Engineering building component model construction method and system based on BIM

Abstract

The invention relates to the technical field of image processing, in particular to an engineering building component model construction method and system based on BIM, and the method comprises the steps: collecting initial point cloud data of a target engineering building component; performing feature analysis processing on the initial point cloud data to correct a noisy value of each point cloud when an existing filtering algorithm is used for calculation; and according to the corrected noisy value, judging a noise point cloud and removing the noise point cloud to obtain effective point cloud data which is used for constructing a BIM engineering building component model. According to the method and the device, the denoising result of the radius filtering algorithm using the fixed radius is corrected by analyzing the point cloud features, so that a more reliable denoising result is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a BIM-based engineering building component model construction method and system. BACKGROUND

[0002] In the field of modern construction engineering, Building Information Modeling (BIM) has become an indispensable tool. It simulates buildings and their components in a digital way, providing an integrated platform for design, construction and operation management. However, in actual operation, especially when renovating or maintaining existing buildings, the problem of how to efficiently and accurately convert the building structure in the real world into a digital model is faced.

[0003] Point cloud technology, as an advanced three-dimensional data acquisition method, can capture the geometric shape of the building surface with high precision. However, due to the complex and variable scanning environment, the obtained point cloud data often contains a large amount of noise, such as "fly points" (i.e. outliers), overlapping areas and irregular distribution, which seriously affects the accuracy of subsequent data processing and BIM model construction. In order to improve the efficiency and quality of the conversion from point cloud data to BIM model, the original point cloud data must first be effectively denoised.

[0004] Most existing filtering algorithms only analyze the noise of point cloud based on point cloud coordinates and point cloud grayscale values, etc. However, in actual processing, such as the existing radius filtering algorithm, the selection of search radius and minimum neighbor number directly affects the filtering effect. Too large a radius may result in the deletion of normal points; too small a radius may not effectively remove noise points. SUMMARY

[0005] To solve the above technical problems, the present application provides a BIM-based engineering building component model construction method and system, and the technical solutions adopted are as follows:

[0006] In a first aspect, an embodiment of the present application provides a BIM-based engineering building component model construction method, which comprises the following steps:

[0007] Scanning the target engineering building component to obtain initial point cloud data containing three-dimensional coordinates and intensity values;

[0008] Based on the preset radius threshold, a radius filtering algorithm is used to calculate the noise value of each point cloud.

[0009] a neighborhood centered on the target point cloud is evenly divided into multiple sub-regions; clustering is performed based on the reflection intensity values of all point clouds in the neighborhood of the target point cloud; a first self-similarity between the target point cloud and any point cloud in its neighborhood is determined based on the difference between the number of point clouds belonging to the same cluster category in the same sub-region in the neighborhood of the target point cloud and the number of point clouds belonging to the same cluster category in the same sub-region in the neighborhood of the target point cloud;

[0010] The reflection intensity value of the target point cloud is compared with the average value of the reflection intensity values of all point cloud data in each sub-region of the target point cloud to calculate a texture feature descriptor of the target point cloud, thereby determining a second self-similarity between the target point cloud and any point cloud in its neighborhood; and combining the first self-similarity to determine a first correction value of the target point cloud as noise data;

[0011] The difference in normal vectors after surface fitting of the target point cloud and the point clouds in its neighborhood is analyzed to determine the regularity of the local neighborhood where the target point cloud is located, and the first correction value is used to correct the noise value of the target point cloud;

[0012] Noise point clouds are determined and removed according to the corrected noise value to obtain effective point cloud data for constructing a BIM engineering building component model.

[0013] Preferably, the method for calculating the noise value is:

[0014] If the number of point clouds in the neighborhood of the target point cloud is greater than or equal to a preset number threshold, the noise value of the target point cloud is set to a first preset value;

[0015] Otherwise, the noise value of the target point cloud is calculated according to the ratio of the absolute value of the difference between the number of point clouds in the neighborhood and the preset number threshold to the preset number threshold.

[0016] Preferably, the neighborhood is a spherical discrimination region centered on the target point cloud; the sub-regions are obtained by evenly dividing the spherical discrimination region into 8 parts using an 8-part division method with the target point cloud as the center.

