A digital construction engineering simulation system based on point cloud data

Abstract

The present invention relates to the technical field of point cloud data registration, and specifically to a digital building engineering simulation system based on point cloud data, comprising: a building point cloud data acquisition module, an auxiliary registration point acquisition module, and a registration result acquisition module; used to acquire a building point cloud data group, downsample to obtain the optimal downsampling result of the building point cloud data group, obtain a number of detection points of the building point cloud data group under the optimal downsampling result, obtain the building rigidity performance coefficient of each detection point of the building point cloud data group under each downsampling scale, and then obtain the building structure of each detection point in the building point cloud data group, obtain a number of auxiliary registration points in the building point cloud data group according to the building structure, and obtain the registration results of all building group point cloud data. The present invention aims to solve the problem of initial point cloud data clutter in point cloud data registration, and achieve the purpose of improving the registration accuracy by replacing it with auxiliary registration points that reflect the building structure.

Description

Technical Field

[0001] The invention relates to the technical field of point cloud data registration, and in particular to a digital construction engineering simulation system based on point cloud data. Background Art

[0002] In the field of digital construction engineering simulation, the three-dimensional point cloud data of buildings is currently mainly collected to construct digital models of buildings or foundation designs for analysis, design, cost estimation and construction progress management of construction projects. Each point cloud data contains the three-dimensional point coordinates of its spatial position, and several point cloud data constitute the surface information of the building. However, due to the limitations of the collection angle and collection range, the point cloud data collected each time only contains the three-dimensional point cloud data of the local area or local surface of the building. Therefore, it is necessary to align the point cloud data collected multiple times.

[0003] ICP (Iterative Closest Point) algorithm is a commonly used point cloud registration method, mainly used to align two sets of point cloud data. Its core idea is to iteratively minimize the distance error between two sets of point cloud data, so as to find their optimal transformation matrix, which contains the translation and rotation information between the two sets of point clouds. The two sets of point cloud data can be registered through the optimal transformation matrix; however, in the registration process using point cloud data, it is necessary to obtain another point cloud data with the closest distance in another set for each point cloud data. This process is easily affected by the quality of noise in the original point cloud data. For example, the position adjustment of the equipment caused by the occlusion caused by the change of the construction site environment and the different scanning angles will cause different batches of point cloud data to have obvious differences in resolution and morphology. That is, in addition to the point cloud data reflecting the rigidity of the building, the original point cloud data also contains point cloud data interfered by non-rigid noise, which reduces the accuracy of the ICP algorithm registration result and increases the calculation amount of the registration process. Summary of the invention

[0004] The invention provides a digital construction engineering simulation system based on point cloud data to solve the existing problems.

[0005] A digital construction engineering simulation system based on point cloud data of the present invention adopts the following technical solution:

[0006] An embodiment of the present invention provides a digital construction engineering simulation system based on point cloud data, the system comprising the following modules:

[0007] A building point cloud data acquisition module 101 is used to acquire a plurality of building point cloud data groups of the construction building, wherein the point cloud data in the building point cloud data group includes a horizontal axis, a vertical axis and a Z axis;

[0008] An auxiliary registration point acquisition module is used to iterate based on the difference in local spatial point cloud density of each point cloud data in the downsampling results of the two groups of building point cloud data groups to obtain the optimal downsampling results of the two groups of building point cloud data groups;

[0009] According to the building rigidity trend of each point cloud data in the optimal downsampling results of the two building point cloud data groups, the directional registration positions of the two building point cloud data groups are obtained;

[0010] Based on the optimal downsampling result of each group of building point cloud data, a number of detection points of each group of building point cloud data under the optimal downsampling result are obtained;

[0011] According to the building rigidity presentation of each detection point of the two building point cloud data groups at each downsampling scale, the building rigidity presentation coefficient of each detection point of each building point cloud data group at each downsampling scale is obtained;

[0012] According to the building rigidity performance coefficient of each detection point in each building point cloud data set at each downsampling scale and the local space point cloud density, the building structurality of each detection point in the two building point cloud data sets is obtained;

[0013] According to the architectural structure of each detection point in the two sets of building point cloud data sets, all preferred detection points in the two sets of building point cloud data sets are selected, and based on the directional registration positions of the two sets of building point cloud data sets and the positions of the preferred detection points, several auxiliary registration points in the two sets of building point cloud data sets are obtained;

[0014] The registration result acquisition module is used to register the auxiliary registration points of all groups of building point cloud data using the registration algorithm to obtain the registration results of all building point cloud data.

[0015] Preferably, the specific steps of obtaining the optimal downsampling result include:

[0016] Preset the three-dimensional voxel grid scale, use the voxel grid filtering algorithm to perform the first downsampling on the K-th group of building point cloud data and the K+1-th group of building point cloud data, and obtain the first downsampling results of the K-th group of building point cloud data and the K+1-th group of building point cloud data;

[0017] Obtaining downsampling supervision conditions of the Kth group of building point cloud data and the K+1th group of building point cloud data in the first downsampling result;

[0018] The downsampling supervision conditions of the first downsampling results of the K-th group of building point cloud data and the K+1-th group of building point cloud data are normalized using a sigmoid function to obtain a normalized result of the downsampling supervision conditions of the first downsampling results;

[0019] A downsampling supervision threshold is preset. When the normalized result of the downsampling supervision condition of the first downsampling result is greater than or equal to the downsampling supervision threshold, the first downsampling result of the two groups of building point cloud data is used as the downsampled original data, and the downsampled original data is used for the second downsampling to obtain the normalized result of the downsampling supervision condition of the second downsampling result, and the judgment is made according to the downsampling supervision threshold; and so on, until there is a constant So that the Kth and K+1th building point cloud data groups are in the When the normalized result of the downsampling supervision condition after the downsampling result is less than the downsampling supervision threshold, it means that the point cloud density of the point cloud data in the two groups of building point cloud data sets has reached the minimum difference, or has reached the threshold of the number of downsampling times, the first downsampling supervision condition of the two groups of building point cloud data sets is reduced to the minimum. The downsampling results are recorded as the optimal downsampling results of the two groups of building point cloud data sets.

