Steel structure virtual assembly prediction method and system based on point cloud data

Abstract

The invention discloses a point cloud data-based steel structure virtual assembly prediction method and system, and relates to the technical field of steel structure virtual assembly prediction, and the method comprises the steps: collecting point cloud data, carrying out the preprocessing, dividing the point cloud data into subsets, calculating the geometric characteristics of the subsets according to the divided subsets, and carrying out the optimization based on the geometric characteristics, and obtaining an optimal feature vector; performing category classification based on the optimal feature vector, further analyzing component features through a classification result, constructing a path optimization model, and taking the optimal feature vector as input to obtain a final assembled path sequence; according to the method, the particle swarm optimization algorithm is combined with the weighted fusion method, high-precision extraction of the geometric characteristics of the point cloud is achieved, the classification and fitting precision is remarkably improved through covariance matrix characteristic decomposition and linear, plane and spherical surface characteristic construction, and then a path optimization model with the combination of the fire-fisch algorithm and the multi-layer perceptron is utilized, so that the classification and fitting precision is improved. And the error, the length and the stability of the assembly path are comprehensively optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of steel structure virtual assembly prediction, in particular to a method and system for steel structure virtual assembly prediction based on point cloud data. Background Art

[0002] In recent years, with the continuous expansion of the scale of modern buildings and infrastructure projects, steel structures have been widely used in engineering construction due to their advantages such as high strength, easy installation, and cost-effectiveness. The assembly process of steel structures has an important impact on project quality, construction efficiency, and safety. Traditional steel structure assembly relies on manual measurement and empirical judgment, which has problems such as low efficiency, large errors, and insufficient stability. To address these challenges, point cloud data technology has gradually become an important tool in the field of steel structure assembly.

[0003] However, the existing steel structure virtual assembly technology based on point cloud data mostly uses static analysis in the geometric feature extraction method, which is difficult to comprehensively capture the complex features of steel components, resulting in insufficient classification and fitting accuracy, and unable to accurately distinguish linear, planar, and spherical features, thus affecting the subsequent feature classification effect. Secondly, most path optimization algorithms are limited to single-object or simple object combination optimization and lack the ability to find the global optimum. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method and system for steel structure virtual assembly prediction based on point cloud data, which solves the problems that the existing steel structure virtual assembly technology based on point cloud data mostly uses static analysis in the geometric feature extraction method, is difficult to comprehensively capture the complex features of steel components, results in insufficient classification and fitting accuracy, and is unable to accurately distinguish linear, planar, and spherical features, thus affecting the subsequent feature classification effect. Secondly, most path optimization algorithms are limited to single-object or simple object combination optimization and lack the ability to find the global optimum.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, the present invention provides a method for steel structure virtual assembly prediction based on point cloud data, which includes: Collect point cloud data, preprocess it and divide it into subsets, calculate the geometric features of the subsets according to the divided subsets, and optimize based on the geometric features to obtain the optimal feature vector; Perform category classification based on the optimal feature vector, further analyze the component features through the classification results, construct a path optimization model, and take the optimal feature vector as the input to obtain the final assembly path sequence; Visualize the final assembly path sequence through a 3D modeling platform and store it in a database.

[0007] As a preferred solution of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, wherein: the sub - set division after pre - processing the collected point cloud data includes: Collect point cloud data through a three - dimensional laser scanner and record the initial coordinates; Use the statistical filtering method to remove isolated points in the point cloud data and then use the median filter to smooth the data to obtain smooth data; Based on the smooth data, use the density clustering algorithm to divide the smooth data into sub - sets.

