A digital point cloud correction processing method for the entire construction process

Abstract

The present invention relates to the technical field of point cloud correction, and discloses a method for correcting and processing digital point clouds throughout the entire construction process, the method comprising: performing coordinate transformation on the collected three-dimensional point cloud data to obtain three-dimensional point cloud data in a world coordinate system; enhancing the three-dimensional point cloud data after coordinate transformation using an optimal three-dimensional point cloud data enhancement model, and performing similarity calculation in combination with the three-dimensional data of the building design target, thereby calculating transformation parameters; if the similarity calculation result is less than a specified threshold, correcting the three-dimensional point cloud data of the current building state based on the transformation parameters, and performing corresponding correction processing on the current building. The present invention obtains the missing point cloud coordinates of the point cloud data at different resolutions by performing multiple rounds of sampling and perception processing on the three-dimensional point cloud data, and then complements and enhances the missing point cloud data, and performs point cloud correction based on the transformation parameters according to the similarity between the current building state and the target building state in the direction.

Description

Technical Field

[0001] The invention relates to the technical field of point cloud correction, and in particular to a digital point cloud correction processing method for the entire construction process. Background Art

[0002] At present, in the modern construction production process, the construction task is mainly completed by manual calibration and mechanical automatic filling. First, the construction skeleton is calibrated offline by construction workers, and then the filling equipment is used to reach the designated location to complete the construction task. However, the target workpieces in the industrial site are mostly planar weak geometric contour structures. Planar weak geometric contour structures mainly refer to the objects to be measured with plane features as the main overall features, with single contour geometric features, limited surface normal vector changes and less surface texture. At the same time, due to the existence of the filling equipment's own structural parameters, workpiece calibration results, system open-loop control and other influencing factors, it is impossible to complete the construction task under strict tolerances, so that the building has a weak deviation from the actual expected building. Although this type of weak deviation has no great impact in low-rise buildings, as the building height increases, there is a deviation accumulation effect, which has certain safety hazards. In response to this problem, the present invention proposes a digital point cloud correction processing method for the entire construction process to achieve precise error control and correction of the entire construction process. Summary of the invention

[0003] In view of this, the present invention provides a method for digital point cloud correction processing of the entire construction process, the purpose of which is to: 1) establish a mapping conversion relationship between the pixel coordinates captured by a camera and the world coordinates based on the 3D point cloud data of the current state of the target building, and then convert the 3D point cloud data from the camera coordinate system to the world coordinate system, and adopt the same method to obtain the transformation parameters between the 3D point cloud data and the 3D data of the building design target, and calculate the similarity between the two by combining the polar angle of the 3D point cloud coordinates in the polar coordinate system and the polar angle of the 3D coordinates of the building design target to characterize the difference in direction of the 3D data. If the similarity calculation result is less than a specified threshold, it means that there is a difference between the current building state and the building design target. In the large difference, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction results, so as to realize the digital point cloud correction processing of the whole process of the building; 2) by constructing a three-dimensional point cloud data enhancement model, the point cloud data is sampled and processed at different resolutions, and the sampling results of different resolutions are encoded to form multi-level coding features, and then the multi-level coding features are perceived and up-sampled to obtain the missing point cloud coordinates at different resolutions, and the missing point cloud coordinates at different resolutions are combined with the original point cloud coordinates to form a completed and enhanced three-dimensional point cloud coordinate set, so as to realize the completion and enhancement of missing point cloud data.

[0004] To achieve the above object, the present invention provides a method for correcting and processing digital point clouds throughout the construction process, comprising the following steps:

[0005] S1: Collect the 3D point cloud data of the current state of the target building, and perform coordinate transformation on the collected 3D point cloud data to obtain the 3D point cloud data in the world coordinate system;

[0006] S2: construct a 3D point cloud data enhancement model, the constructed model takes the 3D point cloud data in the world coordinate system as input and takes the enhanced 3D point cloud data as output;

[0007] S3: Determine the optimization objective function of the 3D point cloud data enhancement model and perform optimization training to obtain the optimal 3D point cloud data enhancement model;

[0008] S4: using the optimal 3D point cloud data enhancement model to enhance the 3D point cloud data after coordinate transformation, performing similarity calculation on the enhanced 3D point cloud data combined with the architectural design target 3D data, and calculating the transformation parameters;

[0009] S5: If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

[0010] As a further improvement method of the present invention:

[0011] Optionally, collecting three-dimensional point cloud data of the current state of the target building in step S1 includes:

[0012] The camera is used to collect the three-dimensional point cloud data of the current state of the target building. The camera consists of an RGB color camera and a depth camera. The RGB color camera is used to obtain the color information of the target building, and the depth camera is used to obtain the depth information of the target building. The image of the target building taken by the camera is I. The formula for converting any pixel I (x, y) in the image I into a three-dimensional point cloud coordinate is:

[0013]

[0014] in:

[0015] g(x, y, z) represents the three-dimensional point cloud coordinates corresponding to the pixel I(x, y) in the target building image, and the pixel I(x, y) represents the pixel in the xth row and yth column in the target building image I;

[0016] c X represents the principal point of the camera in the horizontal direction, c Y Indicates the principal point of the camera in the vertical direction;

[0017] f XIndicates the horizontal focal length of the camera, f Y Indicates the vertical focal length of the camera;

[0018] d x，y Represents the depth information of pixel I(x, y) partially captured by the depth camera;

[0019] An RGB color camera is used to obtain the color value of the pixel in the RGB color channel to form three-dimensional point cloud data, where the three-dimensional point cloud data of pixel I(x, y) is (g(x, y, z), R(x, y), G(x, y), B(x, y)}, where R(x, y), G(x, y), B(x, y) represent the color value of pixel I(x, y) in the RGB color channel respectively.