[0020] Preferably, the specific method of obtaining the downsampling supervision condition is:

[0021] Downsampling supervision conditions of the Kth group of building point cloud data and the K+1th group of building point cloud data in the first downsampling result The specific calculation method is:

[0022] ;

[0023] in, is the number of point cloud data of the Kth group of building point cloud data after the first downsampling, is the point cloud density of the i-th point cloud data in the K-th group of building point cloud data after the first downsampling; is the K+1th group of building point cloud data. The point cloud density of the point cloud data after the first downsampling.

[0024] Preferably, the specific method of obtaining the point cloud density is:

[0025] In the first downsampling results of the Kth group of building point cloud data groups and the K+1th group of building point cloud data groups, for the i-th point cloud data in the Kth group of building point cloud data groups, the point cloud data with the closest spatial Euclidean distance to the corresponding coordinate position of the i-th point cloud data in the Kth group of building point cloud data groups is selected in the K+1th group of building point cloud data groups, and is recorded as the i-th point cloud data in the K+1th group of building point cloud data groups. Point cloud data;

[0026] The default number of nearest neighbors is , the point cloud density of the i-th point cloud data in the K-th group of building point cloud data after the first downsampling The specific calculation method is:

[0027] ;

[0028] in, is the mean of the spatial Euclidean distances between the i-th point cloud data and each nearest neighbor in the K-th group of building point cloud data after the first downsampling.

[0029] Preferably, the specific steps of acquiring the direction registration position include:

[0030] The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the K-th group of building point cloud data is projected to the two-dimensional coordinate system of the horizontal plane to obtain the horizontal plane principal component direction projection of the K-th group of building point cloud data under the optimal downsampling result, and the counterclockwise angle between the horizontal plane principal component direction projection and the horizontal axis is recorded as the horizontal plane registration direction of the K-th group of building point cloud data. ;

[0031] The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the Kth group of building point cloud data set is projected to the two-dimensional coordinate system of the vertical axis and the Z axis to obtain the elevation principal component direction projection of the Kth group of building point cloud data set under the optimal downsampling result, and the counterclockwise angle between the elevation principal component direction projection and the vertical axis is recorded as the elevation registration direction of the Kth group of building point cloud data set ;

[0032] Combine the horizontal registration direction and the elevation registration direction of the Kth group of building point cloud data sets , recorded as the direction registration position of the Kth group of building point cloud data .

[0033] Preferably, the specific steps of obtaining the principal component direction in the optimal downsampling result include:

[0034] Projecting the point cloud data in the optimal downsampling result of the Kth group of building point cloud data sets in a three-dimensional Cartesian coordinate system to obtain a coordinate projection of the optimal downsampling result of the Kth group of building point cloud data sets, wherein the horizontal axis and the vertical axis are respectively the horizontal plane coordinates of the point cloud data, and the Z axis is the elevation coordinate of the point cloud data;

[0035] The principal component analysis algorithm is used to reduce the dimension of the coordinate projection of the optimal downsampling result of the K-th group of building point cloud data, and several principal component directions of the K-th group of building point cloud data under the optimal downsampling result are obtained.

[0036] Preferably, the specific method of obtaining the detection points is:

[0037] Taking the preset three-dimensional voxel grid scale as the division scale of the detection points, the optimal downsampling result of each group of building point cloud data groups is equally divided to obtain a number of voxel blocks. In the optimal downsampling result of each group of building point cloud data groups, the point cloud data with the closest Euclidean distance to each vertex of each voxel block is selected as the detection point of the optimal downsampling result of each group of building point cloud data groups.

[0038] Preferably, the specific steps of obtaining the building rigidity performance coefficient include:

[0039] The preset three-dimensional voxel grid scale is used as the spatial neighborhood range of each detection point under the optimal downsampling result, and the principal component analysis algorithm is used to reduce the dimension of all point cloud data within the spatial neighborhood range of each detection point under the optimal downsampling result to obtain several principal component directions of each detection point under the optimal downsampling result;

[0040] The principal component directions of each detection point under the optimal downsampling result are used as the Several principal component directions under downsampling scales;

[0041] Each detection point in each building point cloud data set is The difference between the eigenvalue corresponding to the direction of the largest principal component under the downsampling scale and the eigenvalue corresponding to the direction of the second largest principal component is recorded as the value of each detection point in each group of building point cloud data set. The building rigidity performance coefficient under the downsampling scale;

[0042] Each detection point in each group of building point cloud data is reversely downsampled according to the preset three-dimensional voxel grid scale to obtain the building rigidity performance coefficient of each detection point in each group of building point cloud data from the 0th downsampling scale to the cth downsampling scale.

[0043] Preferably, the specific steps of obtaining the building structure include:

[0044] The building structure of the vth detection point in the Kth group of building point cloud data The calculation method is:

[0045] ;

[0046] in, Indicates that the vth detection point in the Kth group of building point cloud data is The building rigidity performance coefficient under the downsampling scale is represents the variance of the building rigidity performance coefficient of the vth detection point in the Kth group of building point cloud data under all downsampling scales; Indicates that the vth detection point in the Kth group of building point cloud data is Point cloud density at downsampling scale, Represents the variance of the point cloud density of the vth detection point in the Kth group of building point cloud data at all downsampling scales; It represents the maximum downsampling scale of the Kth group and the K+1th group of building point cloud data, and the maximum downsampling scale corresponds to the optimal downsampling result of the two groups of building point cloud data; norm represents the normalization function.