[0008] As a preferred solution of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, wherein: calculating the geometric characteristics of the sub - sets according to the divided sub - sets and optimizing based on the geometric characteristics to obtain the optimal feature vector includes: Based on the divided sub - sets, use the covariance matrix formula to calculate the covariance matrix of each sub - set; Use the eigenvalue decomposition technique to decompose the covariance matrix to obtain eigenvalues and corresponding eigenvectors; Define the geometric characteristics of the sub - sets according to the obtained eigenvalues and eigenvectors and then calculate the geometric characteristic scores of each sub - set; According to the geometric characteristic scores, use the weighted fusion method for fusion to obtain feature stability ; Take the corresponding eigenvector as the normal vector and construct a fitting plane; Based on the fitting plane, calculate the mean value of the sum of the squares of the perpendicular distances from each point to the fitting plane to obtain the geometric error : , wherein, represents the number of internal points in the sub - set , represents the point to the distance from the fitting geometric shape; Based on the fitting plane, use the Euclidean distance formula to calculate the Euclidean distance between points; Based on the Euclidean distance, calculate the mean value of the Euclidean distances between all points and then further calculate the feature distribution complexity ; , wherein, represents the Euclidean distance between the internal point in the sub - set and the point ; Integrate based on the geometric error to obtain an error set; Calculate the mean of all errors in the error set using the mean formula, and then further calculate the variance of the errors in combination with the variance formula; Normalize the variance of the errors to obtain the normalized error variance, and define the normalized error variance as the regularization constraint term ; Form a particle population by taking the feature vectors of each subset as particles, and after randomly initializing the particle population, take the particle positions as the initial values of the feature vectors; Define the objective function: , wherein, represents the objective function value of the subset, , , , represents the weight factor, represents normalization; Set the initial velocity to a random value, and then update the individual optimal position and the global optimal position through the particle swarm optimization algorithm. During the update process, calculate the objective function value of each particle. During the iteration process, when the objective function value no longer decreases significantly, stop the iteration, and obtain the optimal feature vector according to the final positions of the particles; The optimal feature vector includes a linear component, a planar component, and a spherical component.

[0009] As a preferred solution of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, wherein: the category classification based on the optimal feature vector includes: Based on the components in the optimal feature vector, calculate the proportion of each component in the total component sum, and after obtaining the importance ratio, define the importance ratio as the weight of each component in the optimal feature vector; According to the weights of each component and the geometric property scores, use the weighted fusion method for fusion to obtain the path score; Set the classification threshold to and , , compare the path score with the threshold. When the path score is greater than or equal to , then determine the optimal feature vector as the linear category. When the path score is less than or equal to , then determine the optimal feature vector as the spherical category. When the path score is less than and greater than , then determine the optimal feature vector as the planar category.

[0010] As a preferred embodiment of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, the steps of further analyzing the component features through the classification results, constructing a path optimization model, and obtaining the final assembly path sequence by using the optimal feature vector as the input include: Based on the classification results, calculate the Euclidean distance between each component in each classification result to obtain the Euclidean distance between components; According to the Euclidean distance between components, calculate the sum of the distances between each component and all components in each classification result, and define the sum of distances as geometric centrality; The sum formula is: , In the formula, represents the geometric centrality of the th component in the current category, represents the total number of components included in the current category, represents the th component, represents the th component, represents the th component, represents the th component, is the index variable of the Based on the geometric centrality, define the component with the minimum geometric centrality as the core node, and further calculate the weight of the core node; The weight calculation formula is: , In the formula, represents the weight of the th core node, represents the weight decay coefficient, represents the total number of components included in the current category, represents the th component, Obtain the position coordinates from the core node by the centroid method; Combine the position coordinates, weights, and initial coordinates to calculate the deviation between the assembly positions: , In the formula, represents the deviation between the assembly positions, represents the initial coordinate of the th component, represents the target position of the th component, Indicates the total number of components included in the current category; Based on the classification results and the core nodes, perform integration, generate a node sequence, calculate the Euclidean distance between adjacent nodes, and accumulate all the Euclidean distances to obtain the total path length ; Obtain the displacement data of each component through a finite element simulation tool for integration. After forming a component set, calculate the displacement standard deviation and mean of each component in the component set; According to the standard and the mean, further calculate the stability of the component set: , In the formula, Indicates the stability of the component set, Indicates the total number of components in the component set, Indicates the th component's displacement standard deviation; Use a multi-layer perceptron and a neural network architecture to construct a path optimization model, including an input layer, a hidden layer, and an output layer; After embedding the firefly algorithm in the neural network, use the prediction results output by the output layer as the first-generation population of the firefly algorithm, and define each firefly individual as an assembly path; Define the loss function: , In the formula, Indicates the loss function value, , , Respectively indicate the deviation, the total path length, and the stability optimization weight; Use the gradient descent method to update and optimize the parameters of the path optimization model and the firefly individuals. During the update and optimization process, gradually calculate the loss function value. During the iteration process, when the loss function value no longer decreases significantly, stop the iteration and output the final path optimization model; Input the optimal feature vector and the classification results into the final path optimization model, and then output the final assembly path sequence of each classification result.