[0017] Preferably, the method for calculating the first self-similarity is:

[0018] wherein represents the first self-similarity between the target point cloud u and the vth point cloud in its neighborhood, norm represents a normalization function, represents the absolute value of the difference between the number of point clouds contained in the i th sub-region of the spherical discrimination region centered on the target point cloud u and the v th point cloud in its neighborhood, K represents the number of point cloud intensity categories obtained based on the reflection intensity values of the point clouds in the neighborhood of the target point cloud u, The absolute value of the difference between the number of point clouds in which the reflection intensity value of the point cloud in the i-th region of the spherical discriminant region centered on the v-th point cloud in the neighborhood of the target point cloud u and divided by the target point cloud u and the j-th category belongs to.

[0019] Preferably, the reflection intensity value of the target point cloud is compared with the average value of all point cloud data reflection intensity values in each sub-region thereof to calculate the texture feature descriptor of the target point cloud, thereby determining the second self-similarity between the target point cloud and any point cloud in its neighborhood, comprising:

[0020] Marking the sub-region with a reflection intensity value greater than that of the target point cloud as 1, and vice versa as 0;

[0021] Arranging the marking results in ascending order of the sub-region division sequence, and converting the sorted binary data into decimal as the texture feature descriptor of the target point cloud;

[0022] According to the difference in the texture feature descriptor between the target point cloud and any point cloud in its neighborhood, the second self-similarity is calculated.

[0023] Preferably, the first correction value is positively correlated with the first self-similarity and the second self-similarity between the target point cloud and any point cloud in its neighborhood, respectively.

[0024] Preferably, the method for determining the regularity of the component comprises: obtaining the normal vector of the target point cloud and all point clouds in its neighborhood on the fitted surface; calculating the variance of the angle difference between the normal vector of the target point cloud and the normal vectors of all point clouds in its neighborhood, and using 1 minus the normalized value of the variance as the regularity of the component in the local neighborhood of the target point cloud.

[0025] Preferably, the calculation method for correcting the noise value of the target point cloud is:

[0026] wherein represents the corrected noise value of the target point cloud u, represents the noise value of the target point cloud u, represents the regularity of the component in the local neighborhood of the target point cloud u, represents the first correction value of the target point cloud u as noise data.

[0027] Preferably, when the corrected noise value is greater than the first preset value, the corresponding point cloud is determined to be a noise point cloud and is removed.

[0028] In a second aspect, another embodiment of the present application further provides a BIM-based engineering building component model construction system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the BIM-based engineering building component model construction method described above when executing the computer program.

[0029] The present application has at least the following beneficial effects:

[0030] 1. The traditional radius filtering algorithm is prone to the problems of "radius too large to delete normal points" or "radius too small to miss noise points" due to the use of a fixed radius threshold. The present application introduces local spatial distribution + self-similarity analysis of intensity texture by dynamically correcting noisy values, so that the filtering threshold is adaptively adjusted according to the local features of the component such as regularity and texture consistency, significantly improving the denoising accuracy.

[0031] 2. By joint analysis of the first self-similarity and the second self-similarity, the geometric features and physical properties of the point cloud are deeply integrated. This method improves the recognition accuracy when dealing with noise in complex scenes such as overlapping areas and material abrupt change areas.

[0032] 3. The regularity of the component is introduced as a correction weight, and the local surface smoothness is quantified by the normal vector variance. When the component surface is regular, the reliability of the self-similarity analysis result is high, and the correction weight is increased; otherwise, the influence of self-similarity is reduced. This mechanism effectively avoids the misjudgment of normal points caused by complex local structure. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0034] Figure 1 A flowchart of a BIM-based engineering building component model construction method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0035] Embodiment 1

[0036] An embodiment of the present application provides a BIM-based engineering building component model construction method, which specifically refers to Figure 1 The method comprises the following steps:

[0037] Step S1: Collecting initial point cloud data of the target engineering building component.

