Method and system for constructing three-dimensional model of precast concrete component

Abstract

The invention relates to the technical field of prefabricated part three-dimensional construction, in particular to a concrete prefabricated part three-dimensional model construction method and system. The method comprises the following steps: acquiring point cloud data corresponding to a concrete prefabricated part; obtaining a sorting sequence corresponding to each point in the point cloud data; obtaining the local change intensity of a point according to the sorting sequence corresponding to the point, and further calculating the required detail degree of the point; determining a K value of each point and an adjacent point set; calculating the curvature value of each point according to the adjacent point set of each point, and obtaining suspected noise points; the adjacent suspected noise points are divided into the same connected domain, a first sequence is obtained, and then the structural characteristic value of each suspected noise point is calculated; screening each suspected noise point to obtain a real noise point; denoising real noise points in the point cloud data to obtain denoised point cloud data; and constructing a three-dimensional model by using the noise reduction point cloud data. The construction accuracy of the three-dimensional model of the precast concrete component can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional construction of prefabricated components, and in particular to a method and system for constructing a three-dimensional model of a prefabricated concrete component. Background Art

[0002] The three-dimensional model of precast concrete components refers to the conversion of the shape, structure and other information of real-life precast concrete components into a three-dimensional coordinate model through three-dimensional digital technology. The three-dimensional model of precast concrete components can realize digital quality inspection, virtual assembly simulation, collision detection, etc. of precast components, which can significantly improve the accuracy and efficiency of components throughout the entire process of design, manufacturing, construction and operation and maintenance. Three-dimensional point cloud technology is a commonly used method in the three-dimensional model of precast concrete components. However, because the reflected signal is easily interfered with by other factors when scanning precast concrete components, the point cloud data obtained by scanning often contains noise points. Therefore, denoising the obtained point cloud data is a key step to ensure the accuracy of the constructed three-dimensional model.

[0003] Existing denoising methods for point cloud data often cannot effectively distinguish between noise points and structures such as holes and corners on the surface of prefabricated components, resulting in the loss of some of the true structural information on the surface of prefabricated components after denoising, affecting the accuracy of further quality inspections, collision simulations, and other results of prefabricated components. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a method and system for constructing a three-dimensional model of a precast concrete component. The technical solutions adopted are as follows: In a first aspect, an embodiment of the present invention provides a method for constructing a three-dimensional model of a precast concrete component, the method comprising: Scan the precast concrete components to obtain point cloud data; construct a window with a point in the point cloud data as the center, and sort the points in the window to obtain a sorting sequence; The local variation intensity of a point is obtained based on the coordinate difference between each two adjacent points in the sorted sequence corresponding to the point; the required level of detail of the point is calculated based on the local variation intensity of the point and the adjacent points before and after it in the sorted sequence; Determine the K value of each point according to the required level of detail of each point, and obtain the neighboring point set of each point respectively; calculate the curvature value of each point according to the neighboring point set of each point, and obtain the suspected noise point; Divide adjacent suspected noise points into the same connected domain, sort the suspected noise points in the connected domain using a minimum curvature path starting from a point on the edge of the connected domain to obtain a first sequence; obtain the structural characteristic value of each suspected noise point in the connected domain based on the first sequence; Based on the structural characteristic values ​​and curvature values ​​of each suspected noise point, the suspected noise points are screened to obtain the real noise points; the real noise points in the point cloud data are denoised to obtain the denoised point cloud data; and the denoised point cloud data is used to construct a three-dimensional model.

[0005] Preferably, sorting the points in the window to obtain a sorted sequence includes: Take one of the points in the window corresponding to a point that is closest to the top vertex of the window as the starting point, calculate the distance between each point in the window except the starting point and the starting point, and take the point with the smallest distance as the second point; calculate the distance between the other points in the window except the starting point and the second point and the second point, and take the point with the smallest distance as the third point, and so on, until all points in the window corresponding to the point are traversed, the points in the window are sorted, and the sorting sequence corresponding to the point is obtained.

