CN117743955A

CN117743955A - BIM (building information modeling) acquired data processing method, system, electronic equipment and storage medium

Info

Publication number: CN117743955A
Application number: CN202311768670.3A
Authority: CN
Inventors: 张思中; 古习斌; 曹永雄
Original assignee: Guangdong Renxin Engineering Cost Consulting Co ltd
Current assignee: Guangdong Renxin Engineering Cost Consulting Co ltd
Priority date: 2023-12-21
Filing date: 2023-12-21
Publication date: 2024-03-22

Abstract

The application provides a BIM (building information modeling) acquired data processing method, system, electronic equipment and storage medium, and relates to the technical field of data processing. The method comprises the steps of determining the type of a target building, collecting historical data of the type of the building to construct a vector machine data set, presetting a scale interval, obtaining all scale parameters in the interval, determining a corresponding training set according to each scale parameter, intelligently determining a kernel function of the corresponding parameter according to the data type of each training set, determining an initial value of each scale parameter according to the kernel function and the training set, and calculating the target scale parameter according to the initial value. And the target scale parameters can obtain the classification accuracy through cross verification, and finally, the target interval is determined according to the initial value and the classification accuracy so as to lock the optimal parameter range. The problem that the accuracy of the BIM model is affected due to poor hyperplane construction effect and classification accuracy caused by improper selection of SVM kernel functions and scale parameters of the kernel functions is solved.

Description

BIM (building information modeling) acquired data processing method, system, electronic equipment and storage medium

Technical Field

The present application relates to the field of data processing, and in particular, to a method, a system, an electronic device, and a storage medium for processing acquired data of a BIM.

Background

In the digital transformation and intelligent development of the building industry, the data acquisition and intelligent processing technology of the BIM model is becoming an important field. BIM realizes project cooperation by creating a three-dimensional digital building model and integrating information such as geometry, attribute, material, cost and the like. However, errors in the data acquisition process can lead to reduced model accuracy. To ensure that the model data is reliable, anomaly detection and classification of the acquired data is required.

At present, an SVM support vector machine is commonly used for carrying out abnormal classification. But improper parameter selection of the SVM results in performance degradation. The improper selection of the scale parameters of the kernel function can cause poor hyperplane construction effect, reduce classification accuracy and further influence BIM model accuracy.

Disclosure of Invention

The application provides a method, a system, electronic equipment and a storage medium for processing acquired data of BIM, which are used for solving the problem that performance is reduced due to improper parameter selection of SVM. The problem that the accuracy of the BIM model is affected by poor hyperplane construction effect and classification accuracy rate caused by improper selection of scale parameters of the kernel function is solved.

In a first aspect, the present application provides a method for processing acquired data of a BIM, where the method includes:

determining the building type of a target building, collecting historical building data of the same building type, and constructing a vector machine data set based on the historical building data;

acquiring all scale parameters in a preset scale interval, and determining a training set corresponding to each scale parameter according to each scale parameter and a vector machine data set;

determining the type of building data according to building data in the training set corresponding to each scale parameter, and determining a kernel function corresponding to each scale parameter according to the type of building data;

determining an initial value of each scale parameter according to a kernel function corresponding to each scale parameter and a training set corresponding to each scale parameter, and determining a target scale parameter corresponding to each scale parameter according to the initial value of each scale parameter;

cross-verifying all the target scale parameters to obtain the classification accuracy of each target scale parameter;

determining a target interval according to initial values and classification accuracy of all target scale parameters in the scale interval, and determining optimal scale parameters of a target building according to the target interval;

And constructing a target SVM hyperplane model of the target building type according to the optimal scale parameters, and processing building data to be classified according to the target SVM hyperplane model to obtain building data after separation processing.

By adopting the technical scheme, firstly, the type of the target building is determined, and the historical data of the type of building is collected to construct a vector machine data set, so that a data basis is provided for the training of a subsequent SVM model. Then, the method presets a scale interval, acquires all scale parameters in the interval, and determines a corresponding training set according to each scale parameter, so that the step traverses the scale parameter space, and conditions are created for searching the optimal scale parameters. Then, the method intelligently determines the kernel function of the corresponding parameter according to the data type of each training set, and combines data driving and a rule of thumb, so that the kernel function is selected more specifically. After the kernel function is determined, the method determines the initial value of each scale parameter according to the kernel function and the training set, and calculates the target scale parameter according to the initial value. The target scale parameters can obtain the classification accuracy through cross verification, and then the target interval is determined according to the initial value and the classification accuracy so as to lock the optimal parameter range. And finally, determining final scale parameters by using the target interval, and constructing an SVM model for classification. In this way, the optimal parameters are determined through the steps of data set construction, parameter traversal, kernel function determination, cross verification and the like, so that the BIM acquisition data processing effect is improved, and the problem that the performance is reduced due to improper parameter selection of the SVM is solved. The problem that the accuracy of the BIM model is affected by poor hyperplane construction effect and classification accuracy rate caused by improper selection of scale parameters of the kernel function is solved.

Optionally, determining the type of the building data according to the building data in the training set corresponding to each scale parameter, and determining the kernel function corresponding to each scale parameter according to the type of the building data, including:

classifying building data in the training set corresponding to each scale parameter, and determining the dimension type and the data distribution type of the building data;

when the type of the building data is a low-dimensional type or a linear distribution type, determining a kernel function corresponding to the scale parameter as a linear kernel function;

when the type of the building data is a high-dimensional type or a nonlinear distribution type, determining a kernel function corresponding to the scale parameter as a Gaussian kernel function.