[0020] In the embodiment of the present invention, the principal point and focal length of the camera are both internal parameters of the camera, that is, known parameter information.

[0021] Optionally, in step S1, coordinate transformation is performed on the collected three-dimensional point cloud data to obtain three-dimensional point cloud data in a world coordinate system, including:

[0022] The collected 3D point cloud data is subjected to coordinate transformation, wherein the coordinate transformation object is the 3D point cloud coordinates in the 3D point cloud data, and the 3D point cloud coordinates are transformed into the world coordinate system, wherein the coordinate system transformation formula for the 3D point cloud coordinates is:

[0023] g′(x,y,z)=Rg(x,y,z) T +H

[0024] in:

[0025] R represents the rotation matrix between the current point cloud coordinate system and the world coordinate system, and H represents the translation matrix;

[0026] T stands for transpose;

[0027] g′(x, y, z) represents the coordinate transformation result of the three-dimensional point cloud coordinate g(x, y, z);

[0028] The calculation process of the rotation matrix and the translation matrix is ​​as follows:

[0029] S11: Use the camera to obtain the three-dimensional point cloud coordinates of N known coordinates in the world coordinate system:

[0030] {(p i ,q i )|i∈[1,N]}

[0031] in:

[0032] p irepresents the coordinates of the i-th 3D point cloud captured by the camera, q i Indicates p i Coordinates in the world coordinate system;

[0033] S12: Standardize the coordinates:

[0034]

[0035]

[0036] in:

[0037] They are the standardized results of pi and qi respectively;

[0038] S13: The normalized coordinates are used to form matrices P and Q respectively:

[0039]

[0040]

[0041] S14: Calculate the transformation matrices A and B:

[0042] A=P T Q

[0043] B=PQ T

[0044] S15: Calculate the N eigenvalues ​​and eigenvectors of the transformation matrix A and B respectively, and sort the eigenvectors in descending order of the eigenvalues, where the N eigenvectors of the transformation matrix A are:

[0045] (α1, α2, ..., α i , ..., α N )

[0046] in:

[0047] α i represents the i-th eigenvector in the transformation matrix A;

[0048] The N eigenvectors of the transformation matrix B are:

[0049] (β1, β2,...,βi,...,β N )

[0050] in:

[0051] β i represents the i-th eigenvector in the transformation matrix B;

[0052] S16: Construct the eigenvectors into matrix forms α and β respectively:

[0053] α=[α1,α2,...,α i , ..., α N ]

[0054] β=[β1,β2,...,β i , ..., β N ]

[0055] S17: Calculate the rotation matrix R between the current point cloud coordinate system and the world coordinate system, and the translation matrix H:

[0056] R=αβ T

[0057] H=Q-RP

[0058] The calculated rotation matrix R and translation matrix H are used to transform the three-dimensional point cloud coordinates in the three-dimensional point cloud data into the world coordinate system.

[0059] Optionally, constructing a three-dimensional point cloud data enhancement model in step S2 includes:

[0060] Constructing a 3D point cloud data enhancement model, wherein the constructed 3D point cloud data enhancement model takes the 3D point cloud data in the world coordinate system as input and takes the enhanced 3D point cloud data as output;

[0061] The constructed 3D point cloud data enhancement model includes an input layer, a multi-level encoding layer and a decoding point cloud enhancement layer, wherein the input layer is used to receive 3D point cloud data and perform multi-resolution downsampling processing on the received 3D point cloud data, the multi-level encoding layer is used to encode the sampling results of different resolutions to form multi-level encoding features, and the decoding point cloud enhancement layer is used to decode the multi-level encoding features, complete the missing parts of the sampling results at different resolutions, and obtain enhanced 3D point cloud data;

[0062] The enhancement process based on the 3D point cloud data enhancement model is as follows:

[0063] S21: taking the three-dimensional point cloud data Data in the world coordinate system as input, wherein Data includes D groups of three-dimensional point cloud coordinates in the world coordinate system and color values ​​corresponding to the coordinates;

[0064] S22: The input layer samples the D groups of 3D point cloud coordinates at resolutions of D / 2, D / 4, and D / 8 to obtain three groups of sampling results, where the sampling process for the D / 2 resolution is as follows:

[0065] S221: Select a coordinate from the D group of three-dimensional point cloud coordinates as a starting point, construct a sampling point set, and add the starting point to the sampling point set;

[0066] S222: Calculate the distance from the external coordinates of the sampling point set to the starting point, and select the coordinate with the largest distance to add to the sampling point set;

[0067] S223: Calculate the shortest distance from the external coordinates of the sampling point set to any coordinates in the sampling point set, select the coordinate corresponding to the maximum shortest distance and add it to the sampling point set;

[0068] S224: Repeat steps S222 to S223 until the number of sampling points in the sampling point set reaches

[0069] S23: The multi-layer encoding layer uses a multi-layer perceptron to increase the dimension of each set of sampling results until the dimension of each sampling coordinate reaches 128; the maximum pooling layer is used to extract the feature map with the highest dimension of each set of sampling results. The dimension of each feature map is 128, where the feature map feature under D / 2 resolution sampling is D / 2 is:

[0070] feature D / 2 =Pooling(MLE(sample D / 2 (1))||...||MLE(sample D / 2(D / 2)))

[0071] in:

[0072] || represents the concatenation of the processing results of the multi-layer perceptron;

[0073] MLE(sample D / 2 (D / 2)) represents the dimension-raising result of the D / 2th sampling result under D / 2 resolution sampling, and MLE represents multi-layer perceptron;