[0006] Preferably, obtaining the severity of the local change of a point according to the coordinate difference between every two adjacent points in the sorting sequence corresponding to the point includes: Obtain the sum of the absolute values ​​of the differences between the coordinates of two adjacent points in the sequence corresponding to a point, and record it as the coordinate difference of the two adjacent points; calculate the mean of the coordinate differences between every two adjacent points in the sequence corresponding to the point to obtain the severity of the local change of the point.

[0007] Preferably, the required level of detail of a point is calculated based on the severity of the local change of the point and the adjacent points in the corresponding sorting sequence, including: The required level of detail of a point is obtained by obtaining the sum of the distances between the adjacent points before and after a point in its corresponding sorting sequence, multiplying it by the severity of the local change of the point, and then normalizing it.

[0008] Preferably, the K value of each point is determined according to the required level of detail of each point, and the neighboring point set of each point is obtained respectively, including: The set initial K value is multiplied by the difference between the first preset value and the required level of detail of a point and rounded up to obtain the K value of the point; then the K value of each point is combined with the KNN algorithm to obtain the neighboring point set of each point.

[0009] Preferably, calculating the curvature value of each point based on the neighboring point set of each point and obtaining the suspected noise point includes: The covariance matrix is ​​calculated based on the set of neighboring points of a point, and the eigenvalue decomposition of the covariance matrix is ​​performed to obtain the normal eigenvalue, the first eigenvalue and the second eigenvalue; the normal eigenvalue is divided by the sum of the normal eigenvalue, the first eigenvalue and the second eigenvalue to obtain the curvature value of the point; the curvature value of each point is clustered to obtain different clusters, and the maximum curvature value in the largest cluster is selected as the judgment threshold. When the curvature value of a point is greater than the judgment threshold, the point is a suspected noise point.

[0010] Preferably, obtaining the structural characteristic value of each suspected noise point in the connected domain based on the first sequence includes: Calculate the difference between the curvature value of the latter suspected noise point and the curvature value of the former suspected noise point in the first sequence corresponding to a connected domain, and record it as the curvature difference of the latter suspected noise point; sort the curvature differences of every two adjacent suspected noise points according to the order of the first sequence to obtain a curvature difference sequence; use the least squares method to perform linear fitting on the curvature difference sequence to obtain a fitting curve; calculate the root mean square error of the fitting curve based on the curvature differences in the curvature difference sequence and the corresponding curvature differences on the fitting curve; The inverse of the sum of the root mean square error and the first preset value is obtained, which is recorded as the first structural feature. The absolute value of the difference between the curvature difference of a suspected noise point in the connected domain and the curvature difference corresponding to the point on the fitting curve is obtained, and the difference is added to the first preset value and then inverted to obtain the second structural feature. The first structural feature and the second structural feature are multiplied and normalized to obtain the structural feature value of the suspected noise point in the connected domain.

[0011] Preferably, screening the suspected noise points based on the structural characteristic values ​​and curvature values ​​of each suspected noise point to obtain the real noise points includes: The curvature value of a suspected noise point and the inverse of the structural eigenvalue are multiplied and normalized to obtain the true noise possibility of the suspected noise point; if the true noise possibility of the suspected noise point is greater than a first preset threshold, the suspected noise point is a true noise point.

[0012] In a second aspect, the present invention also provides a system for constructing a three-dimensional model of precast concrete components, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of a method for constructing a three-dimensional model of precast concrete components.

