on BIM technology according to claim 1, characterized in that, The quality analysis module includes: The indicator determination unit is used to match the corresponding predefined indicator from the attribute-indicator lookup table according to the unit attributes of each building decoration unit, and regard it as the first indicator; The deviation determination unit is used to extract the index deviation by performing feature comparison on the actual building decoration based on the index contents of each first index. The set acquisition unit is used to obtain the deviation set based on the combination of deviations of all indicators.

7. The building decoration quality analysis system based on BIM technology according to claim 6, characterized in that, The quality analysis module includes: The set mapping unit is used to map each deviation set to the building decoration unit corresponding to the standard 3D BIM model, forming a spatial heat map with topological relationships based on the spatial dimension, a deviation process stage based on the time dimension, and a deviation attribute vector based on the attribute dimension. The network determines the unit, which is used to construct a knowledge graph with building decoration units as nodes and deviation associations as edges, and uses a graph neural network to determine the deviation propagation path, the responsible party for the deviation, and the probability of the current combination of deviations causing a quality accident. The report generation unit is used to obtain the quality profile of the corresponding building decoration unit based on the determined results, obtain the individual quality information, and at the same time, to integrate and analyze all the individual quality information involved in each construction stage throughout the entire construction cycle to obtain the total quality information. A quality analysis report is generated based on individual quality information and total quality information.

8. The building decoration quality analysis system based on BIM technology according to claim 5, characterized in that, The difference construction unit includes: A position locking subunit is used to lock the first position of a single element in a local feature vector and the second position of a single element in the actual feature vector from an element lookup table, wherein the element lookup table is sorted according to a standard step-decay order of the single elements. The first determining subunit is used to determine the first stability coefficient of the corresponding individual element based on the first distance between the first position and the second position; The second determining subunit is used to analyze the existence variable of each individual element, and to perform individual element elimination and correlation analysis on each actual feature vector in turn to obtain the influence coefficient and existence coefficient of the eliminated individual element, and match them with the double coefficient comparison table to obtain the second stability coefficient. The variable mining subunit is used to find the matching variable that has the highest similarity to each existing variable, and to reconstruct a new vector based on all matching variables; The defect acquisition subunit is used to analyze the first global comparison result between the local feature vector and the actual feature vector and the second global comparison result between the local feature vector and the new vector to obtain the first defect and the second defect. The difference determination subunit is used to obtain deep feature differences based on the first stability coefficient, the second stability coefficient, and in combination with the first defect and the second defect.

Images