[0074] Pooling (·) represents the pooling operation. For the sampling result under D / 2 resolution sampling, the dimensionality-enhanced 128×3 dimensionality-enhanced result MLE (sample D / 2 (1))||...||MLE(sample D / 2 (D / 2)) pooled to 128 dimensions to obtain the feature map feature D / 2 ;

[0075] S24: The multi-level coding layer uses multiple independent convolution layers to perform convolution processing on the feature maps of the three groups of sampling results, and assigns different weights to the convolution processing results of the three groups of feature maps, and splices the weighted convolution processing results to obtain multi-level coding features; in this embodiment of the present invention, the weight of the feature map under D / 2 resolution sampling is 0.6, the weight of the feature map under D / 4 resolution sampling is 0.2, and the weight of the feature map under D / 8 resolution sampling is 0.2;

[0076] S25: The decoding point cloud enhancement layer contains multiple multi-layer perceptrons with different parameters. The multi-layer perceptron is used to reconstruct and upsample the multi-level coding features to obtain the missing point cloud coordinates at different resolutions, and the missing point cloud coordinates at different resolutions are combined with the original point cloud coordinates to form a set of completed and enhanced three-dimensional point cloud coordinates, where the color value of the missing point cloud coordinate is the average color value of the adjacent three-dimensional point cloud coordinates, to obtain enhanced three-dimensional point cloud data.

[0077] Optionally, determining the optimization objective function of the three-dimensional point cloud data enhancement model and performing optimization training in step S3 includes:

[0078] Determine the optimization objective function of the three-dimensional point cloud data enhancement model, and optimize the training of the three-dimensional point cloud data enhancement model based on the optimization objective function to obtain the trained model parameters, wherein the model parameters to be trained are the multi-layer perceptron parameters in the multi-level encoding layer and the convolution layer parameters θ1, and the multi-layer perceptron parameters θ2 in the decoding point cloud enhancement layer, wherein θ1 and θ2 are parameter vectors;

[0079] Obtain a set of 3D point cloud data with missing point cloud data, and obtain a set of real 3D point cloud coordinates of the missing point cloud data. The determined optimization objective function is:

[0080]

[0081] in:

[0082] Ω represents the real 3D point cloud coordinate set of missing point cloud data, num(Ω) represents the number of missing point cloud data, and num represents the number of missing point clouds obtained by the 3D point cloud data enhancement model;

[0083] loc c Represents the true 3D point cloud coordinates of the cth missing point cloud data;

[0084] Represents the distance loc output by the 3D point cloud data enhancement model c The nearest missing point cloud coordinates;

[0085] dis(·) represents the distance calculation formula;

[0086] ε represents the proportional weight, which is set to 0.2;

[0087] Based on the optimization objective function, the Adam optimizer is used to train the 3D point cloud data enhancement model to obtain the trained model parameters, and then the optimal 3D point cloud data enhancement model is constructed based on the trained model parameters.

[0088] Optionally, in step S4, enhancing the three-dimensional point cloud data after coordinate transformation using an optimal three-dimensional point cloud data enhancement model includes:

[0089] The three-dimensional point cloud data after the coordinate transformation in step S2 is input into the optimal three-dimensional point cloud data enhancement model, and the optimal three-dimensional point cloud data enhancement model outputs enhanced three-dimensional point cloud data, wherein the enhanced three-dimensional point cloud data set is:

[0090] {r m |m∈[1,M]}

[0091] in:

[0092] r m It represents the mth 3D point cloud data output by the optimal point cloud enhancement model. The 3D point cloud data includes 3D point cloud coordinates and 3D point cloud color values. M represents the number of 3D point cloud data output by the optimal point cloud enhancement model.

[0093] Optionally, in step S4, similarity calculation is performed on the enhanced three-dimensional point cloud data in combination with the architectural design target three-dimensional data, and transformation parameters are calculated, including:

[0094] The enhanced 3D point cloud data is combined with the 3D data of the architectural design target to perform similarity calculation, and the transformation parameters are calculated. The similarity calculation process is as follows:

[0095] S41: using a Delaunay triangulation method to convert the three-dimensional point cloud coordinates in the three-dimensional point cloud data set into a number of Delaunay triangulation networks, and using the Delaunay triangulation method to convert the three-dimensional data of the architectural design target into a number of Delaunay triangulation networks;

[0096] S42: Extract the three-dimensional point cloud coordinates in the three-dimensional point cloud data, where r m The three-dimensional point cloud coordinates in are L m ;

[0097] S43: Take any three-dimensional point cloud coordinates L m As the vertex, we get L m Several groups of directly connected points are directly connected to form a three-dimensional point cloud coordinate L m As the pole, L m The direction of the ray to any directly connected point is the polar axis, and the pole L is calculated.m The polar angle of the ray formed by different direct connection points is obtained by the pole L m The corresponding polar angle set;

[0098] S44: Select the three-dimensional data of the building design target and the three-dimensional point cloud coordinates L m The closest three-dimensional coordinate point L′ m , with the three-dimensional coordinate point L′ m As the pole, L′ m The direction of the ray to any directly connected point is the polar axis, and the pole L' is calculated m The polar angle of the ray formed by different direct connection points gives the pole L' m The corresponding polar angle set;

[0099] S45: Calculate the pole L m The corresponding polar angle set and the pole L′ m The cosine similarity of the corresponding polar angle set;

[0100] S46: repeating steps S43 to S45 to obtain the cosine similarities between all three-dimensional point cloud coordinates in the three-dimensional point cloud data and the corresponding three-dimensional coordinates of the building design target, and taking the minimum cosine similarity as the similarity calculation result;

[0101] Construct a coordinate matching set of the 3D point cloud coordinates and the 3D coordinates of the architectural design target:

[0102] {(L m , L′ m )|m∈[1,M]}

[0103] According to the calculation process of the rotation matrix and the translation matrix in step S1, the rotation matrix and the translation matrix between the three-dimensional point cloud coordinates and the three-dimensional coordinates of the building design target are calculated, and the calculation results are used as transformation parameters.