[0013] The embodiments of the present invention have at least the following beneficial effects: the present application obtains point cloud data of a precast concrete component, then constructs a window centered on a point in the point cloud data, sorts the points in the window to obtain a sorted sequence, and then calculates the required level of detail for each point based on the sorted sequence corresponding to each point. This adaptively adjusts the K value of each point during curvature calculation (using the K-nearest neighbor method to obtain a set of neighboring points). While minimizing the interference of noise points on non-noise points, it also maximizes the preservation of features of detail locations such as hole edges and component corners, and avoids smoothing out noise points. This improves the accuracy of noise point identification and provides more accurate data for subsequent precast component structural analysis. After the suspected noise points are obtained through the initial screening, the adjacent suspected noise points are divided into the same connected domain. The suspected noise points in the connected domain are sorted using the minimum curvature path starting from a point on the edge of the connected domain to obtain the first sequence, and the structural characteristic values ​​of each suspected noise point in the connected domain are obtained. Based on the structural characteristic values ​​and curvature values ​​of each suspected noise point, the suspected noise points are screened to obtain the real noise points. The structural characteristics of the locations of the suspected noise points are combined to further distinguish the real noise points from the structural points on the components, so as to avoid the real structure on the prefabricated components being smoothed by denoising, retain more component feature information, and improve the accuracy of the denoising process. Finally, the real noise points in the point cloud data are denoised to obtain denoised point cloud data, thereby improving the accuracy of subsequent 3D modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 A flowchart of a method for constructing a three-dimensional model of a precast concrete component provided by an embodiment of the present invention; Figure 2 A schematic diagram of a window illustrating a method for constructing a three-dimensional model of a precast concrete component provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method and system for constructing a three-dimensional model of precast concrete components, including its specific implementation, structure, features, and effectiveness. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0017] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0018] The following describes in detail a method and system for constructing a three-dimensional model of a precast concrete component according to an embodiment of the present invention with reference to the accompanying drawings.

[0019] Example 1: The main application scenarios of this invention are: Laser scanning methods are often used in the scanning process of precast concrete components. When the surface of the precast concrete components is rough or there is jitter during the scanning process, the obtained point cloud data may contain noise points. Because the noise points and the holes and edges on the surface of the precast components have similar characteristics, the traditional denoising method will blur the information of the edges and holes of the precast components, making the constructed three-dimensional model less accurate. This application combines the structural feature differences between noise points and real precast component surface points to distinguish noise points from surface holes and edges, screen out the real noise points, and perform noise reduction processing to improve the accuracy of three-dimensional model construction.

[0020] See also Figure 1 , which shows a method flow chart of a method for constructing a three-dimensional model of a precast concrete component provided by an embodiment of the present invention, the method comprising the following steps: Step S1: Scan the precast concrete component to obtain point cloud data; construct a window with a point in the point cloud data as the center, and sort the points in the window to obtain a sorting sequence.

[0021] A laser scanner is used to scan precast concrete components from multiple angles. By receiving laser light reflected from the surface of the precast component and calculating the distance between each acquisition point and the scanner, the spatial coordinates of the acquisition points on the precast component are generated, and the original point cloud data of the precast component is obtained. Before scanning, the scanning accuracy parameters must be set according to the size of the precast component and the inspection requirements. Here, the angular resolution is set to 0.1° and the point spacing is set to 0.5mm based on experience. During the scanning process, ensure that there are no other objects blocking the scanning. The ground should be placed as flat as possible to reduce jitter during the scanning process. This results in the point cloud data of the precast concrete component.

[0022] Because noise points are caused by unstable variations in reflected laser light due to interference factors such as reflections and jitter, their coordinates vary significantly relative to the coordinates of other normal points. Therefore, curvature detection is used to first identify points with greater curvature—that is, points with greater variation relative to the surrounding surface—as suspected noise points. Furthermore, because precast concrete surfaces often contain structures such as holes and sharp corners, which also exhibit significant curvature, the suspected noise points may also contain holes and sharp corners. Direct denoising will obscure the structural information at these locations, necessitating filtering to identify the true noise points from the suspected noise points. Because holes and sharp corners are real precast component structures, meaning that the points at corresponding locations share certain structural characteristics, noise points are generated and manifested in a highly random manner, meaning they lack structural characteristics. Therefore, the structural characteristic values ​​of the suspected noise points are used to filter out the true noise points.