[0047] Preferably, the specific steps of acquiring the auxiliary registration points include:

[0048] A building structure judgment threshold is preset, and when the building structure of the vth detection point in the Kth group of building point cloud data is greater than or equal to the building structure judgment threshold, the vth detection point is recorded as a preferred detection point;

[0049] Obtain all the preferred detection points in the K-th group of building point cloud data and all the preferred detection points in the K+1-th group of building point cloud data;

[0050] The Euclidean distance between each preferred detection point and the origin in the Kth group of building point cloud data is recorded as the vector registration position of each preferred detection point in the Kth group of building point cloud data, the vector registration position is used as the vector modulus, the direction registration position of the Kth group of building point cloud data is used as the vector direction, and the constructed vector is recorded as an auxiliary registration in the Kth group of building point cloud data.

[0051] The beneficial effects of the technical solution of the present invention are as follows: after acquiring several groups of building point cloud data groups, the present invention downsamples two groups of building point cloud data groups to the same point cloud density, thereby avoiding the registration error problem caused by the difference in point cloud data density during the registration process, and obtaining the directional registration position of the building point cloud data group based on the rigid performance of the building structure in the point cloud data in the optimal downsampling result obtained by downsampling, which is used to represent the main structural direction of the building in the group of building point cloud data groups; and then selecting several detection points based on the optimal downsampling result, and selecting the directional registration position of the building point cloud data group based on the structural directionality of other point cloud data in the spatial neighborhood of the detection point during the reverse downsampling process as the downsampling increases. The architectural structure of each detection point is obtained by changing the characteristics, and is used to screen out the detection points that show the strong structural rigidity characteristics of the building as a whole, so as to avoid the registration error caused by the detailed structure of the building. Then, a number of preferred detection points are selected by using the architectural structure. According to the orientation registration position of the preferred detection points and the building point cloud data group, a number of auxiliary registration points of each building point cloud data group are obtained for the registration of the point cloud data; by quantifying the overall direction and structural rigidity characteristics of the building, a number of auxiliary registration points are selected from the building point cloud data to replace the original point cloud data for registration, so as to improve the registration accuracy and reduce the amount of registration iterative calculation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 creative work.

[0053] Figure 1 This is a structural block diagram of a digital construction engineering simulation system based on point cloud data of the present invention. DETAILED DESCRIPTION

[0054] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a digital construction engineering simulation system based on point cloud data proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0055] 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.

[0056] The following is a detailed description of a specific scheme of a digital construction engineering simulation system based on point cloud data provided by the present invention in conjunction with the accompanying drawings.

[0057] See also Figure 1 , which shows a structural block diagram of a digital building engineering simulation system based on point cloud data provided by an embodiment of the present invention, the system includes the following modules:

[0058] The building point cloud data acquisition module is used to acquire several groups of building point cloud data groups of the construction building.

[0059] It should be noted that the purpose of this embodiment is to register the construction point cloud data collected multiple times, so as to build a digital building engineering model according to the registration results. Therefore, it is necessary to collect the construction point cloud data first. The specific process of collecting the building point cloud data and preprocessing it to obtain the building point cloud data is as follows:

[0060] 1. Configure acquisition equipment: This embodiment uses laser radar equipment to collect point cloud data, specifically using Velodyne laser radar to scan the environment of the construction building. This embodiment sets the following parameters for the laser radar: ranging range: 50 meters to 200 meters; scanning frequency: 15Hz; angle resolution: 0.1 degrees; vertical field of view: 60 degrees in the vertical direction. It should be noted that this embodiment does not specifically limit the laser radar parameters set, and other embodiments may set other values.

[0061] 2. Install the LiDAR on the vehicle-mounted system platform and calibrate the position of the LiDAR so that the position and posture of the LiDAR on the vehicle-mounted system platform remain consistent each time the LiDAR scans.

[0062] 3. Select several time nodes for scanning, and select several fixed geographical locations at different time nodes in a day to scan once each. The fixed geographical locations are locations where all point cloud data of the construction building can be scanned. The time nodes are to avoid interference to the point cloud data caused by the singleness of rain, fog, and ambient light in the environment at the same time node. In this embodiment, 9 o'clock, 13 o'clock, and 18 o'clock in a day are used as time nodes.

[0063] 4. After the collection preparation is performed according to the above steps, the laser radar is used to collect the point cloud data of all geographical locations at each time node. The point cloud data collected each time is used as a group of initial point cloud data. All the collected groups of point cloud data are filtered using the Radius Outlier Filter algorithm to obtain all groups of building point cloud data groups. It should be noted that the radius outlier filter algorithm described in this embodiment is an existing well-known technology. The radius distance set in this embodiment is 0.5 meters, and the threshold of the number of point cloud data within the radius is 4. When the point cloud data within the radius cluster of the point cloud data is less than the number threshold, the point cloud data is filtered and removed as a noise point.