[0104] Optionally, if the similarity calculation result in step S5 is less than a specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, including:

[0105] If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

[0106] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0107] A memory storing at least one instruction;

[0108] Communication interface, enabling electronic equipment to communicate; and

[0109] The processor executes the instructions stored in the memory to implement the above-mentioned digital point cloud correction processing method for the whole process of construction.

[0110] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned method for digital point cloud correction processing of the entire building process.

[0111] Compared with the prior art, the present invention proposes a digital point cloud correction processing method for the entire construction process, which has the following advantages:

[0112] First, this scheme proposes a three-dimensional coordinate mapping and similarity measurement method to perform coordinate transformation on the collected three-dimensional point cloud data, where the coordinate transformation object is the three-dimensional point cloud coordinates in the three-dimensional point cloud data, and the three-dimensional point cloud coordinates are transformed into the world coordinate system. The coordinate system transformation formula for the three-dimensional point cloud coordinates is:

[0113] g′(x,y,z)=Rg(x,y,z) T +H

[0114] Where: R represents the rotation matrix between the current point cloud coordinate system and the world coordinate system, H represents the translation matrix; T represents transposition; g′(x, y, z) represents the coordinate transformation result of the three-dimensional point cloud coordinate g(x, y, z); the calculation process of the rotation matrix and the translation matrix is: use the camera to obtain the three-dimensional point cloud coordinates of N known coordinates in the world coordinate system:

[0115] {(p i ,q i )|i∈[1,N]}

[0116] Where: p i represents the coordinates of the i-th 3D point cloud captured by the camera, q i Indicates p i Coordinates in the world coordinate system; coordinates are normalized:

[0117]

[0118]

[0119] in: They are the standardized results of pi and qi respectively; the standardized coordinates are used to form matrices P and Q respectively:

[0120]

[0121]

[0122] The transformation matrices A and B are calculated:

[0123] A=P T Q

[0124] B=PQ T

[0125] The N eigenvalues ​​and eigenvectors of transformation matrices A and B are calculated respectively, and the eigenvectors are sorted in descending order of eigenvalues. The N eigenvectors of transformation matrix A are:

[0126] (α1, α2, ..., α i , ..., α N )

[0127] Where: α i represents the i-th eigenvector in the transformation matrix A; the N eigenvectors of the transformation matrix B are:

[0128] (β1, β2,...,βi,...,β N )

[0129] Where: β i Represents the i-th eigenvector in the transformation matrix B; the eigenvectors are respectively formed into matrices α and β:

[0130] α=[α1,α2,...,α i , ..., α N ]

[0131] β=[β1, β2,...,βi,...,β N ]

[0132] Calculate the rotation matrix R between the current point cloud coordinate system and the world coordinate system, and the translation matrix H:

[0133] R=αβ T

[0134] H=Q-RP

[0135] The calculated rotation matrix R and translation matrix H are used to transform the 3D point cloud coordinates in the 3D point cloud data into the world coordinate system. The 3D point cloud coordinates in the 3D point cloud data set are converted into several Delaunay triangulations using the Delaunay triangulation method, and the 3D data of the architectural design target are converted into several Delaunay triangulations using the Delaunay triangulation method; the 3D point cloud coordinates in the 3D point cloud data are extracted, where r m The three-dimensional point cloud coordinates in are L m ; Take any three-dimensional point cloud coordinate L m As the vertex, we get Lm Several groups of directly connected points are directly connected to form a three-dimensional point cloud coordinate L m As the pole, L m The direction of the ray to any directly connected point is the polar axis, and the pole L is calculated. m The polar angle of the ray formed by different direct connection points is obtained by the pole L m Corresponding polar angle set; select the three-dimensional data of the architectural design target and the three-dimensional point cloud coordinates L m The closest three-dimensional coordinate point L′ m , with the three-dimensional coordinate point L′ m As the pole, L′ m The direction of the ray to any directly connected point is the polar axis, and the pole L' is calculated m The polar angle of the ray formed by different direct connection points gives the pole L' m Corresponding polar angle set; calculate the pole L m The corresponding polar angle set and the pole L′ m The cosine similarity of the corresponding polar angle set is obtained; the cosine similarity of all three-dimensional point cloud coordinates in the three-dimensional point cloud data and the corresponding three-dimensional coordinates of the architectural design target is obtained, and the minimum cosine similarity is used as the similarity calculation result; based on the three-dimensional point cloud data of the current state of the target building, this scheme establishes a mapping transformation relationship between the pixel coordinates captured by the camera and the world coordinates, and then transforms the three-dimensional point cloud data from the camera coordinate system to the world coordinate system, and adopts the same method to obtain the transformation parameters between the three-dimensional point cloud data and the three-dimensional data of the architectural design target. By combining the polar angle of the three-dimensional point cloud coordinates in the polar coordinate system and the polar angle of the three-dimensional coordinates of the architectural design target, the similarity between the two is calculated to characterize the difference in direction of the three-dimensional data. If the similarity calculation result is less than the specified threshold, it means that there is a large difference between the current building state and the architectural design target. Therefore, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result, so as to realize the digital point cloud correction processing of the whole process of the building.