[0023] The curvature of a point in the collected point cloud data reflects the degree of geometric change relative to the surrounding points. The normal surface of a precast concrete component is relatively flat, so the curvature of the corresponding point is relatively low. Noise points usually appear to be floating in the air or drilling into the interior of the object, which is significantly deviated from the true surface of the precast component, so their corresponding curvature will be relatively large. Therefore, the degree of noise performance of each point is calculated based on the size of the curvature.

[0024] First, for each point, the KNN algorithm is used to obtain its corresponding set of neighboring points. Since the choice of K value directly affects the calculated curvature value, if the K value is too large, the curvature will be smoothed out, making it impossible to accurately identify the component's edges, corners, detailed structures, and noise points. If the K value is too small, the curvature will be severely affected by noise, and normal points near the noise point may be mistakenly identified as noise. Therefore, the K value needs to be adaptively adjusted based on the distribution characteristics of each point and its surrounding points.

[0025] Therefore, the initial K value is set to 30 based on the current scanning accuracy. It is further necessary to calculate the required level of detail for each point based on the severity of the change in the coordinates of the location of each point. Specifically, a window is constructed with a point as the center. The window is a three-dimensional window. The preferred size of the window is 9×9×9. If a point is at the corner or detail structure of a component, the coordinate change of the corresponding point in the window is relatively drastic, while the coordinates of the points in a flat position are usually continuous changes with a smaller amplitude, or only the coordinate values ​​in a certain direction change continuously. The construction of the window needs to be constructed in a unified manner. For example, the upper and lower surfaces and all sides of all windows are parallel to the three-dimensional rectangular coordinate system plane XOY, plane XOZ and plane YOZ, respectively, such as Figure 2 As shown in the figure, X, Y, and Z represent three coordinate axes respectively, O represents the coordinate origin, and the cube in the figure is a constructed window, whose upper and lower surfaces and side surfaces are parallel to the three-dimensional rectangular coordinate system plane XOY, plane XOZ, and plane YOZ respectively.

[0026] Therefore, it is necessary to sort the points in the window. Specifically, take one of the points in the window closest to the top vertex of the window as the starting point, calculate the distance between each point in the window except the starting point and the starting point, and take the point with the smallest distance as the second point; calculate the distance between the other points in the window except the starting point and the second point and the second point, and take the point with the smallest distance as the third point, and so on, until all points in the window corresponding to the point are traversed, the points in the window are sorted, and the sorting sequence corresponding to the window is obtained, that is, the sorting sequence corresponding to the point.

[0027] It should be noted that previously sorted points are not included in the subsequent sorting, that is, they are not placed back into the sorting. In addition, if a point has multiple points with the smallest distance, the one with the smallest distance is selected. After obtaining the sorting sequence corresponding to each point, the severity of the changes around each point is analyzed based on the sorting sequence.

[0028] Step S2: Obtain the local variation intensity of a point based on the coordinate difference between every two adjacent points in the sorting sequence corresponding to the point; and calculate the required detail level of the point based on the local variation intensity of the point and the adjacent points before and after the point in the sorting sequence corresponding to the point.

[0029] In step S1, the sorting sequence corresponding to each point in the point cloud data is obtained, and then the degree of local point change of each point in the sequence is analyzed. The degree of local change of a point is obtained based on the coordinate difference between each two adjacent points in the sorting sequence corresponding to the point.

[0030] Specifically, the sum of the absolute values ​​of the differences between the coordinates of two adjacent points in the arrangement sequence corresponding to a point is obtained, and recorded as the coordinate difference of the two adjacent points; the mean of the coordinate differences between every two adjacent points in the arrangement sequence corresponding to the point is calculated to obtain the severity of the local change of the point.