By adopting the technical scheme, building data classification is firstly carried out on the training set corresponding to each scale parameter, and two characteristics of the dimension type and the distribution type of the data are determined. Then, it is determined which kernel function to use based on these two features: if the building data is of a low-dimensional type or a linear distribution type, selecting a simple linear kernel function due to low data complexity; if the building data is of a high-dimensional type or a non-linear distribution type, a strong gaussian kernel function is selected due to the high complexity of the data. In this way, the method enables the selection of a kernel function of appropriate complexity according to the specific data situation. Compared with the fixed use of a certain kernel function, the method can enable the expression capacity of the kernel function to be matched with the complexity of data, and prevent overestimation or underestimation, so that the classification effect of the finally obtained SVM model is better.

Optionally, determining an initial value of each scale parameter according to the kernel function corresponding to each scale parameter and the training set corresponding to each scale parameter includes:

taking building data corresponding to each scale parameter as input of a corresponding kernel function to obtain a kernel function matrix corresponding to each scale parameter;

determining a kernel function value sequence of each historical building data in the training set corresponding to each scale parameter according to the corresponding kernel function matrix;

and determining initial values of corresponding scale parameters according to the kernel function value sequence.

By adopting the technical scheme, building data is used as input of a corresponding kernel function, and a kernel function matrix is constructed. And then calculating the kernel function value between each historical building data and other data based on the matrix to form a kernel function value sequence. And finally, determining the initial value of the scale parameter according to the kernel function value sequence. In this way, the method does not simply and randomly or empirically set an initial value, but fully considers the inherent correlation among building data by calculating the kernel function mapping feature space distance, and initializes the scale parameters conforming to the data distribution. Compared with other initial value determining modes, the data driving mode based on the kernel function matrix has better interpretability, and the initial value of the parameter is closer to the optimal value in the search space.

Optionally, acquiring the kernel function value sequence of each building data in the building data of the target scale parameter includes:

and acquiring kernel function values of the nth building data and each building data except the nth building data in the Gaussian kernel function matrix in the building data of the target scale parameter, and forming a kernel function value sequence of the nth building data.

By adopting the technical scheme, the nth sample is taken out in sequence from the sample set corresponding to the target scale parameter. And then extracting the kernel function values between the nth sample and all other samples except the nth sample from the Gaussian kernel function matrix of the scale parameter. These kernel function values constitute the sequence of kernel function values for the nth sample at the current scale parameter.

The mode of acquiring the sequence one by one can comprehensively investigate the effect of the target scale parameter on the feature extraction of each sample. And (3) evaluating the distribution adaptation effect of the scale parameters after the samples are mapped to the feature space by analyzing the matching degree of the samples in the kernel matrix. The calculation sequence can quantify the adaptation degree, and provides basis for the subsequent determination of the initial value of the scale parameter.

Optionally, determining the target interval according to the initial values and the classification accuracy of all target scale parameters in the scale interval includes:

According to the initial values and the classification accuracy of all target scale parameters in the scale interval, determining a correlation coefficient between the initial values and the classification accuracy;

taking a target scale parameter corresponding to the maximum classification accuracy in the scale interval as a center parameter of the scale interval;

dividing the scale interval into a first interval and a second interval according to the central parameter, and determining the classification accuracy of the first interval and the second interval;

acquiring all scale parameters in a first interval and a second interval and initial values of each scale parameter;

determining a first average initial value of the first section according to the initial values of all scale parameters of the first section, and determining a second average initial value of the second section according to the initial values of all scale parameters of the second section;

determining the target degree of the first interval according to the classification accuracy of the first average initial value and the first interval and the correlation coefficient between the initial value and the classification accuracy;

determining the target of the second interval according to the classification accuracy of the second average initial value and the second interval and the correlation coefficient between the initial value and the classification accuracy;

and taking the second interval and the interval with the largest target scale in the second interval as target intervals.

By adopting the technical means, the correlation coefficient of the initial value and the classification accuracy is calculated first, and the correlation between the initial value and the classification accuracy is determined. And then dividing the scale interval into two parts by taking the parameter corresponding to the maximum classification accuracy as a central parameter, and calculating the classification accuracy of each interval. And then acquiring scale parameters and initial value information in each section, and calculating the average initial value of each section. And determining the target scale of each section according to the average initial value, the classification accuracy and the correlation coefficient of the two, and comparing the target scales to determine a new target section. The method gradually reduces the selection range of the scale parameters through the iterative process of interval division and target degree comparison, so that the scale parameters approach to the optimization target. Compared with global scanning, the local iterative search mode improves the efficiency of parameter optimization. In conclusion, the target interval determination thought realizes more accurate and efficient scale parameter selection, and the performance of the final BIM acquisition data processing is enhanced.

Optionally, dividing the scale interval into a first interval and a second interval according to the central parameter, and determining the classification accuracy of the first interval and the second interval, including:

taking the target scale parameter which is smaller than the central parameter and has the smallest difference value with the central parameter as the left boundary of the first interval, and taking the central parameter as the right boundary of the first interval to obtain the first interval of the scale interval;

Taking the target scale parameter which is larger than the central parameter and has the smallest difference value with the central parameter as the right boundary of the second interval, and taking the central parameter as the left boundary of the second interval to obtain the second interval of the scale interval;

taking the average value of the classification accuracy rates corresponding to the left boundary and the right boundary of the first interval as the classification accuracy rate of the first interval, and taking the average value of the classification accuracy rates corresponding to the left boundary and the right boundary of the second interval as the classification accuracy rate of the second interval.

By adopting the technical scheme, the central parameter corresponding to the maximum classification accuracy is taken as a basis, the scale parameter with the smallest difference smaller than the central parameter is selected as the left boundary of the first interval, and the scale parameter with the smallest difference larger than the central parameter is selected as the right boundary of the second interval. Therefore, the two sections are symmetrically distributed around the central parameter, the section ranges are basically equivalent, and unreasonable division of one section into one section and one section is avoided. Meanwhile, the method takes the average value of the classification accuracy of the boundary parameters of each interval as the classification accuracy of the interval, and can smoothly and randomly fluctuate to reflect the integral level of the interval. By means of the interval division mode, the method achieves balanced and accurate interval construction, and a reasonable range is provided for subsequent target interval comparison and selection. Compared with a simple random or equal interval generation method, the division mode considering the central sensitivity is more beneficial to enabling the parameter interval to be converged to the optimal, so that the BIM acquisition data processing effect is improved.