[0136] At the same time, this scheme proposes a three-dimensional point cloud completion and enhancement method, by constructing a three-dimensional point cloud data enhancement model, the constructed three-dimensional point cloud data enhancement model takes the three-dimensional point cloud data in the world coordinate system as input, and takes the enhanced three-dimensional point cloud data as output; the constructed three-dimensional point cloud data enhancement model includes an input layer, a multi-level encoding layer and a decoding point cloud enhancement layer, wherein the input layer is used to receive the three-dimensional point cloud data, and perform multi-resolution downsampling processing on the received three-dimensional point cloud data, the multi-level encoding layer is used to encode the sampling results of different resolutions to form multi-level encoding features, and the decoding point cloud enhancement layer is used to decode the multi-level encoding features, complete the missing parts of the sampling results at different resolutions, and obtain enhanced three-dimensional point cloud data. This solution samples point cloud data at different resolutions, encodes the sampling results at different resolutions to form multi-level coding features, and then perceives and upsamples the multi-level coding features to obtain the missing point cloud coordinates at different resolutions. The missing point cloud coordinates at different resolutions are combined with the original point cloud coordinates to form a completed and enhanced three-dimensional point cloud coordinate set, thereby realizing the completion and enhancement of missing point cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0137] Figure 1 A schematic diagram of a flow chart of a digital point cloud correction processing method for the entire construction process provided by an embodiment of the present invention;

[0138] Figure 2 A schematic diagram of the structure of an electronic device for implementing a method for correcting and processing digital point cloud throughout a building process, provided by an embodiment of the present invention.

[0139] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0140] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0141] The embodiment of the present application provides a method for correcting and processing digital point clouds throughout the construction process. The execution subject of the method for correcting and processing digital point clouds throughout the construction process includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for correcting and processing digital point clouds throughout the construction process can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0142] Embodiment 1:

[0143] S1: Collect the three-dimensional point cloud data of the current state of the target building, and perform coordinate transformation on the collected three-dimensional point cloud data to obtain the three-dimensional point cloud data in the world coordinate system.

[0144] The three-dimensional point cloud data of the current state of the target building is collected in step S1, including:

[0145] The camera is used to collect the three-dimensional point cloud data of the current state of the target building. The camera consists of an RGB color camera and a depth camera. The RGB color camera is used to obtain the color information of the target building, and the depth camera is used to obtain the depth information of the target building. The image of the target building taken by the camera is I. The formula for converting any pixel I (x, y) in the image I into a three-dimensional point cloud coordinate is:

[0146]

[0147] in:

[0148] g(x, y, z) represents the three-dimensional point cloud coordinates corresponding to the pixel I(x, y) in the target building image, and the pixel I(x, y) represents the pixel in the xth row and yth column in the target building image I;

[0149] c X represents the principal point of the camera in the horizontal direction, c Y Indicates the principal point of the camera in the vertical direction;

[0150] f X Indicates the horizontal focal length of the camera, f Y Indicates the vertical focal length of the camera;

[0151] d x，y Represents the depth information of pixel I(x, y) partially captured by the depth camera;

[0152] An RGB color camera is used to obtain the color value of the pixel in the RGB color channel to form three-dimensional point cloud data, where the three-dimensional point cloud data of pixel I(x, y) is (g(x, y, z), R(x, y), G(x, y), B(x, y)}, where R(x, y), G(x, y), B(x, y) represent the color value of pixel I(x, y) in the RGB color channel respectively.

[0153] In the embodiment of the present invention, the principal point and focal length of the camera are both internal parameters of the camera, that is, known parameter information.

[0154] In the step S1, coordinate transformation is performed on the collected three-dimensional point cloud data to obtain three-dimensional point cloud data in a world coordinate system, including:

[0155] The collected 3D point cloud data is subjected to coordinate transformation, wherein the coordinate transformation object is the 3D point cloud coordinates in the 3D point cloud data, and the 3D point cloud coordinates are transformed into the world coordinate system, wherein the coordinate system transformation formula for the 3D point cloud coordinates is:

[0156] g′(x,y,z)=Rg(x,y,z) T +H

[0157] in:

[0158] R represents the rotation matrix between the current point cloud coordinate system and the world coordinate system, and H represents the translation matrix;

[0159] T stands for transpose;

[0160] g′(x, y, z) represents the coordinate transformation result of the three-dimensional point cloud coordinate g(x, y, z);

[0161] The calculation process of the rotation matrix and the translation matrix is ​​as follows:

[0162] S11: Use the camera to obtain the three-dimensional point cloud coordinates of N known coordinates in the world coordinate system:

[0163] {(p i ,q i )|i∈[1,N]}

[0164] in:

[0165] p i represents the coordinates of the i-th 3D point cloud captured by the camera, q i Indicates p i Coordinates in the world coordinate system;

[0166] S12: Standardize the coordinates:

[0167]

[0168]

[0169] in:

[0170] They are the standardized results of pi and qi respectively;

[0171] S13: The normalized coordinates are used to form matrices P and Q respectively:

[0172]

[0173]

[0174] S14: Calculate the transformation matrices A and B:

[0175] A=P T Q

[0176] B=PQ T

[0177] S15: Calculate the N eigenvalues ​​and eigenvectors of the transformation matrix A and B respectively, and sort the eigenvectors in descending order of the eigenvalues, where the N eigenvectors of the transformation matrix A are:

[0178] (α1, α2, ..., α i , ..., α N )

[0179] in:

[0180] α i represents the i-th eigenvector in the transformation matrix A;

[0181] The N eigenvectors of the transformation matrix B are:

[0182] (β1,β2...,β i , ..., β N )

[0183] in:

[0184] β i represents the i-th eigenvector in the transformation matrix B;

[0185] S16: Construct the eigenvectors into matrix forms α and β respectively:

[0186] α=[α1,α2,...,α i , ..., α N ]

[0187] β=[β1,β2,...,β i , ..., β N ]

[0188] S17: Calculate the rotation matrix R and translation matrix H between the current point cloud coordinate system and the world coordinate system:

[0189] R=αβ T

[0190] H=Q-RP

[0191] The calculated rotation matrix R and translation matrix H are used to transform the three-dimensional point cloud coordinates in the three-dimensional point cloud data into the world coordinate system.