[0031] The calculation model of the severity of local changes is as follows: , in, Indicates the intensity of the local change corresponding to the i-th point, that is, the intensity of the change of the point coordinates in the window corresponding to the i-th point; Indicates the number of points contained in the window corresponding to the i-th point, that is, the number of points in the permutation sequence; Represents the coordinates of the jth point in the permutation sequence corresponding to the i-th point, The coordinates of the j+1th point in the permutation sequence corresponding to the i-th point, Indicates the absolute value of the difference between the coordinates of two adjacent points, that is, the X-axis coordinates. The larger the value, the more drastic the change in the coordinates of the points in the window, that is, the stronger the local change. It represents the coordinate difference between two adjacent points in the sorted sequence, that is, the coordinate difference between the jth point and the j+1th point in the sorted sequence corresponding to point i. Because the number of points contained in the window corresponding to each point may be different, the average value of the coordinate change of adjacent points in the window is used to represent the overall coordinate change intensity of the window. In this way, the local change intensity of each point in the point cloud data can be obtained.

[0032] Since the coordinate changes of other points in the corresponding window of a single noise point on a relatively flat surface of a component may be less drastic, in order to avoid smoothing the curvature of the noise point when calculating the curvature, the corresponding K value should also be reduced. Therefore, the required level of detail of the point is calculated based on the distance between the current analysis point and its adjacent points.

[0033] Specifically, the sum of the distances between a point and its adjacent points in the corresponding sorting sequence is obtained, multiplied by the local variation intensity of the point, and then normalized to obtain the required detail level of the point.

[0034] The calculation model for the required level of detail is as follows: , in, represents the required level of detail of the i-th point, Indicates the severity of the local change corresponding to the i-th point, and They represent the distance between the i-th point and the previous adjacent point and the distance between the i-th point and the next adjacent point in its corresponding sorted sequence. The larger the sum of these two distances, the greater the degree of mutation of the i-th point relative to its adjacent points, that is, it may be noise data. In order to avoid being smoothed out, a more accurate curvature value of the point needs to be obtained; norm represents the normalization function.

[0035] In this way, the required level of detail for each point can be obtained. When using the KNN algorithm to obtain the set of neighboring points corresponding to each point, the greater the required level of detail of the point, the smaller the corresponding K value should be.

[0036] Step S3, determining the K value of each point according to the required detail level of each point, and obtaining the neighboring point set of each point respectively; calculating the curvature value of each point according to the neighboring point set of each point, and obtaining the suspected noise point.

[0037] When the local change of a point is more drastic, the distance between it and the adjacent points in the corresponding sorting sequence is larger, which means that it needs more attention and more attention to its details. Therefore, when using the KNN algorithm to obtain the set of neighboring points of the point, the corresponding K value is smaller.

[0038] Furthermore, the K value of each point is determined according to the required level of detail of each point. Specifically, the K value of the point is obtained by multiplying the set initial K value by the difference between the first preset value and the required level of detail of a point and rounding up.

[0039] The calculation model of the K value of each point is: , in, represents the K value of the i-th point, Indicates the set initial K value, Indicates the required level of detail of the i-th point, the first preset value is 1, The higher the required level of detail, the smaller the corresponding K value should be. ⌈ ⌉ indicates the rounding symbol. In addition, since the K value cannot be infinitely small, the minimum K value is set to 10. That is, when the calculated value is When setting .

[0040] Then, the K value of each point is combined with the KNN algorithm to obtain the neighboring point set of each point.

[0041] After obtaining the neighboring point set of each point, the covariance matrix is ​​calculated based on the neighboring point set of a point, and the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the normal eigenvalue, the first eigenvalue, and the second eigenvalue; the normal eigenvalue is divided by the sum of the normal eigenvalue, the first eigenvalue, and the second eigenvalue to obtain the curvature value of the point.

[0042] The calculation model of the curvature value is: , in, represents the curvature value of the i-th point, 、 and are the normal eigenvalue, first eigenvalue and second eigenvalue corresponding to the i-th point respectively. These three eigenvalues ​​represent the spatial distribution of the point in different main directions. Corresponds to the principal normal direction, which is the minimum direction.

[0043] The curvature values ​​of noise points are higher than those of the collected points. Since most locations on the surface of precast concrete components are relatively flat, the corresponding curvature values ​​are small and close together. Therefore, the DBSCAN clustering algorithm is used to cluster the curvature values ​​of each point to obtain different clusters. The maximum curvature value in the largest cluster is selected as the judgment threshold. That is, when the curvature value of a point is greater than the judgment threshold, it is considered a suspected noise point. This results in the suspected noise points in the point cloud data.