Optionally, determining the optimal scale parameter of the target building according to the target interval includes:

acquiring target scale parameters in the target interval according to the initial value of each scale parameter in the target interval, performing cross-validation on all the target scale parameters in the target interval to obtain the classification accuracy of each target scale parameter in the target interval, and acquiring a new target interval according to the initial value and the classification accuracy of all the target scale parameters in the target interval;

repeating the target interval updating operation until no scale parameter with higher classification accuracy than the classification accuracy of the central parameter of the latest target interval exists in the first interval and the second interval of the latest target interval, and stopping iteration;

taking the center parameter of the latest target interval as the optimal scale parameter.

By adopting the technical scheme, the classification accuracy of each parameter is obtained through cross verification based on the scale parameter and the initial value thereof in the current target interval. And then comparing the classification accuracy, updating the interval range, and obtaining a new target interval. Such iterative updating operations are performed until the parameter classification accuracy in the latest target section cannot exceed the section center parameter. At this time, the interval is reduced to only contain one optimal parameter, and the central parameter is the finally determined optimal scale parameter. This iterative approach to optimal parameters enables higher accuracy than single-step determinations. And finally, the obtained scale parameters optimize the parameter configuration of the SVM model, and the classification performance is maximized.

In a second aspect of the present application, there is provided a collected data processing system of a BIM, comprising:

the data acquisition module is used for determining the building type of a target building, acquiring historical building data of the same building type and constructing a vector machine data set based on the historical building data;

the training set construction module is used for acquiring all scale parameters in a preset scale interval and determining a training set corresponding to each scale parameter according to each scale parameter and the vector machine data set;

the kernel function determining module is used for determining the type of the building data according to the building data in the training set corresponding to each scale parameter and determining the kernel function corresponding to each scale parameter according to the type of the building data;

the target scale parameter determining module is used for determining an initial value of each scale parameter according to the kernel function corresponding to each scale parameter and the training set corresponding to each scale parameter, and determining a target scale parameter corresponding to each scale parameter according to the initial value of each scale parameter;

the accuracy rate determining module performs cross verification on all the target scale parameters to obtain the classification accuracy rate of each target scale parameter;

the optimal scale parameter determining module is used for determining a target interval according to the initial values and the classification accuracy of all target scale parameters in the scale interval and determining the optimal scale parameters of the target building according to the target interval;

The data classification module is used for constructing a target SVM hyperplane model of the target building type according to the optimal scale parameters, and processing building data to be classified according to the target SVM hyperplane model to obtain building data after separation processing.

In a third aspect the present application provides a computer storage medium having stored thereon a plurality of instructions adapted to be loaded by a processor and to carry out the above-described method steps.

In a fourth aspect of the present application, there is provided an electronic device comprising: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the method steps described above.

In summary, one or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:

1. according to the method, firstly, the type of a target building is determined, historical data of the type of building are collected to construct a vector machine data set, and a data basis is provided for training of a subsequent SVM model. Then, the method presets a scale interval, acquires all scale parameters in the interval, and determines a corresponding training set according to each scale parameter, so that the step traverses the scale parameter space, and conditions are created for searching the optimal scale parameters. Then, the method intelligently determines the kernel function of the corresponding parameter according to the data type of each training set, and combines data driving and a rule of thumb, so that the kernel function is selected more specifically. After the kernel function is determined, the method determines the initial value of each scale parameter according to the kernel function and the training set, and calculates the target scale parameter according to the initial value. The target scale parameters can obtain the classification accuracy through cross verification, and then the target interval is determined according to the initial value and the classification accuracy so as to lock the optimal parameter range. And finally, determining final scale parameters by using the target interval, and constructing an SVM model for classification. In this way, the optimal parameters are determined through the steps of data set construction, parameter traversal, kernel function determination, cross verification and the like, so that the BIM acquisition data processing effect is improved, and the problem that the performance is reduced due to improper parameter selection of the SVM is solved. The problem that the accuracy of the BIM model is affected by poor hyperplane construction effect and classification accuracy rate caused by improper selection of scale parameters of the kernel function is solved.

2. The relevance between the initial value and the classification accuracy is determined by calculating the correlation coefficient of the initial value and the classification accuracy. And then dividing the scale interval into two parts by taking the parameter corresponding to the maximum classification accuracy as a central parameter, and calculating the classification accuracy of each interval. And then acquiring scale parameters and initial value information in each section, and calculating the average initial value of each section. And determining the target scale of each section according to the average initial value, the classification accuracy and the correlation coefficient of the two, and comparing the target scales to determine a new target section. The method gradually reduces the selection range of the scale parameters through the iterative process of interval division and target degree comparison, so that the scale parameters approach to the optimization target. Compared with global scanning, the local iterative search mode improves the efficiency of parameter optimization. In conclusion, the target interval determination thought realizes more accurate and efficient scale parameter selection, and the performance of the final BIM acquisition data processing is enhanced.

3. The method and the device obtain the classification accuracy of each parameter through cross verification based on the scale parameter and the initial value thereof in the current target interval. And then comparing the classification accuracy, updating the interval range, and obtaining a new target interval. Such iterative updating operations are performed until the parameter classification accuracy in the latest target section cannot exceed the section center parameter. At this time, the interval is reduced to only contain one optimal parameter, and the central parameter is the finally determined optimal scale parameter. This iterative approach to optimal parameters enables higher accuracy than single-step determinations. And finally, the obtained scale parameters optimize the parameter configuration of the SVM model, and the classification performance is maximized.

Drawings

Fig. 1 is a schematic flow chart of a method for processing collected data of a BIM according to an embodiment of the present application;

FIG. 2 is a block diagram of a BIM data acquisition processing system according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of an electronic device provided in the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only some embodiments of the present application, but not all embodiments.