[0038] The target engineering building component is scanned using a three-dimensional laser scanner to obtain initial point cloud data of the target engineering building component. The point cloud data format is P={x, y, z, a}, wherein x, y, and z are three-dimensional coordinates of the point cloud data, and a represents a reflection intensity value of the point cloud data.

[0039] Step S2: performing feature analysis processing on the initial point cloud data to correct the noise value of each point cloud calculated by using the existing filtering algorithm.

[0040] In combination with the autocorrelation characteristics of the point cloud data, the noise analysis result obtained by using the existing filtering algorithm is corrected to obtain a more accurate and effective point cloud data preprocessing result. Details are as follows:

[0041] a. Based on the preset radius threshold, a radius filtering algorithm is used on the initial point cloud data to calculate the noise value of each point cloud.

[0042] The initial point cloud data obtained is denoised by using the existing filtering algorithm. In this embodiment, a radius filtering algorithm with a preset first radius threshold r=15 is used to denoise the obtained point cloud data. In the existing radius filtering algorithm, the target point cloud whose number of contained point clouds in the radius is less than the preset number threshold is recorded as a noise point cloud and is removed.

[0043] In order to better analyze the noise of the point cloud data and correct the analysis result, the application records the noise value of the target point cloud greater than or equal to the preset number threshold as C, C is a first preset value, and the value in this embodiment is 0.5. The calculation method of the noise value of the target point cloud whose number of contained point clouds is less than the preset number threshold is as follows: the absolute value of the difference between the number of contained point clouds in the preset radius neighborhood of the target point cloud and the preset number threshold is calculated, the ratio of the difference absolute value to the preset number threshold is divided by 2, and then 0.5 is added to obtain the noise value of the target point cloud.

[0044] That is, the greater the difference between the number of contained point clouds in the radius neighborhood of the target point cloud and the preset number threshold, the greater the noise value of the target point cloud.

[0045] According to the above result, if the noise value of the target point cloud is greater than C, it can be concluded that the point cloud is noise data.

[0046] b. Based on the local self-similarity of the point cloud data, the noise value of the point cloud is corrected to obtain a corrected noise value.

[0047] For relatively regular engineering building components, adjacent point cloud regions have the feature of self-similarity. Therefore, the acquired point cloud data can be analyzed and judged by analyzing the self-similarity characteristics between the point cloud data and the adjacent point cloud data, so as to determine whether the point cloud is noise data.

[0048] In the application, a self-similarity analysis model is constructed, a spherical discrimination region with a preset second radius threshold is constructed around the target point cloud as a neighborhood of the target point cloud, and then the correlation between the target point cloud data and the spherical discrimination region and the correlation between the spherical discrimination region of the neighboring point cloud of the target point cloud are analyzed by calculating the correlations, so as to determine whether the target point cloud is noise point cloud data. Specifically, the following steps are performed.

[0049] The spherical discrimination region is divided into eight parts by using an eight-part division method with the target point cloud as the center, and eight sub-regions of the eight parts are obtained. The number of point clouds contained in each sub-region is counted.

[0050] In the classification, the reflection intensity data of the point cloud data are further considered. Therefore, in the analysis, the reflection intensity values of all point clouds in the neighborhood of the target point cloud are clustered. In the classification, the distribution difference of the reflection intensity values of the point cloud data in each sub-region is further analyzed, so that a more accurate autocorrelation analysis result is obtained. In this embodiment, the DBSCAN clustering algorithm is used to cluster the reflection intensity values of all point cloud data obtained by scanning. In this embodiment, the clustering parameters are r=5 and minpts=5. The number of clusters is denoted as K.

[0051] In the judgment process of whether the target point cloud is noise based on self-similarity analysis, whether the number of point clouds belonging to each cluster in the above-mentioned one sub-region is consistent is analyzed.