[0044] Step S4: divide adjacent suspected noise points into the same connected domain, sort the suspected noise points in the connected domain using a minimum curvature path starting from a point on the edge of the connected domain to obtain a first sequence; and obtain a structural feature value of each suspected noise point in the connected domain based on the first sequence.

[0045] Because the curvature values ​​of corners, hole edges, and other locations on prefabricated components are also higher than other flat locations, the suspected noise points obtained above may also include points at corners, hole edges, and other locations. These points reflect the size, shape, and quality characteristics of the prefabricated components. Therefore, when denoising, the characteristics of these points should be preserved as much as possible, that is, it is necessary to distinguish between noise points and points of the actual structure of these prefabricated components. Although the curvature of corners, hole edges, and other locations is higher than other planar locations, because these points are the actual structure of the prefabricated components, there is a strong structural relationship between these points and the surrounding points. Noise points are highly random and do not have structural characteristics. Therefore, based on the structural characteristic values ​​of the suspected noise points, noise points and non-noise points are distinguished.

[0046] First, adjacent suspected noise points are divided into the same connected domain. Because the curvature of the edge points of prefabricated components changes continuously, while the curvature values ​​of noise points are disorderly and irregular, the structurality of each connected domain is calculated based on the difference in curvature changes between adjacent points in the connected domain.

[0047] Furthermore, starting from a point on the edge of the connected domain, an edge trend curve is generated using the path with the minimum curvature gradient, and the points on the connected domain are sorted based on this curve. The path with the minimum curvature gradient refers to the path with the smallest curvature change, connecting points on the component surface with continuously changing curvature. This results in a sorted sequence of suspected noise points in the connected domain, which is also known as the first sequence. It should be noted that this part is existing technology and will not be elaborated on in detail here.

[0048] Thus, the first sequence corresponding to each connected domain composed of adjacent suspected noise points can be obtained. Furthermore, the difference between the curvature value of the latter suspected noise point and the curvature value of the former suspected noise point in the first sequence corresponding to a connected domain is calculated, and recorded as the curvature difference of the latter suspected noise point; the curvature difference of each two adjacent suspected noise points is sorted according to the order of the first sequence to obtain a curvature difference sequence; the curvature difference sequence is linearly fitted using the least squares method to obtain a fitting curve; and the root mean square error of the fitting curve is calculated based on the curvature difference in the curvature difference sequence and the corresponding curvature difference on the fitting curve.

[0049] If a suspected noise point is adjacent to an edge point, it will be divided into the same connected domain. Because the other points are more structural, the randomness of the suspected noise point may be weakened. Therefore, the structural characteristic value of each suspected noise point is calculated by combining the difference between the curvature difference corresponding to each suspected noise point and the corresponding point on the fitting curve.

[0050] Specifically, the reciprocal of the sum of the root mean square error and a first preset value is calculated and recorded as the first structural feature. The absolute value of the difference between the curvature difference value of a suspected noise point in the connected domain and the curvature difference value corresponding to the point on the fitting curve is calculated, added to the first preset value, and then inverted to obtain the second structural feature. The first structural feature and the second structural feature are multiplied and normalized to obtain the structural feature value of the suspected noise point in the connected domain. The curvature difference value corresponding to the first point in the first sequence is the curvature difference value of the second point in the first sequence.

[0051] The calculation model of structural eigenvalue is as follows: , in, represents the structural eigenvalue of the i-th suspected noise point, represents the connected domain where the i-th suspected noise point is located, It represents the root mean square error of the fitting curve after fitting the curvature difference sequence corresponding to the connected domain where the i-th suspected noise point is located. is the first structural feature, indicating that the smaller the RMS error value corresponding to the connected domain where the i-th suspected noise point is located, the stronger the structure of the connected domain, that is, the stronger the structure of the suspected noise point. It is added to the first preset value 1 to ensure that the fraction is meaningful. represents the curvature difference of the i-th suspected noise point, Represents the curvature difference of the i-th suspected noise point on the corresponding fitting curve, The second structural feature indicates that the closer the actual curvature difference of the point is to the fitted value, the stronger the structure of the point is; norm represents the normalization operation. In this way, the structural feature value of each point can be obtained.