In the description of embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described herein as "such as" or "for example" should not be construed as being more target or advantageous than other embodiments or designs. Rather, the use of words such as "or" for example "is intended to present related concepts in a concrete fashion.

In the description of the embodiments of the present application, the term "plurality" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating an indicated technical feature. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.

In order to facilitate understanding of the methods and systems provided in the embodiments of the present application, a description of the background of the embodiments of the present application is provided before the description of the embodiments of the present application.

In the digital transformation and intelligent development of the building industry, an intelligent processing method for collected data of a BIM model is becoming an important field gradually. BIM technology is an information management and collaborative work platform for building and infrastructure projects. BIM technology integrates a variety of information including geometric data, component properties, materials, costs, and progress, etc., by creating a three-dimensional, digitized building model. When building data is collected, all the building data enter a BIM building information model, and data modeling is carried out on the collected building data, but the collected building data contains abnormal data due to errors or errors in the building data collecting process inevitably, so that the accuracy of the established model is affected. In order to prevent low accuracy of a final model caused by errors or errors in the building data acquisition process, data analysis processing is required to be performed on the acquired building data, so that abnormal classification monitoring on abnormal acquired data is realized, and the reliability of the data entering the BIM building information model is ensured.

At present, abnormal classification monitoring of abnormal collected data is usually carried out through an SVM support vector machine, but when the collected building data is classified through the existing SVM support vector machine, if a kernel function and a scale parameter corresponding to the kernel function are improperly set, a hyperplane construction effect is poor, the accuracy of abnormal classification of the collected building data is affected, and the accuracy of a BIM building model is further affected.

In view of the foregoing background description, those skilled in the art will appreciate that the problems occurring in the prior art, and it is evident that the following description of the embodiments of the present application, taken in conjunction with the accompanying drawings, refers to only a portion of the embodiments of the present application, and not to all of the embodiments.

On the basis of the background art, further, please refer to fig. 1, fig. 1 is a schematic flow chart of a method for processing collected data of a BIM according to an embodiment of the present application, the system may be implemented by a computer program, and may also operate as an independent tool application, specifically, in the embodiment of the present application, the method may be applied to a server, but may also be applied to an electronic device such as a server, and the method for processing collected data of a BIM includes the following steps:

S101, determining the building type of a target building, collecting historical building data of the same building type, and constructing a vector machine data set based on the historical building data;

specifically, the project attribute of the target building is determined, and which building type, such as office building, residence, business complex, and the like, the project attribute of the target building is determined. BIM data for the completed project for the type of building is then collected, which may originate from a historic archive of the design house or construction unit. To guarantee data size and representativeness, model data for a plurality of projects needs to be collected. And then preprocessing the collected building data, constructing sample characteristics, and generating a structured data set which can be processed by a vector machine.

Through the execution of the step, the method obtains sufficient and homogenized building data, and creates a data base for the subsequent construction of the SVM model. The adoption of the historical building data of the same type can improve the adaptability of the model to the target building, so that the model learns more effective classification knowledge. Compared with the method of directly adopting cross-type data, the method can greatly improve the training quality of the SVM model, thereby ensuring the effect of BIM acquisition data processing.

S102, acquiring all scale parameters in a preset scale interval, and determining a training set corresponding to each scale parameter according to each scale parameter and a vector machine data set;

Specifically, a scale interval is formed by presetting a value range of the scale parameter, namely determining a maximum value and a minimum value. And then gradually increasing the values of the scale parameters in the interval to acquire all parameter points in the interval. And training an SVM model by using the parameter value and the constructed vector machine data set for each sampled scale parameter to obtain a training set. Since the scale parameters influence the construction of the classification face, different parameters correspond to different training sets. Thus, by traversing the parameters within the interval, a series of training sets is obtained.

By executing the step, the method realizes the comprehensive scanning sampling of the scale parameter space, and obtains all possible training sets. The method creates conditions for the follow-up determination of the optimal parameters, so that the optimization process can be considered globally, and the situation of sinking into a local optimal solution is prevented. Compared with the direct selection of a plurality of discrete points, the traversal sampling mode can more accurately approximate the optimal parameters, so that the BIM acquisition data processing effect is improved.

S103, determining the type of building data according to building data in the training set corresponding to each scale parameter, and determining a kernel function corresponding to each scale parameter according to the type of building data;

Specifically, it is necessary to detect the number of dimensions and the type of distribution of the building data in each training set. For example, the principal component analysis method is used for analyzing the data dimension, and the kernel density estimation method is used for judging the distribution type. Then, according to the data characteristics, selecting a kernel function type: if the data dimension is low and the distribution is simple, selecting a linear kernel; if the data dimension is high and the distribution is complex, a Gaussian kernel is selected. Thus, different parameters may select different kernel functions due to the corresponding training set differences.

By executing the step, the method realizes the kernel function of suitable complexity according to the complexity of specific acquired data. Compared with a fixed kernel function, the method enables the expression capacity of the kernel function to be adapted to the data characteristics, and prevents overestimation or underestimation, so that the SVM model can train a more accurate classification surface. Therefore, the kernel function selection mode of parameter adaptation can improve the effect of BIM acquisition data processing.

On the basis of the foregoing embodiment, as an optional embodiment, S103, determining a type of building data according to building data in the training set corresponding to each scale parameter, and determining a kernel function corresponding to each scale parameter according to the type of building data, includes:

S201, classifying building data in a training set corresponding to each scale parameter, and determining the dimension type and the data distribution type of the building data;

specifically, an unsupervised clustering method is used to classify the building data. For example, using a K-means clustering algorithm, groups according to inter-sample distances, and then calculates a covariance matrix of data within each cluster. If the covariance matrix diagonal element value is far greater than the non-diagonal element value, judging the covariance matrix diagonal element value as low-dimensional data; otherwise, the data is judged to be high-dimensional data. Meanwhile, the data distribution of each cluster can be fitted, and if the data distribution accords with normal distribution, the data distribution is judged to be simple distribution; if irregular, the distribution is judged to be complex.