[0011] As a preferred solution of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, wherein: the visualization display of the final assembly path sequence through the three-dimensional modeling platform refers to using different colors for identification based on the final assembly path sequence, with green for installed, yellow for being installed, and gray for to be installed. Use the Blender three-dimensional modeling platform to display the assembly process, and during the assembly process, real-time display the geometric deviation, path length, and stability index.

[0012] As a preferred solution of the steel structure virtual assembly prediction method based on point cloud data according to the present invention, wherein: the storage through the database means storing the display result and the final assembly path sequence as a CSV file, and transmitting each file to the data for storage. After transmission, an encryption operation is performed by the AES encryption method.

[0013] In a second aspect, the present invention provides a steel structure virtual assembly prediction system based on point cloud data, including An acquisition and optimization module, configured to acquire point cloud data, preprocess it, divide it into subsets, calculate the geometric characteristics of the subsets according to the divided subsets, and optimize based on the geometric characteristics to obtain the optimal feature vector; A classification and path optimization module, configured to perform category classification according to the optimal feature vector, further analyze the component characteristics through the classification result, construct a path optimization model, and obtain the final assembly path sequence after taking the optimal feature vector as the input; A display and storage module, configured to visually display the final assembly path sequence through a three-dimensional modeling platform and store it through a database.

[0014] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and wherein: when the computer program is executed by the processor, any step of the steel structure virtual assembly prediction method based on point cloud data as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and wherein: when the computer program is executed by the processor, any step of the steel structure virtual assembly prediction method based on point cloud data as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: The present invention realizes the high-precision extraction of the geometric characteristics of point clouds through the particle swarm optimization algorithm combined with the weighted fusion method, and constructs linear, planar, and spherical features through the eigenvalue decomposition of the covariance matrix, significantly improving the classification and fitting accuracy. Secondly, a path optimization model combining the firefly algorithm and the multi-layer perceptron is used to comprehensively optimize the error, length, and stability of the assembly path, and finally output the optimized assembly path sequence. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1Flowchart of the steel structure virtual assembly prediction method based on point cloud data in Embodiment 1.

[0019] Figure 2 Structure diagram of the steel structure virtual assembly prediction system based on point cloud data in Embodiment 1. Detailed implementation manners

[0020] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention with reference to the accompanying drawings of the specification.

[0021] In the following description, many specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0022] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0023] Embodiment 1, referring to Figure 1 and Figure 2 , which is the first embodiment of the present invention. This embodiment provides a steel structure virtual assembly prediction method based on point cloud data, including the following steps: S1. After collecting point cloud data and preprocessing it, divide it into subsets, calculate the geometric characteristics of the subsets according to the divided subsets, and optimize based on the geometric characteristics to obtain the optimal feature vector; Specifically, the process of collecting point cloud data and preprocessing it to divide it into subsets includes: Collect point cloud data through a 3D laser scanner and record the initial coordinates; Use the statistical filtering method to remove the isolated points in the point cloud data and then use the median filter to smooth the data to obtain the smoothed data; Based on the smoothed data, use the density clustering algorithm to divide the smoothed data into subsets.

[0024] Through the preprocessing method combining 3D laser scanning technology, statistical filtering, and median filtering, the quality and geometric accuracy of the point cloud data have been significantly improved, laying a solid foundation for subsequent processing. The use of statistical filtering to remove isolated points and median filtering to retain geometric characteristics ensures data smoothness and boundary integrity. The introduction of the density clustering algorithm realizes the dynamic partitioning of the point cloud data, enabling the accurate partitioning of the data into subsets according to the density characteristics of the point cloud, so that each subset corresponds to a steel structure component, thereby significantly reducing the subsequent computational complexity.