[0192] S2: Construct a 3D point cloud data enhancement model. The constructed model takes the 3D point cloud data in the world coordinate system as input and takes the enhanced 3D point cloud data as output.

[0193] The three-dimensional point cloud data enhancement model is constructed in step S2, including:

[0194] Constructing a 3D point cloud data enhancement model, wherein the constructed 3D point cloud data enhancement model takes the 3D point cloud data in the world coordinate system as input and takes the enhanced 3D point cloud data as output;

[0195] The constructed 3D point cloud data enhancement model includes an input layer, a multi-level encoding layer and a decoding point cloud enhancement layer, wherein the input layer is used to receive 3D point cloud data and perform multi-resolution downsampling processing on the received 3D point cloud data, the multi-level encoding layer is used to encode the sampling results of different resolutions to form multi-level encoding features, and the decoding point cloud enhancement layer is used to decode the multi-level encoding features, complete the missing parts of the sampling results at different resolutions, and obtain enhanced 3D point cloud data;

[0196] The enhancement process based on the 3D point cloud data enhancement model is as follows:

[0197] S21: taking the three-dimensional point cloud data Data in the world coordinate system as input, wherein Data includes D groups of three-dimensional point cloud coordinates in the world coordinate system and color values ​​corresponding to the coordinates;

[0198] S22: The input layer samples the D groups of 3D point cloud coordinates at resolutions of D / 2, D / 4, and D / 8 to obtain three groups of sampling results, where the sampling process for the D / 2 resolution is as follows:

[0199] S221: Select a coordinate from the D group of three-dimensional point cloud coordinates as a starting point, construct a sampling point set, and add the starting point to the sampling point set;

[0200] S222: Calculate the distance from the external coordinates of the sampling point set to the starting point, and select the coordinate with the largest distance to add to the sampling point set;

[0201] S223: Calculate the shortest distance from the external coordinates of the sampling point set to any coordinates in the sampling point set, select the coordinate corresponding to the maximum shortest distance and add it to the sampling point set;

[0202] S224: Repeat steps S222 to S223 until the number of sampling points in the sampling point set reaches

[0203] S23: The multi-layer encoding layer uses a multi-layer perceptron to increase the dimension of each set of sampling results until the dimension of each sampling coordinate reaches 128; the maximum pooling layer is used to extract the feature map with the highest dimension of each set of sampling results. The dimension of each feature map is 128, where the feature map feature under D / 2 resolution sampling is D / 2 for:

[0204] feature D / 2 =Pooling(MLE(sample D / 2 (1))||...||MLE(sample D / 2 (D / 2)))

[0205] in:

[0206] || represents the concatenation of the processing results of the multi-layer perceptron;

[0207] MLE(sample D / 2 (D / 2)) represents the dimension-raising result of the D / 2th sampling result under D / 2 resolution sampling, and MLE represents multi-layer perceptron;

[0208] Pooling (·) represents the pooling operation. For the sampling result under D / 2 resolution sampling, the dimensionality-enhanced 128×3 dimensionality-enhanced result MLE (sample D / 2 (1))||...||MLE(sample D / 2 (D / 2)) pooled to 128 dimensions to obtain the feature map feature D / 2 ;

[0209] S24: The multi-level coding layer uses multiple independent convolution layers to perform convolution processing on the feature maps of the three groups of sampling results, and assigns different weights to the convolution processing results of the three groups of feature maps, and splices the weighted convolution processing results to obtain multi-level coding features; in this embodiment of the present invention, the weight of the feature map under D / 2 resolution sampling is 0.6, the weight of the feature map under D / 4 resolution sampling is 0.2, and the weight of the feature map under D / 8 resolution sampling is 0.2;

[0210] S25: The decoding point cloud enhancement layer contains multiple multi-layer perceptrons with different parameters. The multi-layer perceptron is used to reconstruct and upsample the multi-level coding features to obtain the missing point cloud coordinates at different resolutions, and the missing point cloud coordinates at different resolutions are combined with the original point cloud coordinates to form a set of completed and enhanced three-dimensional point cloud coordinates, where the color value of the missing point cloud coordinate is the average color value of the adjacent three-dimensional point cloud coordinates, to obtain enhanced three-dimensional point cloud data.

[0211] S3: Determine the optimization objective function of the 3D point cloud data enhancement model and perform optimization training to obtain the optimal 3D point cloud data enhancement model.

[0212] In the step S3, the optimization objective function of the three-dimensional point cloud data enhancement model is determined and optimization training is performed, including:

[0213] Determine the optimization objective function of the three-dimensional point cloud data enhancement model, and optimize the training of the three-dimensional point cloud data enhancement model based on the optimization objective function to obtain the trained model parameters, wherein the model parameters to be trained are the multi-layer perceptron parameters in the multi-level encoding layer and the convolution layer parameters θ1, and the multi-layer perceptron parameters θ2 in the decoding point cloud enhancement layer, wherein θ1 and θ2 are parameter vectors;

[0214] Obtain a set of 3D point cloud data with missing point cloud data, and obtain a set of real 3D point cloud coordinates of the missing point cloud data. The determined optimization objective function is:

[0215]

[0216] in:

[0217] Ω represents the real 3D point cloud coordinate set of missing point cloud data, num(Ω) represents the number of missing point cloud data, and num represents the number of missing point clouds obtained by the 3D point cloud data enhancement model;

[0218] loc c Represents the true 3D point cloud coordinates of the cth missing point cloud data;

[0219] Represents the distance loc output by the 3D point cloud data enhancement model c The nearest missing point cloud coordinates;

[0220] dis(·) represents the distance calculation formula;

[0221] ε represents the proportional weight, which is set to 0.2;

[0222] Based on the optimization objective function, the Adam optimizer is used to train the 3D point cloud data enhancement model to obtain the trained model parameters, and then the optimal 3D point cloud data enhancement model is constructed based on the trained model parameters.