By executing this step, the method can determine the dimensional complexity and distribution type characteristics of the building dataset. These features can directly reflect the intrinsic rules of the data, which are the basis for selecting a linear core or a nonlinear core. Compared with a rule of thumb, the data driving mode can enable the kernel function to be more in line with actual building data conditions, and therefore the BIM acquisition data processing effect is improved.

S202, when the type of the building data is a low-dimensional type or a linear distribution type, determining a kernel function corresponding to the scale parameter as a linear kernel function;

Specifically, according to the data classification result of the previous step, if the building data set is found to be of a low-dimensional or linear distribution type, it is determined that the SVM parameter corresponding to the data set adopts a linear kernel function. The theory behind the choice of linear kernel functions is that such simple data structures can be modeled efficiently by linear functions. Compared with the complex kernel function, the linear kernel function has fewer parameters, is efficient in calculation, and can avoid the risk of overfitting.

By executing the step, the method realizes the selection of the kernel function with high matching degree according to the complexity of the data. The linear kernel function is suitable for simple data and can prevent excessive complications. Compared with the direct use of complex kernels, the method has the advantages that the model has higher data matching degree, and more accurate classification surface and better effect can be brought.

And S203, when the type of the building data is a high-dimensional type or a nonlinear distribution type, determining a kernel function corresponding to the scale parameter as a Gaussian kernel function.

Specifically, according to the data classification result of the previous step, if the building data set is found to be of a high-dimensional or nonlinear distribution type, determining that the SVM parameter corresponding to the data set adopts a Gaussian kernel function. The theoretical basis for selecting gaussian kernel functions is that such complex data structures require nonlinear functions to model. Compared with the linear kernel, the Gaussian kernel function has more parameters and stronger expression capability, and can fit complex mapping relations. And selecting a kernel function with high matching degree according to the complexity of the data. The Gaussian kernel function is suitable for complex data, and can prevent the problem of under fitting. Compared with the direct use of a simple kernel, the method ensures that the model has higher data conformity, and can bring more accurate decision surface and better effect.

S104, determining an initial value of each scale parameter according to the kernel function corresponding to each scale parameter and the training set corresponding to each scale parameter, and determining a target scale parameter corresponding to each scale parameter according to the initial value of each scale parameter;

specifically, a kernel function matrix, that is, a mapping relationship between the feature space and the historical building data of the target building data is calculated. And then calculating the kernel function distance between each piece of historical building data and other data based on the kernel matrix to obtain a kernel function value sequence. According to the kernel function sequence, the distribution of the historical building data in the parameter space can be evaluated, and the reasonable initial value of the scale parameter is determined. And finally, defining an optimized searching direction by taking the initial value as a starting point to obtain a first target scale parameter. By the method, the intelligent initialization of the scale parameters according to the specific data set is realized. Compared with random initialization, the data driving mode combined with the kernel matrix can enable the initial value of the parameter to be close to global optimum, and accelerate subsequent optimizing convergence. Meanwhile, the generation of the target parameters also has a certain basis, and no optimal blind search can occur.

On the basis of the above embodiment, as an alternative embodiment, determining an initial value of each scale parameter according to a kernel function corresponding to each scale parameter and a training set corresponding to each scale parameter, including the following steps:

S301, building data corresponding to each scale parameter is used as input of a corresponding kernel function, and a kernel function matrix corresponding to each scale parameter is obtained;

specifically, the collected target building data and the historical building data are sequentially input into corresponding parameter kernel functions, kernel function distances between the target data and each historical data are calculated, and a matrix is formed. The matrix element values reflect the similarity of different building data under the kernel function mapping. The diagonal of the matrix is the kernel function of each data, and the off-diagonal is the kernel function between different data.

By performing this step, the method obtains distance distribution information of the data set in the feature space. This reflects the correlation between different data more directly, which is an important basis for parameter initialization.

S302, determining a kernel function value sequence of each historical building data in the training set corresponding to each scale parameter according to the corresponding kernel function matrix;

based on the above embodiment, as an optional embodiment, S302, determining, according to the corresponding kernel function matrix, a kernel function value sequence of each scale parameter corresponding to each historical building data in the training set, further includes the following steps:

S401, determining a kernel function value between the nth historical building data and each historical building data except the nth historical building data in a kernel function matrix in a training set corresponding to the scale parameter;

specifically, by traversing the kernel function matrix, a row (or column) index n of each historical building data in the matrix is located. Then, after the n-th row (column) element is removed, the intersection elements of the remaining rows (columns) corresponding to the n-th row (column) are extracted. The element values are the kernel function calculated values of the nth historical building data and all other data.

S402, forming a kernel function value sequence of the nth building data based on kernel function values of the nth historical building data between the kernel function matrix and each historical building data except the nth historical building data.

Specifically, all the kernel function values extracted in the previous step are rearranged into a sequence according to the corresponding historical building data index sequence. I.e. each column of values constitutes a sequence, the element values in the sequence representing the kernel distance of the historic building data from all other data. Thus, each row of kernel function values constitutes a sequence of kernel functions.

By performing this step, the method obtains a sequence of coordinates for each historical building data in the parameter mapping space. This can intuitively reflect the distribution and density of data in the parameter space, and provide data support for initializing parameters.

Specifically, the kernel function sequence more directly reflects the data distribution state in the parameter space and provides a basis for parameter initialization. Element values of rows (or columns) of each historical building data in the kernel function matrix are extracted to form a kernel function sequence. The sequence consists of the kernel function values calculated for the historical building data and all other historical building data. The magnitude of the kernel function values in the sequence reflects the distance of the data relationship in different positions in the parameter space.