[0025] Furthermore, calculate the geometric characteristics of the subsets based on the partitioned subsets and optimize them based on the geometric characteristics. The optimal feature vectors obtained include: Based on the partitioned subsets, use the covariance matrix formula to calculate the covariance matrix of each subset; Use eigenvalue decomposition technology to decompose the covariance matrix to obtain eigenvalues and corresponding eigenvectors; Among them, the eigenvalues should satisfy , represents the linear eigenvalue, represents the planar eigenvalue, represents the spherical eigenvalue. The linear characteristic is determined by the difference ratio of the eigenvalues , reflecting the main axis distribution. The planar characteristic is determined by the difference ratio of the eigenvalues , reflecting the planar distribution. The spherical characteristic is determined by the ratio of the eigenvalues and , reflecting the uniformity; After defining the geometric characteristics of the subsets based on the obtained eigenvalues and eigenvectors, calculate the geometric characteristic scores of each subset; The formula for the linear characteristic score in geometric characteristics is: , In the formula, represents the linear characteristic score, and represent the eigenvalues; The formula for the planar characteristic score in geometric characteristics is: , In the formula, represents the planar characteristic score, represents the eigenvalue; The formula for the spherical characteristic score in geometric characteristics is: , In the formula, represents the spherical characteristic score; According to the geometric characteristic scores, use the weighted fusion method for fusion to obtain the feature stability ; Use the corresponding eigenvector as the normal vector and construct a fitting plane; Based on the fitting plane, calculate the mean of the sum of the squares of the perpendicular distances from each point to the fitting plane to obtain the geometric error : , wherein, represents the number of inliers in the subset , represents the point to the distance from the fitting geometry; Based on the fitting plane, use the Euclidean distance formula to calculate the Euclidean distance between points; Based on the Euclidean distance, calculate the mean of the Euclidean distances between all points and further calculate the feature distribution complexity ; , wherein, represents the Euclidean distance between the inlier in the subset and the point ; Integrate based on the geometric error to obtain an error set; Use the mean formula to calculate the mean of all errors in the error set and then combine with the variance formula to further calculate the variance of the errors; Normalize the variance of the errors to obtain the normalized error variance, and define the normalized error variance as the regularization constraint term ; Form a particle population by using the eigenvectors of each subset as particles, randomly initialize the particle population, and use the particle positions as the initial values of the eigenvectors; The initial eigenvectors of the particles include linear components, planar components, and spherical components; Define the objective function: , wherein, represents the objective function value of the subset, , , , represents the weight factor, which can be set by experience and experiments, represents normalization; Set the initial velocity to a random value and then update the individual optimal position and the global optimal position through the particle swarm optimization algorithm. During the update process, calculate the objective function value of each particle. During the iteration process, when the objective function value no longer decreases significantly, stop the iteration and obtain the optimal eigenvector according to the final positions of the particles; The optimal eigenvector contains linear components, planar components, and spherical components.

[0026] By combining covariance matrix calculation and eigen-decomposition techniques, the spatial distribution characteristics of point cloud data can be accurately captured, and linear, planar, and spherical characteristics can be extracted. Through a weighted fusion method, the characteristic components are dynamically optimized, which can effectively cope with the diversity of the forms of complex steel members, and at the same time greatly improve the accuracy of characteristic classification, providing a scientific basis for the accurate identification of steel beams, steel columns, and connection nodes. By constructing an objective function based on geometric error, characteristic complexity, and regularization constraints, and introducing a particle swarm optimization algorithm to achieve the global optimum of the eigenvector, the particle swarm optimization algorithm significantly reduces the geometric error by dynamically adjusting the individual and global optimum positions, and ensures that the eigenvector has higher stability and reliability, which can effectively cope with the possible noise and abnormal distribution in the point cloud data, improving the accuracy of characteristic fitting and the robustness of the model.

[0027] S2. Perform category classification based on the optimal eigenvector, further analyze the component characteristics through the classification results, construct a path optimization model, and obtain the final assembly path sequence after using the optimal eigenvector as the input; Specifically, performing category classification based on the optimal eigenvector includes: Based on the components within the optimal eigenvector, calculate the proportion of each component in the total sum of components, and after obtaining the importance ratio, define the importance ratio as the weight of each component within the optimal eigenvector; According to the weight of each component and the geometric characteristic score, use the weighted fusion method for fusion to obtain the path score; Set the classification threshold as and , , compare the path score with the threshold. When the path score is greater than or equal to , then determine the optimal eigenvector as the linear category. When the path score is less than or equal to , then determine the optimal eigenvector as the spherical category. When the path score is less than and greater than , then determine the optimal eigenvector as the planar category. The linear category refers to the assembled steel beam, the spherical category refers to the assembled node, and the planar category refers to the assembled steel column.