[0064] The auxiliary registration point acquisition module 102 is used to iterate based on the difference in local spatial point cloud density of each point cloud data in the downsampling results of the two groups of building point cloud data groups to obtain the optimal downsampling results of the two groups of building point cloud data groups; obtain the directional registration points of the two groups of building point cloud data groups according to the architectural rigidity trend of each point cloud data in the optimal downsampling results of the two groups of building point cloud data groups; obtain a number of detection points of each group of building point cloud data groups under the optimal downsampling results based on the optimal downsampling results of each group of building point cloud data groups; obtain the architectural rigidity trend of each detection point of the two groups of building point cloud data groups under each downsampling scale according to the architectural rigidity trend of each point cloud data in the optimal downsampling results of the two groups ...; obtain the architectural rigidity trend of each detection point of the two groups of building point cloud data groups under each downsampling scale; obtain the directional registration points of the two groups of building point cloud data groups according to the architectural rigidity trend of each point cloud data in the optimal downsampling results; obtain the directional registration points of the two groups of building point cloud data groups according to the architectural rigidity trend of each point cloud data in the According to the rigid presentation situation, the building rigidity expression coefficient of each detection point in each group of building point cloud data sets at each downsampling scale is obtained; according to the building rigidity expression coefficient of each detection point in each group of building point cloud data sets at each downsampling scale and the local space point cloud density, the building structural property of each detection point in the two groups of building point cloud data sets is obtained; according to the building structural property of each detection point in the two groups of building point cloud data sets, all the preferred detection points in the two groups of building point cloud data sets are screened, and based on the directional registration positions of the two groups of building point cloud data sets and the positions of the preferred detection points, several auxiliary registration points in the two groups of building point cloud data sets are obtained.

[0065] Preferably, based on the difference in local spatial point cloud density of each point cloud data in the downsampling results of the two groups of building point cloud data groups, iteration is performed to obtain the optimal downsampling results of the two groups of building point cloud data groups. The specific steps for obtaining the optimal downsampling results are as follows:

[0066] It should be noted that when modeling building point cloud data, the key disadvantage is that the differences and influences of the environment at different time nodes lead to a large amount of invalid point cloud data in the building point cloud data groups collected at different time nodes. The ICP alignment is based on the closest points in different building point cloud data groups as the initial matching point cloud data for analysis. Too much invalid point cloud data is easily used as the initial point cloud data for alignment, which leads to a large amount of calculation for adjusting the alignment and reduces the accuracy of the alignment results.

[0067] It should be further explained that the ICP registration algorithm has extremely high accuracy when performing rigid registration, and in the building point cloud data, the point cloud data belonging to the building except the invalid point cloud data can show extremely strong rigid features, which are mainly manifested in the structural performance of the flatness and angles reflected by the walls and the ground. When the resolution of the point cloud data is reduced, the invalid point cloud data will be erased, while the rigid features of the building will be retained. In order to avoid the registration error caused by the difference in point cloud density between the two groups of building point cloud data groups, it is first necessary to downsample the two groups of building point cloud data groups to the same point cloud density, and the density should be as small as possible when the point cloud density is the same to reduce the interference of invalid point cloud data; therefore, this embodiment obtains the optimal downsampling result of the two groups of building point cloud data groups to be registered by downsampling them to the same point cloud density.

[0068] Preferably, as an example, in order to further reduce the impact of environmental interference at different time nodes, this embodiment takes any two groups of building point cloud data groups with adjacent collected time nodes as examples for analysis, and the alignment methods of other groups of building point cloud data groups with adjacent collected time nodes are the same.

[0069] Furthermore, the voxel grid filtering algorithm divides the point cloud data into multiple three-dimensional voxel grids, and only retains one representative point in each voxel grid to replace multiple point cloud data in the original three-dimensional voxel grid. The preset three-dimensional voxel grid scale is 0.05 meters, and the voxel grid filtering algorithm is used to perform the first downsampling on the two groups of building point cloud data to obtain the first downsampling results of the two groups of building point cloud data.

[0070] Preferably, in the first downsampling results of the Kth group of building point cloud data groups and the K+1th group of building point cloud data groups, for the i-th point cloud data in the Kth group of building point cloud data groups, point cloud data with the closest spatial Euclidean distance to the corresponding coordinate position of the i-th point cloud data in the Kth group of building point cloud data groups is selected in the K+1th group of building point cloud data groups as the registration point cloud data of the i-th point cloud data in the Kth group of building point cloud data groups in the K+1th group of building point cloud data groups, and is recorded as the registration point cloud data of the i-th point cloud data in the Kth group of building point cloud data groups. Point cloud data.

[0071] Furthermore, the K-Nearest Neighbors (KNN) algorithm is used to calculate the point cloud density of the i-th point cloud data in the K-th building point cloud data group and the i-th point cloud data in the K+1-th building point cloud data group. The point cloud density of point cloud data is as follows:

[0072] The number of preset nearest neighbors of the K neighbor algorithm is , then the point cloud density of the i-th point cloud data in the K-th group of building point cloud data after the first downsampling is The specific calculation method is:

[0073] ;

[0074] in, is the mean of the spatial Euclidean distances between the i-th point cloud data and each nearest neighboring point in the K-th group of building point cloud data after the first downsampling.

[0075] Similarly, we obtain the K+1th group of building point cloud data. The point cloud density of the point cloud data after the first downsampling , then the downsampling supervision condition of the Kth group of building point cloud data and the K+1th group of building point cloud data in the first downsampling result is The specific calculation method is:

[0076] ;

[0077] in, is the number of point cloud data of the Kth group of building point cloud data after the first downsampling;

[0078] It represents the difference in point cloud density between the i-th point cloud data in the K-th building point cloud data group after the first downsampling and its registration point cloud data in the K+1-th building point cloud data group. When the point cloud density difference is the smallest, it means that the two building point cloud data groups have been downsampled to the same point cloud density. The threshold parameter of the number of downsampling iterations as the downsampling supervision condition avoids excessive downsampling when the same point cloud density cannot be achieved, so that the number of point cloud data in the two groups of building point cloud data is smaller, resulting in downsampling distortion.