[0052] Step S5: screening the suspected noise points based on the structural characteristic value and curvature value of each suspected noise point to obtain the real noise points; denoising the real noise points in the point cloud data to obtain denoised point cloud data; and constructing a three-dimensional model using the denoised point cloud data.

[0053] In step S4, the structural characteristics of the suspected noise points are analyzed based on the structure of the prefabricated part, and the structural characteristic value of each suspected noise point is obtained. Because the curvature value of noise points is often higher than that of edge points and their structure is poor, the curvature value and structural characteristic value of the suspected noise point are combined to calculate the probability that the suspected noise point is a real noise point.

[0054] The suspected noise points are screened based on their structural eigenvalues ​​and curvature values ​​to obtain the true noise points. Specifically, the curvature value of a suspected noise point is multiplied by the inverse of the structural eigenvalue and the normalized value to obtain the true noise probability of the suspected noise point.

[0055] The larger the curvature value and the smaller the structural characteristic value, the more likely the suspected noise point is to be a real noise point. Because noise points and edge points have significant structural differences, their real noise likelihoods differ significantly. Therefore, we can set the median of the real noise likelihood range, 0.5, as the first preset threshold. If the real noise likelihood of a suspected noise point is greater than the first preset threshold, the suspected noise point is considered a real noise point.

[0056] After obtaining the actual noise points from the point cloud data using the aforementioned steps, we used the local surface projection (MLS) method to smooth the noise points. To avoid issues like structural incompleteness that might result from simply removing them, we used a search radius of 3 mm and a fitted surface order of 2. The 3D model of the precast concrete component was constructed based on the denoised point cloud data.

[0057] In summary, this application uses a curvature detection method to first identify suspected noise points in the data. Then, based on the fact that holes and corners on the surface of precast components have certain structural characteristics, while noise points are random and irregular, the structural characteristics of the suspected noise points are analyzed. Finally, the noise level and structural characteristic values ​​of the suspected noise points are combined to identify the actual noise points and perform denoising. This method can effectively solve the problem of processing noise data in point cloud data of precast concrete components and improve the accuracy of constructing 3D models of precast concrete components.

[0058] Example 2: This embodiment provides a system for constructing a three-dimensional model of a precast concrete component, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of a method for constructing a three-dimensional model of a precast concrete component. Since the method for constructing a three-dimensional model of a precast concrete component has already been described in detail in Example 1, it will not be further elaborated here.

[0059] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing a three-dimensional model of a precast concrete component, characterized in that: The method includes: Scan the precast concrete components to obtain point cloud data; construct a window with a point in the point cloud data as the center, and sort the points in the window to obtain a sorting sequence; The local variation intensity of a point is obtained based on the coordinate difference between each two adjacent points in the sorted sequence corresponding to the point; the required level of detail of the point is calculated based on the local variation intensity of the point and the adjacent points before and after it in the sorted sequence; Determine the K value of each point according to the required level of detail of each point, and obtain the neighboring point set of each point respectively; calculate the curvature value of each point according to the neighboring point set of each point, and obtain the suspected noise point; Divide adjacent suspected noise points into the same connected domain, sort the suspected noise points in the connected domain using a minimum curvature path starting from a point on the edge of the connected domain to obtain a first sequence; obtain the structural characteristic value of each suspected noise point in the connected domain based on the first sequence; Based on the structural characteristic values ​​and curvature values ​​of each suspected noise point, the suspected noise points are screened to obtain the real noise points; the real noise points in the point cloud data are denoised to obtain the denoised point cloud data; and the denoised point cloud data is used to construct a three-dimensional model.