S303, determining initial values of corresponding scale parameters according to the kernel function value sequence.

Specifically, the kernel function sequence of each history building data is analyzed, and a high-density region in the parameter space is judged. These areas reflect a dense distribution of data and can be considered as reasonable intervals for parameter values. Then, a point which is in a high-density area and is close to the target building data is selected as an initial value of the parameter in combination with the kernel function sequence of the target building data. The simple random initialization of parameters is avoided, and the initial values of the parameters are scientifically determined according to the distribution condition of the data in the parameter space. The method can enable the initial value of the parameter to be closer to the global optimal solution, accelerate the subsequent optimizing process and prevent the parameter from being trapped into the local optimal solution.

S105, performing cross verification on all target scale parameters to obtain the classification accuracy of each target scale parameter;

specifically, a corresponding SVM model is trained based on each target scale parameter and the training set. And grouping the training sets, and cross-testing the classification accuracy of each model. I.e., training with some samples, other samples are validated, and the cycle is cross-looped until all samples are involved in the validation. Through the cross-validation, sample errors can be eliminated, and the classification accuracy of each parameter can be obtained.

S106, determining a target interval according to initial values and classification accuracy of all target scale parameters in the scale interval, and determining optimal scale parameters of a target building according to the target interval;

and according to the accuracy, determining an initial value corresponding to the target parameter, and selecting two parameter initial values with the smallest difference value and larger than and smaller than the initial value to form a target interval range. Then iteratively updating in the interval: and calculating the classification accuracy of the newly added parameters, and modifying the interval range until the interval is reduced to one point. The interval center parameter is the final optimal scale parameter. By means of the interval approximation mechanism similar to the bipartite strategy, the parameter optimization direction is clear, and the optimal value is quickly locked. Compared with the random traversal global, the interval iterative search mode considering the central sensitivity can effectively prevent the local optimum from being trapped. The single optimal parameters finally obtained can maximally improve the classification effect of the SVM model.

On the basis of the foregoing embodiment, as an optional implementation manner, S106 determines the target interval according to the initial values and the classification accuracy of all the target scale parameters in the scale interval, and includes the following steps:

s501, determining a correlation coefficient between an initial value and a classification accuracy according to the initial values and the classification accuracy of all target scale parameters in a scale interval;

specifically, a pearson correlation coefficient algorithm is adopted, and a linear correlation relationship between all the initial values of the target parameters and the classification accuracy obtained by corresponding cross verification is calculated. The resulting correlation coefficient is between-1 and 1, positive values representing positive correlations and negative values representing negative correlations.

By performing this step, the method can quantify the degree of association of the initial value and the classification effect. The high correlation coefficient indicates that the initial value of the parameter has a large influence on the classification accuracy, and the target interval needs to be determined around the initial value. Conversely, if the correlation coefficient is low, the range of the section needs to be enlarged.

S502, taking a target scale parameter corresponding to the maximum classification accuracy in the scale interval as a central parameter of the scale interval;

specifically, the target parameter corresponding to the value with the maximum classification accuracy is searched first. It can be decided that this parameter is the candidate point of all the current parameters that is most likely to be the final optimal parameter. It can thus be taken as the center point of the interval around which the interval range is subsequently adjusted for shrinkage optimization.

S503, dividing the scale interval into a first interval and a second interval according to the central parameter, and determining the classification accuracy of the first interval and the second interval;

specifically, the central parameter is taken as a demarcation point, the scale interval is equally divided into a left subinterval and a right subinterval, and the classification accuracy is obtained by cross verification aiming at the newly added target parameter in each subinterval. Thus, the classification accuracy level of each of the two sections is obtained. Through the execution of the step, the method realizes interval division under the guidance of the accuracy. Compared with equipartition, the dividing mode considering the accuracy distribution condition can lead the subsequent interval to be reduced more directionally and lead the optimization to be more efficient. And at the same time, whether the current interval already contains the optimal parameters can be checked.

S504, acquiring all scale parameters in the first interval and the second interval and initial values of each scale parameter;

specifically, traversing the two sections obtained by the previous step, extracting all generated target scale parameters in the sections, and recording initial values corresponding to the parameters. Thus, a section parameter initial value mapping table, namely a first section parameter initial value list and a second section parameter initial value list, can be constructed.

By executing the step, the method can obtain the complete parameter initial value information of the current two intervals. This provides a complete candidate set for subsequent re-determination of the parameter selection of the target interval according to the new optimization strategy. Compared with random parameters, the method for acquiring the initial value of the full-quantity parameters is more beneficial to the convergence of interval optimization, and the efficiency of BIM acquisition data processing can be remarkably improved.

S505, determining a first average initial value of the first section according to the initial values of all scale parameters of the first section, and determining a second average initial value of the second section according to the initial values of all scale parameters of the second section;

specifically, the initial value of the parameter of the first section is traversed, and an arithmetic average value is calculated as the average initial value of the first section. Similarly, the initial value of the parameter of the second section is traversed, and an arithmetic average value is calculated as the average initial value of the second section.

Through the execution of the step, the method can intuitively judge the overall distribution condition of the initial values of the parameters of the two sections at present. This provides a guideline for the overall trend for subsequent re-determination of the target interval, i.e. whether it is more concentrated in a larger or smaller initial value region.

S506, determining the target degree of the first interval according to the classification accuracy of the first average initial value and the first interval and the correlation coefficient between the initial value and the classification accuracy;

Specifically, by constructing an objective function, an average initial value, a classification accuracy and a correlation coefficient of the first interval are input as parameters. The objective function may be configured to: the more centralized the interval average initial value is, the higher the classification accuracy is, the larger the correlation coefficient is, and the higher the targeting degree is.

By executing the step, the method can obtain a quantitative index to judge whether the parameters of the first interval are concentrated and tend to the global optimal target. The range of the mesh-scale high-specification interval can be reduced; a low mesh size requires an enlarged interval.