[0079] Furthermore, the downsampling supervision condition for the first downsampling result Use the sigmoid function for normalization to obtain the normalized result of the downsampling supervision condition of the first downsampling result; preset the downsampling supervision threshold of 0.2, when the normalized result of the downsampling supervision condition of the first downsampling result is greater than or equal to the downsampling supervision threshold, it means that the point cloud density of the point cloud data in the two groups of building point cloud data has not reached the minimum difference, and the first downsampling result of the two groups of building point cloud data is used as the original data to be downsampled, and the original data to be downsampled is used for the second downsampling to obtain the normalized result of the downsampling supervision condition of the second downsampling result, and judged according to the downsampling supervision threshold; and so on, until there is a constant So that the Kth and K+1th building point cloud data groups are in the When the normalized result of the downsampling supervision condition after the downsampling result is less than the downsampling supervision threshold, it means that the point cloud density of the point cloud data in the two groups of building point cloud data sets has reached the minimum difference, or has reached the threshold of the number of downsampling times, the first downsampling supervision condition of the two groups of building point cloud data sets is reduced to the minimum. The downsampling results are recorded as the optimal downsampling results of the two groups of building point cloud data sets.

[0080] Preferably, according to the building rigidity trend of each point cloud data in the optimal downsampling results of the two building point cloud data groups, the directional registration positions of the two building point cloud data groups are obtained, and the specific steps of obtaining the directional registration positions are as follows:

[0081] It should be noted that the optimal downsampling results of the two groups of building point cloud data obtained in the above steps mainly include point cloud data of building structures and some invalid point cloud data with strong features. Therefore, the building structure presents obvious distribution directionality in the point cloud data. For example, the building body and the floor, the building body as a whole presents a directionality perpendicular to the ground plane and upward, and the floor presents a directionality parallel to the ground plane. Since it is necessary to obtain the offset vector of each pair of matching points for iteration in the ICP registration process, and this embodiment obtains auxiliary registration points by screening the original point cloud data, wherein the auxiliary registration points represent point cloud data that can reflect the rigidity of the building structure, and the building structure mainly reflects the rigid distribution directionality. Therefore, this embodiment extracts the main distribution directions of all point cloud data in the optimal downsampling results as the directional registration positions of the auxiliary registration points in each group of building point cloud data, which are used to replace the coordinates of the original point cloud data.

[0082] Preferably, for the optimal downsampling result of the Kth group of building point cloud data, which includes a number of downsampled point cloud data, the downsampled point cloud data are projected in a three-dimensional Cartesian coordinate system to obtain a coordinate projection of the optimal downsampling result of the Kth group of building point cloud data, wherein the horizontal axis and the vertical axis are respectively the horizontal plane coordinates of the point cloud data, and the Z axis is the elevation coordinate of the point cloud data;

[0083] Use the principal component analysis algorithm to reduce the dimension of the coordinate projection of the optimal downsampling result of the K-th group of building point cloud data, and obtain several principal component directions of the K-th group of building point cloud data under the optimal downsampling result;

[0084] The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the K-th group of building point cloud data is projected to the two-dimensional coordinate system of the horizontal plane to obtain the horizontal plane principal component direction projection of the K-th group of building point cloud data under the optimal downsampling result, and the counterclockwise angle between the horizontal plane principal component direction projection and the horizontal axis is recorded as the horizontal plane registration direction of the K-th group of building point cloud data. ;

[0085] The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the Kth group of building point cloud data set is projected to the two-dimensional coordinate system of the vertical axis and the Z axis to obtain the elevation principal component direction projection of the Kth group of building point cloud data set under the optimal downsampling result, and the counterclockwise angle between the elevation principal component direction projection and the vertical axis is recorded as the elevation registration direction of the Kth group of building point cloud data set .

[0086] The principal component direction with the largest eigenvalue of the Kth group of building point cloud data sets under the optimal downsampling result represents the building rigidity trend of the Kth group of building point cloud data sets under the optimal downsampling result.

[0087] Furthermore, the horizontal registration direction and the elevation registration direction represent the floor direction and the building body direction of the building respectively. Therefore, the horizontal registration direction and the elevation registration direction of the Kth group of building point cloud data are combined into , recorded as the direction registration position of the Kth group of building point cloud data Similarly, the directional registration position of each building point cloud data set is obtained.

[0088] It should be noted that the directional registration position represents the direction of the structural features of the building belonging to the building body and floors in the collected building point cloud data group. The purpose of this embodiment is to extract the point cloud data belonging to the building from all the point cloud data, and then avoid the interference of invalid point cloud data during matching. The point cloud data of the building has extremely strong structural rigidity characteristics, so the structural rigidity characteristics are quantified to obtain the directional registration position.

[0089] It should be further explained that the above steps obtain the directional registration position of each group of building point cloud data groups, which represents the directional characteristics of the building structure in the group of building point cloud data groups, but does not further screen all the point cloud data in the group, that is, a lot of invalid point cloud data are still retained. Therefore, this embodiment performs reverse downsampling iterative analysis, and then optimizes the detection points screened out by analyzing the optimal downsampling results to obtain the detection points representing the rigidity of the building structure for ICP alignment.

[0090] Preferably, based on the optimal downsampling result of each group of building point cloud data groups, a number of detection points of each group of building point cloud data groups under the optimal downsampling result are obtained; according to the building rigidity presentation of each detection point of the two groups of building point cloud data groups at each downsampling scale, the building rigidity expression coefficient of each detection point of each group of building point cloud data groups at each downsampling scale is obtained, and the specific steps for obtaining the building rigidity expression coefficient are as follows:

[0091] It should be noted that the process of obtaining the optimal downsampling result in the above steps is to iteratively downsample at the same preset three-dimensional voxel grid scale. The result of the downsampling is the optimal downsampling result. The point cloud data contained in the optimal downsampling result is the most and has the most rigid performance of the building structure. Therefore, several points are selected as detection points in the optimal downsampling result, and the detection points are marked. The reverse downsampling of the same three-dimensional voxel grid scale is performed, and the building rigidity performance coefficient of each detection point at each downsampling scale is obtained according to the building structure rigidity presented by other point cloud data in the three-dimensional voxel grid where each detection point is located.