[0028] Through the dynamic weight allocation method, the weights are adaptively adjusted according to the characteristics of the point cloud data, avoiding the irrationality of artificially setting weights, improving the accuracy of feature component evaluation, and through the weighted fusion method, the linear, planar, and spherical components of the geometric characteristics are effectively integrated, comprehensively considering the relative importance of each feature, significantly enhancing the adaptability of the present invention to the diversity and complexity of the point cloud data. Secondly, through the classification threshold set by domain knowledge and personal experience, the path score is clearly corresponded to the component category, the classification rule is clear, and the judgment result is accurate, ensuring the accurate identification of steel beams, steel columns, and connection nodes.

[0029] Furthermore, by further analyzing the component features based on the classification results, a path optimization model is constructed, and the optimal feature vector is used as the input to obtain the final assembly path sequence, including: Based on the classification results, the Euclidean distance is used to calculate the distance between each component in each classification result, obtaining the Euclidean distance between the components. According to the Euclidean distance between the components, the sum of the distances between each component and all components in each classification result is calculated, and the sum of the distances is defined as the geometric centrality. The sum formula is: , In the formula, represents the geometric centrality of the th component in the current category, represents the total number of components included in the current category, represents the th component and the th component, represents the th component in the current category, represents the th component in the current category, represents the th index variable of the component; Based on the geometric centrality, the component with the minimum geometric centrality is defined as the core node, and then the weight of the core node is further calculated. The core node represents the geometric center of the classification result and the key point of the classification shape. The weight calculation formula is: , In the formula, represents the weight of the th core node, represents the weight decay coefficient, represents the total number of components included in the current category, represents the th index variable of the component; Obtain the position coordinates from the core nodes through the centroid method; Combine the position coordinates, weights, and initial coordinates to calculate the deviation between the assembly positions: , In the formula, represents the deviation between the assembly positions, represents the initial coordinate of the th component, represents the target position of the th component, represents the total number of components included in the current category; Integrate based on the classification results and core nodes, generate a node sequence, calculate the Euclidean distance between adjacent nodes, and accumulate all the Euclidean distances to obtain the total path length ; The content of the nodes in the node sequence includes the core nodes in the classification results, the representative nodes corresponding to the optimal feature vectors, and the meaning represented by the nodes is the representativeness of the geometric center; Obtain the displacement data of each component through a finite element simulation tool for integration. After forming a component set, calculate the displacement standard deviation and mean of each component in the component set; According to the standard and mean, further calculate the stability of the component set: , In the formula, represents the stability of the component set, represents the total number of components in the component set, represents the th component's displacement standard deviation; Use a multi-layer perceptron and neural network architecture to construct a path optimization model, including an input layer, a hidden layer, and an output layer; After embedding the firefly algorithm in the neural network, take the prediction result output by the output layer as the first-generation population of the firefly algorithm, and define each firefly individual as an assembly path; After embedding the firefly algorithm in the neural network, the prediction result output by the output layer is an initial solution, which is used to represent the initial state of the assembly path. The initial solution is the node sequence in the assembly path, which can be directly used as the initial population of the firefly algorithm. And the assembly path is represented by the firefly algorithm individuals, which can ensure the population diversity; Define the loss function: , In the formula, represents the loss function value, , , respectively represent the deviation, total path length, and stability optimization weight, which can be set based on expert experience; Use the gradient descent method to update and optimize the parameters of the path optimization model and the firefly individuals. During the update and optimization process, gradually calculate the loss function value. During the iteration process, when the loss function value no longer decreases significantly, stop the iteration and output the final path optimization model; After inputting the optimal feature vector and classification results into the final path optimization model, output the final assembled path sequence for each classification result.