[0223] S4: Use the optimal 3D point cloud data enhancement model to enhance the 3D point cloud data after coordinate transformation, perform similarity calculation on the enhanced 3D point cloud data and the target 3D data of architectural design, and calculate the transformation parameters.

[0224] In the step S4, the three-dimensional point cloud data after coordinate transformation is enhanced by using the optimal three-dimensional point cloud data enhancement model, including:

[0225] The three-dimensional point cloud data after the coordinate transformation in step S2 is input into the optimal three-dimensional point cloud data enhancement model, and the optimal three-dimensional point cloud data enhancement model outputs enhanced three-dimensional point cloud data, wherein the enhanced three-dimensional point cloud data set is:

[0226] {r m |m∈[1,M]}

[0227] in:

[0228] r m It represents the mth 3D point cloud data output by the optimal point cloud enhancement model. The 3D point cloud data includes 3D point cloud coordinates and 3D point cloud color values. M represents the number of 3D point cloud data output by the optimal point cloud enhancement model.

[0229] In the step S4, similarity calculation is performed on the enhanced three-dimensional point cloud data and the three-dimensional data of the architectural design target, and transformation parameters are calculated, including:

[0230] The enhanced 3D point cloud data is combined with the 3D data of the architectural design target to perform similarity calculation, and the transformation parameters are calculated. The similarity calculation process is as follows:

[0231] S41: using a Delaunay triangulation method to convert the three-dimensional point cloud coordinates in the three-dimensional point cloud data set into a number of Delaunay triangulation networks, and using the Delaunay triangulation method to convert the three-dimensional data of the architectural design target into a number of Delaunay triangulation networks;

[0232] S42: Extract the three-dimensional point cloud coordinates in the three-dimensional point cloud data, where r m The three-dimensional point cloud coordinates in are L m ;

[0233] S43: Take any three-dimensional point cloud coordinates L m As the vertex, we get L m Several groups of directly connected points are directly connected to form a three-dimensional point cloud coordinate L m As the pole, L m The direction of the ray to any directly connected point is the polar axis, and the pole L is calculated. m The polar angle of the ray formed by different direct connection points is obtained by the pole L m The corresponding polar angle set;

[0234] S44: Select the three-dimensional data of the building design target and the three-dimensional point cloud coordinates L m The closest three-dimensional coordinate point L′ m , with the three-dimensional coordinate point L′ m As the pole, L′ mThe direction of the ray to any directly connected point is the polar axis, and the pole L' is calculated m The polar angle of the ray formed by different direct connection points gives the pole L' m The corresponding polar angle set;

[0235] S45: Calculate the pole L m The corresponding polar angle set and the pole L′ m The cosine similarity of the corresponding polar angle set;

[0236] S46: repeating steps S43 to S45 to obtain the cosine similarities between all three-dimensional point cloud coordinates in the three-dimensional point cloud data and the corresponding three-dimensional coordinates of the building design target, and taking the minimum cosine similarity as the similarity calculation result;

[0237] Construct a coordinate matching set of the 3D point cloud coordinates and the 3D coordinates of the architectural design target:

[0238] {(L m , L′ m )|m∈[1,M]}

[0239] According to the calculation process of the rotation matrix and the translation matrix in step S1, the rotation matrix and the translation matrix between the three-dimensional point cloud coordinates and the three-dimensional coordinates of the building design target are calculated, and the calculation results are used as transformation parameters.

[0240] S5: If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

[0241] If the similarity calculation result in step S5 is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, including:

[0242] If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

[0243] Embodiment 2:

[0244] like Figure 2 , which is a schematic diagram of the structure of an electronic device for implementing a method for correcting and processing digital point cloud of a building throughout the entire process, provided by an embodiment of the present invention.

[0245] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

[0246] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may be used not only to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also to temporarily store data that has been output or is to be output.

[0247] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as the program 12 for realizing the digital point cloud correction processing of the entire building process), and calls the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0248] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0249] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0250] Figure 2 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0251] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0252] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0253] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0254] The program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0255] Collect the 3D point cloud data of the current state of the target building, and perform coordinate transformation on the collected 3D point cloud data to obtain the 3D point cloud data in the world coordinate system;

[0256] Construct a 3D point cloud data enhancement model;

[0257] Determine the optimization objective function of the 3D point cloud data enhancement model and perform optimization training to obtain the optimal 3D point cloud data enhancement model;

[0258] The 3D point cloud data after coordinate transformation is enhanced by using the optimal 3D point cloud data enhancement model, and similarity calculation is performed on the enhanced 3D point cloud data combined with the 3D data of the architectural design target, and the transformation parameters are calculated;