[0092] Preferably, as an example, this embodiment uses a preset three-dimensional voxel grid scale as a division scale for detection points, and equally divides the optimal downsampling result of each group of building point cloud data groups to obtain a number of voxel blocks, each voxel block has 8 vertices, and adjacent voxel blocks share multiple vertices. From the optimal downsampling result of each group of building point cloud data groups, point cloud data with the closest Euclidean distance to each vertex of each voxel block is selected as the detection point of the optimal downsampling result of each group of building point cloud data groups.

[0093] It should be further explained that if the detection point of each group of building point cloud data groups under the optimal downsampling result is an edge of a building structure, such as a core part such as a building body or floor, it exhibits obvious flatness and rigidity characteristics in the local range, that is, the point cloud data in the local range of each detection point has an obvious orientation trend and the direction of the trend is the same; therefore, this embodiment obtains several principal component directions of each detection point at each downsampling scale based on the trend of each detection point. Since it exhibits an obvious trend performance and is single in local performance, the corresponding eigenvalues ​​of the maximum principal component direction and the second largest principal component are quite different. Therefore, based on the several principal component directions of each detection point at each downsampling scale, the building rigidity performance coefficient of each detection point at each downsampling scale is obtained.

[0094] Preferably, a preset three-dimensional voxel grid scale is used as the spatial neighborhood range of each detection point under the optimal downsampling result, and a principal component analysis algorithm is used to reduce the dimension of all point cloud data within the spatial neighborhood range of each detection point under the optimal downsampling result to obtain a plurality of principal component directions of each detection point under the optimal downsampling result, wherein each principal component direction corresponds to an eigenvalue;

[0095] Since the optimal downsampling result is Therefore, the principal component directions of each detection point under the optimal downsampling result are taken as the principal component directions of each detection point in the first Several principal component directions under downsampling scales;

[0096] Furthermore, each detection point in each building point cloud data set is The difference between the eigenvalue corresponding to the direction of the largest principal component under the downsampling scale and the eigenvalue corresponding to the direction of the second largest principal component is recorded as the value of each detection point in each group of building point cloud data set. The building rigidity performance coefficient under the downsampling scale.

[0097] Similarly, each detection point in each group of building point cloud data is reversely downsampled according to the preset three-dimensional voxel grid scale, and the building rigidity performance coefficient of each detection point in each group of building point cloud data is obtained from the 0th downsampling scale to the cth downsampling scale;

[0098] Among them, after each reverse downsampling, the spatial neighborhood range of each detection point increases compared with the previous reverse downsampling, and the increase is the preset three-dimensional voxel grid scale size; the 0th downsampling scale corresponds to all point cloud data corresponding to each detection point in the original building point cloud data group before downsampling.

[0099] At this point, the building rigidity performance coefficient of each detection point in the Kth group and the K+1th group of building point cloud data groups at each downsampling scale is obtained.

[0100] Preferably, the building structural property of each detection point in the two building point cloud data sets is obtained according to the building rigidity performance coefficient of each detection point in each downsampling scale and the local space point cloud density of each detection point in each building point cloud data set; all preferred detection points in the two building point cloud data sets are screened according to the building structural property of each detection point in the two building point cloud data sets, and a number of auxiliary registration points in the two building point cloud data sets are obtained based on the directional registration positions of the two building point cloud data sets and the positions of the preferred detection points; the specific steps of obtaining the auxiliary registration points include:

[0101] It should be noted that the building rigidity expression coefficient of each detection point at each downsampling scale only expresses the building structure rigidity characteristics in the spatial neighborhood range at the current downsampling scale. Since the building structure includes parts with local building structure rigidity characteristics such as stairs, windows, and elevator shafts in addition to the building body and floors, it is possible that some detection points have larger building rigidity expression coefficients at a specific downsampling scale, while smaller or larger building rigidity expression coefficients have smaller values. Therefore, this embodiment obtains the building structurality of each detection point in each group of building point cloud data sets based on the consistency of the overall building structure rigidity expression of the building, that is, the difference in the building structure rigidity expression coefficient at different downsampling scales, and the fact that it will not be affected by the local structure and therefore the point cloud density difference of the overall structure of the detection point is small.

[0102] Preferably, as an example, this embodiment takes the vth detection point in the Kth group of building point cloud data as an example to calculate the building structurality, and the calculation method of the detection points in other groups of building point cloud data is the same as the calculation method of the vth monitoring point. Then the building structurality of the vth detection point in the Kth group of building point cloud data is The calculation method is:

[0103] ;

[0104] in, Indicates that the vth detection point in the Kth group of building point cloud data is The building rigidity performance coefficient under the downsampling scale is represents the variance of the building rigidity performance coefficient of the vth detection point in the Kth group of building point cloud data under all downsampling scales; Indicates that the vth detection point in the Kth group of building point cloud data is Point cloud density at downsampling scale, Represents the variance of the point cloud density of the vth detection point in the Kth group of building point cloud data at all downsampling scales; It represents the maximum downsampling scale of the Kth group and the K+1th group of building point cloud data, and the maximum downsampling scale corresponds to the optimal downsampling result of the two groups of building point cloud data; norm represents the linear normalization function.

[0105] It should be noted that the variance of the building rigidity performance coefficient And the variance of the point cloud density The smaller the value, the weaker the change of the building structure rigidity of the vth detection point in the Kth group of building point cloud data with the change of the monitoring downsampling scale, and the more likely it is the main structure of the building rather than the detailed structure. It represents the cumulative sum of the point cloud density at all downsampling scales. The larger the value, the higher the point cloud data density within the spatial neighborhood of the detection point, that is, the more representative it can be used for ICP registration.