[0052] Further, based on the difference between the number of point clouds belonging to the same cluster in the same sub-region in the neighborhood of different point clouds, the first self-similarity between the target point cloud and any point cloud in the neighborhood of the target point cloud is determined. The calculation method of the first self-similarity between the target point cloud and the point cloud in the neighborhood of the target point cloud is as follows:

[0053] wherein S(u,v) represents the first self-similarity between the target point cloud u and the vth point cloud in the neighborhood of the target point cloud, norm represents a normalization function, S(u,v) represents the first self-similarity between the target point cloud u and the vth point cloud in the neighborhood of the target point cloud, norm represents a normalization function, S(u,v) represents the first self-similarity between the target point cloud u and the vth point cloud in the neighborhood of the target point cloud, norm represents a normalization function, ​The construction not only considers the spatial distribution information, but also combines the texture consistency. That is, if the same region has similar point cloud data, but the corresponding point cloud data has different materials, the above method can also find the corresponding differences.

[0054] Therefore, the above self-similarity index can reflect the comprehensive similarity degree of the spatial distribution and intensity characteristics of the two in the local spherical neighborhood. The value closer to 1 indicates that the local structures of the two point clouds are more similar; the smaller the value, the greater the difference, indicating that there may be noise points.

[0055] Meanwhile, the application further considers the self-similarity of the point cloud texture information based on the reflection intensity value of the point cloud, analyzes the texture self-similarity between the target point cloud and the point cloud in its neighborhood, and analyzes the noise value of the target point cloud.

[0056] The application obtains the point cloud data contained in each sub-region of the target point cloud, records the average value of the reflection intensity values of all point cloud data in each sub-region as the point cloud average reflection intensity value of the corresponding sub-region, compares the average value of the point cloud data intensity in each sub-region with the reflection intensity value of the target point cloud, marks the sub-region greater than the reflection intensity value of the target point cloud as 1, and marks the sub-region smaller than the reflection intensity value of the target point cloud as 0. The marking result is arranged from small to large according to the division order of the 8 sub-regions, and the sorted binary data is converted into decimal as the texture feature descriptor of the target point cloud. Thus, the texture feature descriptor based on the reflection intensity value of the target point cloud and its neighborhood point cloud data is obtained.

[0057] The texture feature descriptor of the target point cloud and the point cloud contained in its neighborhood is calculated using this method. Taking the target point cloud u and the vth point cloud data in its neighborhood as an example, the absolute value of the difference between the texture feature descriptors of the target point cloud and the corresponding point cloud in its neighborhood is normalized, and 1 is subtracted from the normalized value to record the second self-similarity between the target point cloud u and the vth point cloud in its neighborhood. The greater the difference between the texture feature descriptors of the target point cloud and the corresponding point cloud in its neighborhood, the more likely the target point cloud is a noise point cloud data.

[0058] According to the above self-similarity analysis, the first correction value of the target point cloud data as noise data is obtained:

[0059] In the application, the first correction value is positively correlated with the first self-similarity and the second self-similarity between the target point cloud and any point cloud in its neighborhood, respectively.

[0060] It can be understood that the positive correlation means that the dependent variable increases with the increase of the independent variable, and the dependent variable decreases with the decrease of the independent variable. The actual application determines, and the application does not make special limitation.

[0061] wherein the first correction value of the target point cloud u as noise data For example, the calculation relationship is:

[0062] wherein The first correction value of the target point cloud u as noise data is represented by norm, and the normalization function is represented by norm. The number of point clouds contained in the target point cloud u in the spherical discrimination region with a neighborhood radius of r is represented by norm. The first self-similarity of the target point cloud u and the vth point cloud in its neighborhood is represented by norm. The second self-similarity between the target point cloud u and the vth point cloud in its neighborhood is represented by norm. The obtained first correction value reflects the degree of comprehensive deviation of the target point cloud in the local neighborhood from the surrounding point clouds in terms of structural self-similarity and texture feature consistency. This value will be an important input for subsequent “noise-containing correction value” calculation, which is used to more accurately identify and retain real structure points and eliminate noise points.

[0063] At the same time, when performing local similarity analysis, the reliability of the result is highly related to the regularity of the building component, so the regularity of the corresponding building component needs to be analyzed.