2. A method for constructing a three-dimensional model of a precast concrete component according to claim 1, characterized in that: Sorting the points in the window to obtain a sorted sequence includes: Take one of the points in the window corresponding to a point that is closest to the top vertex of the window as the starting point, calculate the distance between each point in the window except the starting point and the starting point, and take the point with the smallest distance as the second point; calculate the distance between the other points in the window except the starting point and the second point and the second point, and take the point with the smallest distance as the third point, and so on, until all points in the window corresponding to the point are traversed, the points in the window are sorted, and the sorting sequence corresponding to the point is obtained.

3. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The step of obtaining the severity of the local change of a point based on the coordinate difference between every two adjacent points in the sorting sequence corresponding to the point includes: Obtain the sum of the absolute values ​​of the differences between the coordinates of two adjacent points in the sequence corresponding to a point, and record it as the coordinate difference of the two adjacent points; calculate the mean of the coordinate differences between every two adjacent points in the sequence corresponding to the point to obtain the severity of the local change of the point.

4. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The step of calculating the required level of detail of a point based on the severity of the local change of the point and the adjacent points in the corresponding sorting sequence includes: The required level of detail of a point is obtained by obtaining the sum of the distances between the adjacent points before and after a point in its corresponding sorting sequence, multiplying it by the severity of the local change of the point, and then normalizing it.

5. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The step of determining the K value of each point according to the required level of detail of each point and obtaining a set of neighboring points of each point respectively includes: The set initial K value is multiplied by the difference between the first preset value and the required level of detail of a point and rounded up to obtain the K value of the point; then the K value of each point is combined with the KNN algorithm to obtain the neighboring point set of each point.

6. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The step of calculating the curvature value of each point based on the neighboring point set of each point and obtaining the suspected noise point includes: The covariance matrix is ​​calculated based on the set of neighboring points of a point, and the eigenvalue decomposition of the covariance matrix is ​​performed to obtain the normal eigenvalue, the first eigenvalue and the second eigenvalue; the normal eigenvalue is divided by the sum of the normal eigenvalue, the first eigenvalue and the second eigenvalue to obtain the curvature value of the point; the curvature value of each point is clustered to obtain different clusters, and the maximum curvature value in the largest cluster is selected as the judgment threshold. When the curvature value of a point is greater than the judgment threshold, the point is a suspected noise point.

7. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The obtaining of the structural characteristic value of each suspected noise point in the connected domain based on the first sequence includes: Calculate the difference between the curvature value of the latter suspected noise point and the curvature value of the former suspected noise point in the first sequence corresponding to a connected domain, and record it as the curvature difference of the latter suspected noise point; sort the curvature differences of every two adjacent suspected noise points according to the order of the first sequence to obtain a curvature difference sequence; use the least squares method to perform linear fitting on the curvature difference sequence to obtain a fitting curve; calculate the root mean square error of the fitting curve based on the curvature differences in the curvature difference sequence and the corresponding curvature differences on the fitting curve; The inverse of the sum of the root mean square error and the first preset value is obtained, which is recorded as the first structural feature. The absolute value of the difference between the curvature difference of a suspected noise point in the connected domain and the curvature difference corresponding to the point on the fitting curve is obtained, and the difference is added to the first preset value and then inverted to obtain the second structural feature. The first structural feature and the second structural feature are multiplied and normalized to obtain the structural feature value of the suspected noise point in the connected domain.

8. The method for constructing a three-dimensional model of a precast concrete component according to claim 1, wherein: The method of screening the suspected noise points based on the structural characteristic value and curvature value of each suspected noise point to obtain the real noise point includes: The curvature value of a suspected noise point and the inverse of the structural eigenvalue are multiplied and normalized to obtain the true noise possibility of the suspected noise point; if the true noise possibility of the suspected noise point is greater than a first preset threshold, the suspected noise point is a true noise point.

9. A system for constructing a three-dimensional model of a precast concrete component, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by a processor, the steps of the method for constructing a three-dimensional model of a precast concrete component are implemented as described in any one of claims 1 to 8.

Images