[0106] Furthermore, a building structure judgment threshold of 0.8 is preset, and when the building structure of the vth detection point in the Kth group of building point cloud data is greater than or equal to the building structure judgment threshold, the vth detection point is recorded as a preferred detection point.

[0107] Similarly, all the preferred detection points in the Kth group of building point cloud data sets and all the preferred detection points in the K+1th group of building point cloud data sets are obtained.

[0108] Furthermore, the Euclidean distance between each preferred detection point and the origin in the Kth group of building point cloud data is recorded as the vector registration position of each preferred detection point in the Kth group of building point cloud data, the vector registration position is used as the vector modulus, the direction registration position of the Kth group of building point cloud data is used as the vector direction, and the constructed vector is recorded as an auxiliary registration point in the Kth group of building point cloud data.

[0109] Similarly, all auxiliary registration points of the Kth group of building point cloud data and all auxiliary registration points of the K+1th group of building point cloud data are obtained.

[0110] The registration result acquisition module 103 is used to register the auxiliary registration points of all groups of building point cloud data using a registration algorithm to obtain the registration results of all building point cloud data.

[0111] All auxiliary registration points of the Kth group of building point cloud data and all auxiliary registration points of the K+1th group of building point cloud data are input into the ICP registration algorithm for registration to obtain the optimal transformation matrix of the Kth group of building point cloud data and the K+1th group of building point cloud data, and the optimal transformation matrix of the Kth group of building point cloud data and the K+1th group of building point cloud data is output as the registration result of the Kth group of building point cloud data and the K+1th group of building point cloud data; it should be noted that the ICP registration algorithm performs registration on the input point cloud data to obtain the optimal transformation matrix, which is an existing public technology and will not be repeated in this embodiment.

[0112] This is the end of the present invention.

[0113] 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 principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A digital construction engineering simulation system based on point cloud data, characterized in that: The system includes the following modules: A building point cloud data acquisition module, used to acquire a plurality of building point cloud data groups of the construction building, wherein the point cloud data in the building point cloud data group includes a horizontal axis, a vertical axis and a Z axis; The auxiliary registration point acquisition module is used to iterate based on the difference in local spatial point cloud density of each point cloud data in the downsampling results of the two groups of building point cloud data groups to obtain the optimal downsampling results of the two groups of building point cloud data groups; the specific steps of obtaining the optimal downsampling results include: Preset the three-dimensional voxel grid scale, use the voxel grid filtering algorithm to perform the first downsampling on the K-th group of building point cloud data and the K+1-th group of building point cloud data, and obtain the first downsampling results of the K-th group of building point cloud data and the K+1-th group of building point cloud data; Obtaining downsampling supervision conditions of the Kth group of building point cloud data and the K+1th group of building point cloud data in the first downsampling result; The downsampling supervision conditions of the first downsampling results of the K-th group of building point cloud data and the K+1-th group of building point cloud data are normalized using a sigmoid function to obtain a normalized result of the downsampling supervision conditions of the first downsampling results; A downsampling supervision threshold is preset. When the normalized result of the downsampling supervision condition of the first downsampling result is greater than or equal to the downsampling supervision threshold, the first downsampling result of the two groups of building point cloud data is used as the downsampled original data, and the downsampled original data is used for the second downsampling to obtain the normalized result of the downsampling supervision condition of the second downsampling result, and the judgment is made according to the downsampling supervision threshold; and so on, until there is a constant So that the Kth and K+1th building point cloud data groups are in the When the normalized result of the downsampling supervision condition after the downsampling result is less than the downsampling supervision threshold, it means that the point cloud density of the point cloud data in the two groups of building point cloud data sets has reached the minimum difference, or has reached the threshold of the number of downsampling times, the first downsampling supervision condition of the two groups of building point cloud data sets is reduced to the minimum. The downsampling results are recorded as the optimal downsampling results of the two groups of building point cloud data sets; According to the building rigidity trend of each point cloud data in the optimal downsampling results of the two building point cloud data groups, the directional alignment positions of the two building point cloud data groups are obtained; the specific steps of obtaining the directional alignment positions include: The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the K-th group of building point cloud data is projected to the two-dimensional coordinate system of the horizontal plane to obtain the horizontal plane principal component direction projection of the K-th group of building point cloud data under the optimal downsampling result, and the counterclockwise angle between the horizontal plane principal component direction projection and the horizontal axis is recorded as the horizontal plane registration direction of the K-th group of building point cloud data. ; The principal component direction with the largest eigenvalue of the principal component direction in the optimal downsampling result of the Kth group of building point cloud data set is projected to the two-dimensional coordinate system of the vertical axis and the Z axis to obtain the elevation principal component direction projection of the Kth group of building point cloud data set under the optimal downsampling result, and the counterclockwise angle between the elevation principal component direction projection and the vertical axis is recorded as the elevation registration direction of the Kth group of building point cloud data set ; Combine the horizontal registration direction and the elevation registration direction of the Kth group of building point cloud data sets , recorded as the direction registration position of the Kth group of building point cloud data ; Based on the optimal downsampling result of each group of building point cloud data, a number of detection points of each group of building point cloud data under the optimal downsampling result are obtained; According to the building rigidity presentation of each detection point of the two building point cloud data groups at each downsampling scale, the building rigidity presentation coefficient of each detection point of each building point cloud data group at each downsampling scale is obtained; According to the building rigidity performance coefficient of each detection point in each building point cloud data set at each downsampling scale and the local space point cloud density, the building structurality of each detection point in the two building point cloud data sets is obtained; According to the architectural structure of each detection point in the two sets of building point cloud data sets, all preferred detection points in the two sets of building point cloud data sets are selected, and based on the directional registration positions of the two sets of building point cloud data sets and the positions of the preferred detection points, several auxiliary registration points in the two sets of building point cloud data sets are obtained; the specific steps of obtaining the auxiliary registration points include: A building structure judgment threshold is preset, and when the building structure of the vth detection point in the Kth group of building point cloud data is greater than or equal to the building structure judgment threshold, the vth detection point is recorded as a preferred detection point; Obtain all the preferred detection points in the K-th group of building point cloud data and all the preferred detection points in the K+1-th group of building point cloud data; The Euclidean distance between each preferred detection point in the K-th group of building point cloud data and the origin is recorded as the vector registration position of each preferred detection point in the K-th group of building point cloud data, the vector registration position is used as the vector modulus, the direction registration position of the K-th group of building point cloud data is used as the vector direction, and the constructed vector is recorded as an auxiliary registration point in the K-th group of building point cloud data; The registration result acquisition module is used to register the auxiliary registration points of all groups of building point cloud data using the registration algorithm to obtain the registration results of all building point cloud data.