S507, determining the target of the second interval according to the classification accuracy of the second average initial value and the second interval and the correlation coefficient between the initial value and the classification accuracy;

specifically, by constructing an objective function, the average initial value, the classification accuracy and the correlation coefficient of the two are input as parameters. The objective function may be configured to: the more centralized the interval average initial value is, the higher the classification accuracy is, the larger the correlation coefficient is, and the higher the targeting degree is. By executing the step, the method can obtain a quantitative index to judge whether the parameters of the second interval are concentrated and tend to the global optimal target. The range of the mesh-scale high-specification interval can be reduced; a low mesh size requires an enlarged interval.

S508, taking the second section and the section with the largest target scale in the second section as target sections.

Specifically, the target degree of the two intervals is compared, and the interval with the larger target degree is determined as a new target interval of the next iteration. Thus, the subdivision and reduction of the interval is realized.

Through the execution of the step, the method constructs an interval self-adaptive optimization strategy based on the target scale index. The mechanism can make the interval iteration direction more definite, quickly converge to the subspace containing the optimal parameters, and remarkably improve the parameter optimizing efficiency. Compared with a fixed interval, the interval dynamic adjustment driven by the mesh scale can greatly shorten the parameter optimization process, thereby enhancing the performance of BIM acquisition data processing.

On the basis of the above embodiment, as an alternative embodiment, determining, according to the target section, an optimal scale parameter of the target building includes:

s601, acquiring target scale parameters in a target section according to initial values of all the scale parameters in the target section, performing cross-validation on all the target scale parameters in the target section to obtain classification accuracy of each target scale parameter in the target section, and acquiring a new target section according to the initial values and classification accuracy of all the target scale parameters in the target section;

Specifically, all parameters in the target interval are cross-validated to obtain respective classification accuracy. And then, analyzing the correlation between the parameter initial value and the classification accuracy, and judging the influence degree of the initial value on the accuracy. And readjusting the range of the target interval according to the correlation result to form a new round of target interval.

S602, repeatedly performing the updating operation of the target interval until no scale parameter with higher classification accuracy than the classification accuracy of the central parameter of the latest target interval exists in the first interval and the second interval of the latest target interval, and stopping iteration;

specifically, after updating the target interval once, judging whether the two new partitioned intervals have scale parameters with classification accuracy higher than the central parameter of the current interval. If no such parameters exist, indicating that the central parameter may be a locally optimal solution, stopping the interval iterative optimization. By setting reasonable interval iteration stop conditions, the parameter space can be ensured to be fully optimized while avoiding the goal-free iteration.

S603, taking the center parameter of the latest target interval as the optimal scale parameter.

Specifically, when the stopping condition is met, the central parameter value of the current target interval is directly taken as the finally determined optimal scale parameter.

S107, constructing a target SVM hyperplane model of the target building type according to the optimal scale parameters, and processing building data to be classified according to the target SVM hyperplane model to obtain building data after separation processing.

Specifically, the optimal scale parameters are determined to be substituted into an SVM training process, and a hyperplane model is constructed. And inputting building BIM data to be classified into the model, and judging the data type according to the relationship between the data and the hyperplane, namely finishing the classification processing of the data.

By executing the step, the intelligent classification of the SVM model after the global optimization of the parameters to the acquired data is realized. Compared with a simple SVM, the model based on the optimized scale parameters learns a more accurate and effective decision surface, and classification performance is greatly improved. Depending on parameter optimization, the SVM can realize efficient and accurate processing on complex and diverse building data.

Referring to fig. 2, fig. 2 is a schematic diagram of a BIM data acquisition processing system according to an embodiment of the present application, where the BIM data acquisition processing system may include:

the data acquisition module 1 is used for determining the building type of a target building, acquiring historical building data of the same building type and constructing a vector machine data set based on the historical building data;

The training set construction module 2 is used for acquiring all scale parameters in a preset scale interval and determining a training set corresponding to each scale parameter according to each scale parameter and the vector machine data set;

the kernel function determining module 3 is used for determining the type of the building data according to the building data in the training set corresponding to each scale parameter and determining the kernel function corresponding to each scale parameter according to the type of the building data;

the target scale parameter determining module 4 is configured to determine an initial value of each scale parameter according to the kernel function corresponding to each scale parameter and the training set corresponding to each scale parameter, and determine a target scale parameter corresponding to each scale parameter according to the initial value of each scale parameter;

the accuracy determining module 5 performs cross verification on all the target scale parameters to obtain the classification accuracy of each target scale parameter;

the optimal scale parameter determining module 6 is used for determining a target interval according to the initial values and the classification accuracy of all the target scale parameters in the scale interval and determining the optimal scale parameters of the target building according to the target interval;

the data classification module 7 is used for constructing a target SVM hyperplane model of the target building type according to the optimal scale parameters, and processing building data to be classified according to the target SVM hyperplane model to obtain building data after separation processing.

It should be noted that: in the system provided in the above embodiment, when implementing the functions thereof, only the division of the above functional modules is used as an example, in practical application, the above functional allocation may be implemented by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to implement all or part of the functions described above. In addition, the system and method embodiments provided in the foregoing embodiments belong to the same concept, and specific implementation processes of the system and method embodiments are detailed in the method embodiments, which are not repeated herein.

Please refer to fig. 3, the present application also discloses an electronic device. Fig. 3 is a schematic structural diagram of an electronic device according to the disclosure in an embodiment of the present application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, at least one communication bus 302.

Wherein the communication bus 302 is used to enable connected communication between these components.

The user interface 303 may include a Display screen (Display), a Camera (Camera), and the optional user interface 303 may further include a standard wired interface, and a wireless interface.

The network interface 304 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface).