[0030] Through Euclidean distance and geometric centrality calculations, the spatial distribution characteristics of each component during the assembly process can be quantified. Geometric centrality calculates the sum of the distances between each component and other components to accurately identify the geometric center position, and defines the component with the minimum geometric centrality as the core node, thereby effectively reducing path redundancy and unnecessary assembly steps, laying a solid foundation for the optimization of the assembly path. Moreover, by introducing the weight of the core node, the present invention assigns dynamically adjustable priorities to different nodes. The weight of the core node is dynamically calculated based on geometric centrality and component distribution to ensure that important nodes are given priority in path planning. By using 3D modeling software to obtain the spatial coordinates of the core node, and combining the weight and the target position, dynamically adjust the deviation of the assembly position. Through formulaic geometric deviation calculations, while reducing geometric errors, it ensures the precise alignment of the assembled components, improves the stability of the assembly process, and uses a multi-layer perceptron to construct a path optimization model and embeds a firefly algorithm to enhance the optimization performance. The multi-layer perceptron realizes efficient modeling of path planning by learning the characteristics of non-linear components, and the firefly algorithm dynamically adjusts the objective function value of the assembly path by simulating swarm optimization behavior and quickly converges to the global optimal path sequence, which significantly improves the path planning efficiency of the present invention, reduces the optimization search time, and at the same time ensures the overall performance of the assembly path. Secondly, the introduction of a finite element simulation tool effectively guarantees the stability of the assembly path. By calculating the displacement standard deviation and mean of the component set, evaluate the anti-disturbance ability of the path in a complex engineering environment to ensure that the final assembled path still has high stability and reliability under various constraint conditions.

[0031] S3. Visualize and display the final assembled path sequence through a 3D modeling platform and store it in a database; Specifically, visualizing and displaying the final assembled path sequence through a 3D modeling platform means using different colors for identification based on the final assembled path sequence. Green represents installed, yellow represents being installed, and gray represents to be installed. Use the Blender 3D modeling platform to display the assembly process, and during the assembly process, real-time display geometric deviation, path length, and stability indicators.

[0032] The assembly path sequence is visually displayed through a 3D modeling platform, combined with dynamic updates of color identification. Green indicates installed, yellow indicates being installed, and gray indicates to be installed, presenting the assembly progress clearly and intuitively, facilitating construction management. Moreover, the real-time display of geometric deviations can dynamically evaluate the assembly accuracy, quickly detect and correct errors, significantly reducing the risk of construction rework. The real-time monitoring of the path length optimizes the movement trajectory of construction equipment and resource allocation, improving construction efficiency. The dynamic display of stability indicators monitors the structural stiffness and seismic performance during the assembly process in real time, enhancing construction safety and overall reliability.

[0033] Furthermore, storing through the database means storing the display results and the final assembly path sequence as CSV files, and transmitting each file to the database for storage. After transmission, encryption operations are performed using the AES encryption method.

[0034] By storing the assembly path sequence and display results as CSV files and performing AES encryption after transmitting them to the database, the data management, transmission, and security performance are significantly improved. Using CSV file storage has the advantages of strong universality, high portability, and convenience for analysis, and is suitable for rapid data interaction between the engineering site and the background. Through centralized storage in the database, efficient data management, multi-user access, and associated storage of the assembly path and display results are achieved, enhancing the intuitiveness of the visual display and the traceability of the assembly process. Secondly, the AES encryption technology provides strong security protection for data transmission and storage, preventing data leakage or tampering.

[0035] This embodiment also provides a steel structure virtual assembly prediction system based on point cloud data, including: A collection and optimization module for collecting point cloud data, preprocessing it, dividing it into subsets, calculating the geometric characteristics of the subsets according to the divided subsets, and optimizing based on the geometric characteristics to obtain the optimal feature vector; A classification and path optimization module for classifying categories according to the optimal feature vector, further analyzing the component characteristics through the classification results, constructing a path optimization model, and obtaining the final assembly path sequence after taking the optimal feature vector as the input; A display and storage module for visually displaying the final assembly path sequence through a 3D modeling platform and storing it through the database.

[0036] This embodiment also provides a computer device applicable to the case of the steel structure virtual assembly prediction method based on point cloud data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the steel structure virtual assembly prediction method based on point cloud data proposed in the above embodiment.