2. A digital construction engineering simulation system based on point cloud data according to claim 1, characterized in that: The specific method of obtaining the downsampling supervision condition is: Downsampling supervision conditions of the Kth group of building point cloud data and the K+1th group of building point cloud data in the first downsampling result The specific calculation method is: in, is the number of point cloud data of the Kth group of building point cloud data after the first downsampling, is the point cloud density of the i-th point cloud data in the K-th group of building point cloud data after the first downsampling; is the K+1th group of building point cloud data. The point cloud density of the point cloud data after the first downsampling.

3. A digital construction engineering simulation system based on point cloud data according to claim 2, characterized in that: The specific method for obtaining the point cloud density is: In the first downsampling results of the Kth group of building point cloud data groups and the K+1th group of building point cloud data groups, for the i-th point cloud data in the Kth group of building point cloud data groups, the point cloud data with the closest spatial Euclidean distance to the corresponding coordinate position of the i-th point cloud data in the Kth group of building point cloud data groups is selected in the K+1th group of building point cloud data groups, and is recorded as the i-th point cloud data in the K+1th group of building point cloud data groups. Point cloud data; The default number of nearest neighbors is , the point cloud density of the i-th point cloud data in the K-th group of building point cloud data after the first downsampling The specific calculation method is: in, is the mean of the spatial Euclidean distances between the i-th point cloud data and each nearest neighbor in the K-th group of building point cloud data after the first downsampling.

4. The digital construction engineering simulation system based on point cloud data according to claim 1, characterized in that: The specific method of obtaining the principal component direction in the optimal downsampling result is: Projecting the point cloud data in the optimal downsampling result of the Kth group of building point cloud data sets in a three-dimensional Cartesian coordinate system to obtain a coordinate projection of the optimal downsampling result of the Kth group of building point cloud data sets, wherein the horizontal axis and the vertical axis are respectively the horizontal plane coordinates of the point cloud data, and the Z axis is the elevation coordinate of the point cloud data; The principal component analysis algorithm is used to reduce the dimension of the coordinate projection of the optimal downsampling result of the Kth group of building point cloud data groups, and the principal component direction of the Kth group of building point cloud data groups under the optimal downsampling result is obtained.

5. The digital construction engineering simulation system based on point cloud data according to claim 1, characterized in that: The specific method of obtaining the detection point is: Taking the preset three-dimensional voxel grid scale as the division scale of the detection points, the optimal downsampling result of each group of building point cloud data groups is equally divided to obtain a number of voxel blocks. In the optimal downsampling result of each group of building point cloud data groups, the point cloud data with the closest Euclidean distance to each vertex of each voxel block is selected as the detection point in the optimal downsampling result of each group of building point cloud data groups.

6. The digital construction engineering simulation system based on point cloud data according to claim 1, characterized in that: The specific steps of obtaining the building rigidity performance coefficient include: The preset three-dimensional voxel grid scale is used as the spatial neighborhood range of each detection point under the optimal downsampling result, and the principal component analysis algorithm is used to reduce the dimension of all point cloud data within the spatial neighborhood range of each detection point under the optimal downsampling result to obtain several principal component directions of each detection point under the optimal downsampling result; The principal component directions of each detection point under the optimal downsampling result are used as the Several principal component directions under downsampling scales; Each detection point in each building point cloud data set is The difference between the eigenvalue corresponding to the direction of the largest principal component under the downsampling scale and the eigenvalue corresponding to the direction of the second largest principal component is recorded as the value of each detection point in each group of building point cloud data set. The building rigidity performance coefficient under the downsampling scale; Each detection point in each group of building point cloud data is reversely downsampled according to the preset three-dimensional voxel grid scale to obtain the building rigidity performance coefficient of each detection point in each group of building point cloud data from the 0th downsampling scale to the cth downsampling scale.

7. The digital construction engineering simulation system based on point cloud data according to claim 1, characterized in that: The specific steps of obtaining the building structure include: The building structure of the vth detection point in the Kth group of building point cloud data The calculation method is: in, Indicates that the vth detection point in the Kth group of building point cloud data is The building rigidity performance coefficient under the downsampling scale is represents the variance of the building rigidity performance coefficient of the vth detection point in the Kth group of building point cloud data under all downsampling scales; Indicates that the vth detection point in the Kth group of building point cloud data is Point cloud density at downsampling scale, Represents the variance of the point cloud density of the vth detection point in the Kth group of building point cloud data at all downsampling scales; It represents the maximum downsampling scale of the Kth group and the K+1th group of building point cloud data, and the maximum downsampling scale corresponds to the optimal downsampling result of the two groups of building point cloud data; norm represents the normalization function.

Images