[0064] The application calculates the local regularity of the building component in the local area where the corresponding point cloud is located, and then modifies the noise judgment result obtained based on the local similarity analysis according to the obtained local regularity of the building component. The calculation method of the local regularity of the building component is as follows:

[0065] First, analyze the regularity of the point cloud region where all point clouds in the neighborhood range of the target point cloud are located. In the present application, the more flat the regularity of the building component is, the smaller the difference between the normal vectors of the fitting planes is.

[0066] Take one of the target point clouds u as an example for analysis. First, perform surface fitting on the point cloud and the point cloud data in its neighborhood, wherein the surface fitting is performed using the least squares method. Accordingly, the normal vector of each point cloud based on the fitted surface can be obtained. If the local area of the building component where the point cloud is located is regular, then the normal vector of the fitted surface obtained from the corresponding point cloud data should also be similar. Therefore, the regularity of the fitting plane can be obtained by calculating the angle difference of the normal vectors.

[0067] The application obtains the normal vector corresponding to the target point cloud, calculates the variance of the angle difference between the normal vector of all point clouds in the neighborhood of the target point cloud and the normal vector of the target point cloud, and uses 1 minus the normalized value of the variance as the regularity degree Tn of the component in the local neighborhood of the target point cloud. The smaller the variance of the sequence is, the higher the local regularity degree of the component in the neighborhood of the target point cloud is, and the greater the reliability of the judgment result of whether the target point cloud is noise obtained according to the autocorrelation analysis of the component point cloud is.

[0068] c. Point cloud noise correction.

[0069] Through the point cloud texture information analysis constructed by combining the point cloud distribution and the reflection intensity value, the analysis result obtained by using the radius filtering is corrected based on the self-similarity analysis of the target point cloud and its neighborhood, and the specific calculation method is as follows:

[0070] Wherein represents the corrected noise value of the target point cloud u, represents the noise value of the target point cloud u, represents the regularity degree of the component in the local neighborhood of the target point cloud u, represents the first correction value of the target point cloud u as noise data.

[0071] That is, the greater the regularity degree of the neighborhood region of the target point cloud is, the more reliable the first correction value obtained according to the analysis is, and the greater the weight value of the corresponding target point cloud is. On the contrary, the smaller the regularity degree of the neighborhood region of the target point cloud is, the less reliable the first correction value obtained according to the self-similarity analysis is.

[0072] Thus, the corrected noise value of the target point cloud based on the local self-similarity analysis of the point cloud is obtained.

[0073] Step S3: judging and removing the noise point cloud according to the corrected noise value to obtain effective point cloud data for constructing the BIM engineering building component model.

[0074] According to the calculation method of the initial noise value of the target point cloud in the above steps, when the corrected noise value of the target point cloud is greater than the first preset value C, it can be concluded that the target point cloud is a noise point cloud, which needs to be removed. The removed point cloud is used as effective point cloud. The obtained effective point cloud is used to construct the BIM engineering building component model.

[0075] Example 2

[0076] Another embodiment of the present application further provides a BIM-based engineering building component model construction system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the above-mentioned BIM-based engineering building component model construction method when executing the computer program.

[0077] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the application embrace any and all variations of the present application that fall within the scope of the general inventive concept as defined by the appended claims and their equivalents.

[0078] It is to be understood that the application is not limited to the precise construction described above and shown in the attached drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application.

Claims

1. A method for constructing an engineering building component model based on BIM, characterized in that: The method comprises the following steps: Scan the target engineering building components to obtain initial point cloud data containing three-dimensional coordinates and strength values; Based on the preset radius threshold, the radius filtering algorithm is applied to the initial point cloud data to calculate the noise value of each point cloud; The neighborhood centered on the target point cloud is evenly divided into multiple sub-regions; all point clouds in the neighborhood of the target point cloud are clustered based on their reflection intensity values; and the first self-similarity between the target point cloud and any point cloud in its neighborhood is determined based on the difference in the number of point clouds belonging to the same cluster category in the same sub-region within the neighborhood of different point clouds. Comparing the reflection intensity value of the target point cloud with the average reflection intensity values ​​of all point cloud data in each of its sub-regions to calculate the texture feature descriptor of the target point cloud, thereby determining the second self-similarity between the target point cloud and any point cloud in its neighborhood; and combining the first self-similarity to determine the target point cloud as a first correction value of noise data; Analyzing the difference in normal vectors between the target point cloud and the point clouds in its neighborhood after surface fitting, determining the degree of component regularity in the local neighborhood where the target point cloud is located, and correcting the noise value of the target point cloud using the first correction value; The noise point cloud is determined and removed based on the corrected noise value to obtain valid point cloud data for constructing the BIM engineering building component model.