[0037] The computer device may be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0038] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for predicting virtual assembly of a steel structure based on point cloud data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), Erasable Programmable Read-Only Memory (EPROM for short), Programmable Read-Only Memory (PROM for short), Read-Only Memory (ROM for short), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0039] In summary, the present invention combines the particle swarm optimization algorithm with the weighted fusion method to achieve high-precision extraction of the geometric characteristics of point clouds, and constructs linear, planar, and spherical features through covariance matrix eigenvalue decomposition, significantly improving the classification and fitting accuracy. Secondly, a path optimization model combining the firefly algorithm and a multi-layer perceptron is used to comprehensively optimize the error, length, and stability of the assembly path, and finally outputs an optimized assembly path sequence.

[0040] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A steel structure virtual assembly prediction method based on point cloud data, characterized in that: Including, Collecting point cloud data, preprocessing it, dividing it into subsets, calculating the geometric characteristics of each subset based on the division, and optimizing based on the geometric characteristics to obtain the optimal feature vector; Performing category classification based on the optimal feature vector, further analyzing the component features through the classification results, constructing a path optimization model, and taking the optimal feature vector as the input to obtain the final assembly path sequence; Visualizing the final assembly path sequence through a 3D modeling platform and storing it in a database.

2. The virtual assembly prediction method for steel structures based on point cloud data according to claim 1, wherein: The step of dividing the subsets after preprocessing the collected point cloud data includes: Collecting point cloud data with a 3D laser scanner and recording the initial coordinates; Using the statistical filtering method to remove the isolated points in the point cloud data and then using the median filter to smooth the data to obtain the smoothed data; Based on the smoothed data, the density clustering algorithm is used to divide the smoothed data into subsets.

3. The virtual assembly prediction method for steel structures based on point cloud data according to claim 2, wherein: The step of calculating the geometric characteristics of each subset based on the division, optimizing based on the geometric characteristics, and obtaining the optimal feature vector includes: Based on the divided subsets, using the covariance matrix formula to calculate the covariance matrix of each subset; Using the eigenvalue decomposition technique to decompose the covariance matrix to obtain the eigenvalues and the corresponding eigenvectors; Defining the geometric characteristics of each subset according to the obtained eigenvalues and eigenvectors and then calculating the geometric characteristic scores of each subset; According to the geometric characteristic scores, a weighted fusion method is used for fusion to obtain the feature stability ; Taking the corresponding eigenvector as the normal vector and constructing a fitting plane; Based on the fitted plane, calculate the mean value of the sum of the squares of the perpendicular distances from each point to the fitted plane to obtain the geometric error : ， wherein, represents the number of interior points in the subset and represents the distance from the point to the fitted geometry; Based on the fitting plane, using the Euclidean distance formula to calculate the Euclidean distance between points; Based on the Euclidean distance, calculate the mean of the Euclidean distances between all pairs of points and then further calculate the complexity of the feature distribution ; ， wherein, represents the subset interior point and the point the Euclidean distance between; Integrating based on the geometric error to obtain an error set; Using the mean formula to calculate the mean of all errors in the error set and then further calculating the variance of the errors in combination with the variance formula; The variance of the error is standardized to obtain the standardized error variance, and the standardized error variance is defined as the regularization constraint term ; Taking the feature vector of each subset as a particle to form a particle population, randomly initializing the particle population, and taking the particle position as the initial value of the feature vector; Defining the objective function: ， In the formula, represents the objective function value of the subset, , , , represents the weight factor, represents normalization; Setting the initial velocity as a random value and then updating the individual optimal position and the global optimal position through the particle swarm optimization algorithm. During the update process, calculating the objective function value of each particle. During the iteration process, when the objective function value no longer decreases significantly, stop the iteration, and obtain the optimal feature vector according to the final position of the particle; The optimal feature vector contains linear components, planar components, and spherical components.

4. The method for predicting virtual assembly of steel structures based on point cloud data according to claim 3, wherein: The step of performing category classification based on the optimal feature vector includes: Based on the components in the optimal feature vector, calculating the proportion of each component in the total component sum, obtaining the importance ratio, and defining the importance ratio as the weight of each component in the optimal feature vector; Using the weighted fusion method to fuse according to the weight of each component and the geometric characteristic score to obtain the path score; Set the classification threshold to and , , compare the path score with the threshold. When the path score is greater than or equal to , then determine the optimal feature vector as a linear class. When the path score is less than or equal to , then determine the optimal feature vector as a spherical class. When the path score is less than and greater than , then determine the optimal feature vector as a planar class.