Wherein the processor 301 may include one or more processing cores. The processor 301 utilizes various interfaces and lines to connect various portions of the overall server, perform various functions of the server and process data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and invoking data stored in the memory 305. Alternatively, the processor 301 may be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable logic array (Programmable Logic Array, PLA). The processor 301 may integrate one or a combination of several of a central processing unit (Central Processing Unit, CPU), an image processor (Graphics Processing Unit, GPU), and a modem etc. The CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for rendering and drawing the content required to be displayed by the display screen; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 301 and may be implemented by a single chip. The memory 305 may include a random access memory (RandomAccessMemory, RAM) or a Read-only memory (rom). Optionally, the memory 305 includes a non-transitory computer readable medium (non-transitoroomputter-readabblestonemachineum). Memory 305 may be used to store instructions, programs, code, sets of codes, or sets of instructions. The memory 305 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-described respective method embodiments, etc.; the storage data area may store data or the like involved in the above respective method embodiments. Memory 305 may also optionally be at least one storage system located remotely from the aforementioned processor 301. Referring to fig. 3, an operating system, a network communication module, a user interface module, and an application program of a collected data processing method of the BIM may be included in the memory 305 as a computer storage medium.

In the electronic device 300 shown in fig. 3, the user interface 303 is mainly used for providing an input interface for a user, and acquiring data input by the user; and the processor 301 may be used to invoke an application program in the memory 305 that stores road assessment methods, which when executed by the one or more processors 301, causes the electronic device 300 to perform the methods as in one or more of the embodiments described above. It should be noted that, for simplicity of description, the foregoing method embodiments are all expressed as a series of action combinations, but it should be understood by those skilled in the art that the present application is not limited by the order of actions described, as some steps may be performed in other order or simultaneously in accordance with the present application. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all intended to be examples, and that the acts and modules referred to are not necessarily required in the present application. In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to related descriptions of other embodiments.

In the several embodiments provided herein, it should be understood that the disclosed system may be implemented in other ways. For example, the system embodiments described above are merely illustrative, e.g., the partitioning of elements, merely a logical functional partitioning, and there may be additional partitioning in actual implementation, e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted, or not implemented. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some service interface, system or unit indirect coupling or communication connection, electrical or otherwise.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

The embodiment of the present application further provides a computer storage medium, where the computer storage medium may store a plurality of instructions, where the instructions are adapted to be loaded by a processor and executed by the processor, where the specific execution process may refer to the specific description of the embodiment shown in fig. 1, and details are not repeated herein.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a memory, including several instructions for causing a computer device (which may be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of the embodiments of the present application. And the aforementioned memory includes: various media capable of storing program codes, such as a U disk, a mobile hard disk, a magnetic disk or an optical disk.

The above are merely exemplary embodiments of the present disclosure and are not intended to limit the scope of the present disclosure. That is, equivalent changes and modifications are contemplated by the teachings of this disclosure, which fall within the scope of the present disclosure. Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure.

This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a scope and spirit of the disclosure being indicated by the claims.

Claims

1. A method for processing collected data of a BIM, the method comprising:

acquiring all scale parameters in a preset scale interval, and determining a training set corresponding to each scale parameter according to each scale parameter and the vector machine data set;

Determining the type of building data according to building data in a training set corresponding to each scale parameter, and determining a kernel function corresponding to each scale parameter according to the type of the building data;

determining a target interval according to the initial values and the classification accuracy of all the target scale parameters in the scale interval, and determining the optimal scale parameters of a target building according to the target interval;

2. The method for processing collected data of BIM according to claim 1, wherein the determining the type of the building data according to the building data in the training set corresponding to each scale parameter, and determining the kernel function corresponding to each scale parameter according to the type of the building data includes:

and when the type of the building data is a high-dimensional type or a nonlinear distribution type, determining a kernel function corresponding to the scale parameter as a Gaussian kernel function.

3. The method for processing collected data of BIM according to claim 1, wherein the determining the initial value of each scale parameter according to the kernel function corresponding to each scale parameter and the training set corresponding to each scale parameter includes:

4. A method of processing collected data of a BIM according to claim 3, wherein the determining a sequence of kernel function values for each of the historical building data in the training set for each of the scale parameters according to the corresponding kernel function matrix includes:

Determining a kernel function value between the nth historical building data and each historical building data except the nth historical building data in a kernel function matrix in a training set corresponding to the scale parameter;

and forming a kernel function value sequence of the nth building data based on the kernel function value between the nth historical building data and each historical building data except the nth historical building data in the kernel function matrix.

5. The method for processing collected data of BIM according to claim 1, wherein the determining the target interval according to the initial values and the classification accuracy of all the target scale parameters in the scale interval includes:

6. The method for processing collected data of BIM according to claim 1, wherein the dividing the scale section into a first section and a second section according to the central parameter, and determining the classification accuracy of the first section and the second section includes:

7. The method for processing collected data of BIM according to claim 1, wherein the determining the optimal scale parameter of the target building according to the target section includes:

8. A system for processing acquired data from a BIM, comprising:

the target scale parameter determining module is used for determining an initial value of each scale parameter according to a kernel function corresponding to each scale parameter and a training set corresponding to each scale parameter, and determining a target scale parameter corresponding to each scale parameter according to the initial value of each scale parameter;

the accuracy rate determining module is used for carrying out cross verification on all the target scale parameters to obtain the classification accuracy rate of each target scale parameter;

The optimal scale parameter determining module is used for determining a target interval according to the initial values and the classification accuracy of all the target scale parameters in the scale interval and determining the optimal scale parameters of a target building according to the target interval;

the data classification module is used for constructing a target SVM hyperplane model of a target building type according to the optimal scale parameters, and processing building data to be classified according to the target SVM hyperplane model to obtain building data after separation processing.

9. A computer readable storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the method of any one of claims 1 to 7.

10. An electronic device comprising a processor, a memory for storing instructions, and a transceiver for communicating with other devices, the processor for executing the instructions stored in the memory to cause the electronic device to perform the method of any one of claims 1-7.