2. A method for constructing a BIM-based engineering building component model according to claim 1, characterized in that: The calculation method of the noise value is: If the number of point clouds of the target point cloud in its neighborhood is greater than or equal to a preset number threshold, the noise value of the target point cloud is set to a first preset value; Otherwise, the noise value of the target point cloud is calculated based on the ratio of the absolute value of the difference between the number of point clouds in the neighborhood and the preset number threshold to the preset number threshold.

3. The method for constructing a BIM-based engineering building component model according to claim 1, wherein: The neighborhood is a spherical discrimination area divided with the target point cloud as the center; the sub-area is 8 8-point sub-areas obtained by evenly dividing the spherical discrimination area into 8 parts using an 8-point division method with the target point cloud as the center.

4. A method for constructing a BIM-based engineering building component model according to claim 3, characterized in that: The calculation method of the first self-similarity is: in represents the first self-similarity between the target point cloud u and the vth point cloud in its neighborhood, norm represents the normalization function, represents the absolute value of the difference between the number of point clouds contained in the ith subregion of the spherical discrimination region divided by the target point cloud u and the vth point cloud in its neighborhood. K represents the number of point cloud intensity categories obtained by clustering the point cloud reflection intensity values ​​in the neighborhood of the target point cloud u. It represents the absolute value of the difference between the number of point clouds whose reflection intensity values ​​belong to the jth category in the spherical discrimination area divided by the target point cloud u and the vth point cloud in its neighborhood with each point cloud as the center.

5. The method for constructing a BIM-based engineering building component model according to claim 3, wherein: The step of comparing the reflection intensity value of the target point cloud with the average reflection intensity values ​​of all point cloud data in each of its sub-regions to calculate the texture feature descriptor of the target point cloud, thereby determining the second self-similarity between the target point cloud and any point cloud in its neighborhood, includes: The sub-regions with reflection intensity values ​​greater than the target point cloud are marked as 1, and vice versa; Arrange the marking results in ascending order according to the sub-region division order, and convert the sorted binary data into decimals as the texture feature descriptor of the target point cloud; The second self-similarity is calculated according to the difference in texture feature descriptors between the target point cloud and any point cloud in its neighborhood.

6. A method for constructing a construction component model based on BIM according to claim 4 or 5, characterized in that: The first correction value is positively correlated with the first self-similarity and the second self-similarity between the target point cloud and any point cloud in its neighborhood.

7. A method for constructing a BIM-based engineering building component model according to claim 6, characterized in that: The method for determining the degree of component regularity is as follows: obtaining the normal vectors of the target point cloud and all point clouds in its neighborhood on the fitting surface; calculating the variance of the angular differences between the normal vectors of all point clouds in the neighborhood of the target point cloud and the normal vector of the target point cloud, and using 1 minus the normalized value of the variance to record it as the degree of component regularity of the local neighborhood where the target point cloud is located.

8. A method for constructing a BIM-based engineering building component model according to claim 7, characterized in that: The calculation method for correcting the noise value of the target point cloud is: in Indicates the corrected noise value of the target point cloud u, Represents the noise value of the target point cloud u, Indicates the degree of component regularity in the local neighborhood where the target point cloud u is located, Indicates that the target point cloud u is the first correction value of the noise data.

9. The method for constructing a BIM-based engineering building component model according to claim 8, wherein: When the corrected noise value is greater than the first preset value, the corresponding point cloud is determined to be a noise point cloud and is removed.

10. A BIM-based engineering building component model construction system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method for constructing an engineering building component model based on BIM as claimed in claim 1 is implemented.

Images