5. The method for predicting virtual assembly of steel structures based on point cloud data according to claim 4, characterized in that: The step of further analyzing the component features through the classification results, constructing a path optimization model, and taking the optimal feature vector as the input to obtain the final assembly path sequence includes: Based on the classification results, using the Euclidean distance to calculate the distance between each component in each classification result to obtain the Euclidean distance between components; According to the Euclidean distance between components, calculating the sum of the distances between each component and all components in each classification result and then defining the sum of the distances as the geometric centrality; The sum formula is: ， Wherein, represents the geometric centrality of the -th component in the current category, represents the total number of components included in the current category, represents the Euclidean distance between the -th component and the -th component, represents the -th component in the current category, represents the -th component in the current category, represents the -th component's index variable; Based on geometric centrality, the component with the minimum geometric centrality is defined as the core node, and then the weight of the core node is further calculated; The weight calculation formula is as follows: ， In the formula, represents the weight of the th core node, represents the weight decay coefficient, represents the total number of components included in the current category, represents the th index variable of the component; The position coordinates are obtained from the core nodes by the centroid method; The position coordinates, weights, and initial coordinates are combined to calculate the deviation between the assembly positions: ， In the formula, represents the deviation between the assembly positions, represents the initial coordinate of the th component, represents the target position of the th component, represents the total number of components included in the current category; Integrate based on the classification results and core nodes, generate a node sequence, calculate the Euclidean distance between adjacent nodes, and accumulate all the Euclidean distances to obtain the total path length ; The displacement data of each component are obtained through a finite element simulation tool and integrated. After forming a component set, the displacement standard deviation and mean value of each component in the component set are calculated; According to the standard and mean value, the stability of the component set is further calculated; ， In the formula, represents the stability of the component set, represents the total number of components in the component set, represents the displacement standard deviation of the A path optimization model is constructed using a multi-layer perceptron and a neural network architecture, including an input layer, a hidden layer, and an output layer; After embedding the firefly algorithm in the neural network, the prediction results output by the output layer are used as the first-generation population of the firefly algorithm, and each firefly individual is defined as an assembly path; Define the loss function: ， In the formula, represents the loss function value, , , respectively represent the deviation, the total path length, and the stability optimization weight; The gradient descent method is used to update and optimize the parameters of the path optimization model and the firefly individuals. During the update and optimization process, the loss function value is gradually calculated. During the iteration process, when the loss function value no longer decreases significantly, the iteration is stopped, and the final path optimization model is output; The optimal feature vector and classification results are input into the final path optimization model, and then the final assembly path sequence of each classification result is output.

6. The method for predicting virtual assembly of steel structures based on point cloud data according to claim 5, wherein: The visualization display of the final assembly path sequence through the 3D modeling platform means that different colors are used for identification based on the final assembly path sequence. Green indicates installed, yellow indicates being installed, and gray indicates to be installed. The Blender 3D modeling platform is used to display the assembly process, and during the assembly process, geometric deviation, path length, and stability indicators are displayed in real time.

7. The method for predicting virtual assembly of steel structures based on point cloud data according to claim 6, wherein: The storage through the database means that the display results and the final assembly path sequence are stored as CSV files, and each file is transmitted to the data for storage. After transmission, encryption operations are performed through the AES encryption method.

8. A steel structure virtual assembly prediction system based on point cloud data, based on the steel structure virtual assembly prediction method based on point cloud data according to any one of claims 1 to 7, characterized in that: Including, The acquisition and optimization module is used to acquire point cloud data, perform preprocessing, divide subsets, calculate the geometric characteristics of the subsets according to the divided subsets, and optimize based on the geometric characteristics to obtain the optimal feature vector; The classification and path optimization module is used to perform category classification according to the optimal feature vector, further analyze the component characteristics through the classification results, construct a path optimization model, and obtain the final assembly path sequence with the optimal feature vector as the input; The display and storage module is used to perform visual display of the final assembly path sequence through the 3D modeling platform and store it through the database.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the steel structure virtual assembly prediction method based on point cloud data according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the steel structure virtual assembly prediction method based on point cloud data according to any one of claims 1 to 7.

Images