[0259] If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

[0260] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figure 1 to Figure 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0261] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0262] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0263] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A digital point cloud correction processing method for the entire construction process, characterized in that: The method comprises: S1: Collect the 3D point cloud data of the current state of the target building, and perform coordinate transformation on the collected 3D point cloud data to obtain the 3D point cloud data in the world coordinate system; S2: construct a 3D point cloud data enhancement model, the constructed model takes the 3D point cloud data in the world coordinate system as input and takes the enhanced 3D point cloud data as output; The constructed 3D point cloud data enhancement model includes an input layer, a multi-level encoding layer and a decoding point cloud enhancement layer, wherein the input layer is used to receive 3D point cloud data and perform multi-resolution downsampling processing on the received 3D point cloud data, the multi-level encoding layer is used to encode the sampling results of different resolutions to form multi-level encoding features, and the decoding point cloud enhancement layer is used to decode the multi-level encoding features, complete the missing parts of the sampling results at different resolutions, and obtain enhanced 3D point cloud data; S3: Determine the optimization objective function of the 3D point cloud data enhancement model and perform optimization training to obtain the optimal 3D point cloud data enhancement model; S4: using the optimal 3D point cloud data enhancement model to enhance the 3D point cloud data after coordinate transformation, performing similarity calculation on the enhanced 3D point cloud data combined with the architectural design target 3D data, and calculating the transformation parameters; The similarity calculation process is: S41: using a Delaunay triangulation method to convert the three-dimensional point cloud coordinates in the three-dimensional point cloud data set into a number of Delaunay triangulation networks, and using the Delaunay triangulation method to convert the three-dimensional data of the architectural design target into a number of Delaunay triangulation networks; S42: Extract the three-dimensional point cloud coordinates in the three-dimensional point cloud data, where r m The three-dimensional point cloud coordinates in are L m ; S43: Take any three-dimensional point cloud coordinates L m As the vertex, we get L m Several groups of directly connected points are directly connected to form a three-dimensional point cloud coordinate L m As the pole, L m The direction of the ray to any directly connected point is the polar axis, and the pole L is calculated. m The polar angle of the ray formed by different direct connection points is obtained by the pole L m The corresponding polar angle set; S44: Select the three-dimensional data of the building design target and the three-dimensional point cloud coordinates L m The closest three-dimensional coordinate point L′ m , with the three-dimensional coordinate point L′ m As the pole, L′ m The direction of the ray to any directly connected point is the polar axis, and the pole L' is calculated m The polar angle of the ray formed by different direct connection points gives the pole L' m The corresponding polar angle set; S45: Calculate the pole L m The corresponding polar angle set and the pole L′ m The cosine similarity of the corresponding polar angle set; S46: repeating steps S43 to S45 to obtain the cosine similarities between all three-dimensional point cloud coordinates in the three-dimensional point cloud data and the corresponding three-dimensional coordinates of the building design target, and taking the minimum cosine similarity as the similarity calculation result; Construct a coordinate matching set of the 3D point cloud coordinates and the 3D coordinates of the architectural design target: {(L m ,L′ m )|m∈[1,M]} According to the calculation process of the rotation matrix and the translation matrix in step S1, the rotation matrix and the translation matrix between the three-dimensional point cloud coordinates and the three-dimensional coordinates of the architectural design target are calculated, and the calculation results are used as transformation parameters; S5: If the similarity calculation result is less than the specified threshold, the three-dimensional point cloud data of the current building state is corrected based on the transformation parameters, and the current building is corrected accordingly according to the correction result.

2. A method for correcting and processing digital point clouds during the entire construction process as claimed in claim 1, characterized in that: The three-dimensional point cloud data of the current state of the target building is collected in step S1, including: The camera is used to collect the three-dimensional point cloud data of the current state of the target building. The camera consists of an RGB color camera and a depth camera. The RGB color camera is used to obtain the color information of the target building, and the depth camera is used to obtain the depth information of the target building. The image of the target building taken by the camera is I. The formula for converting any pixel I (x, y) in the image I into a three-dimensional point cloud coordinate is: in: g(x, y, z) represents the three-dimensional point cloud coordinates corresponding to the pixel I(x, y) in the target building image, and the pixel I(x, y) represents the pixel in the xth row and yth column in the target building image I; c X represents the principal point of the camera in the horizontal direction, c Y Indicates the principal point of the camera in the vertical direction; f X Indicates the horizontal focal length of the camera, f Y Indicates the vertical focal length of the camera; d x，y Represents the depth information of pixel I(x, y) partially captured by the depth camera; An RGB color camera is used to obtain the color value of the pixel in the RGB color channel to form three-dimensional point cloud data, where the three-dimensional point cloud data of pixel I(x, y) is {g(x, y, z), R(x, y), G(x, y), B(x, y)}, where R(x, y), G(x, y), B(x, y) represent the color value of pixel I(x, y) in the RGB color channel, respectively.

3. A method for correcting and processing digital point clouds during the entire construction process as claimed in claim 2, characterized in that: In the step S1, coordinate transformation is performed on the collected three-dimensional point cloud data to obtain three-dimensional point cloud data in a world coordinate system, including: The collected 3D point cloud data is subjected to coordinate transformation, wherein the coordinate transformation object is the 3D point cloud coordinates in the 3D point cloud data, and the 3D point cloud coordinates are transformed into the world coordinate system, wherein the coordinate system transformation formula for the 3D point cloud coordinates is: g′(x,y,z)=Rg(x,y,z) T +H in: R represents the rotation matrix between the current point cloud coordinate system and the world coordinate system, and H represents the translation matrix; T stands for transpose; g′(x, y, z) represents the coordinate transformation result of the three-dimensional point cloud coordinate g(x, y, z); The calculation process of the rotation matrix and the translation matrix is ​​as follows: S11: Use the camera to obtain the three-dimensional point cloud coordinates of N known coordinates in the world coordinate system: {(p i ,q i )|i∈[1,N]} in: p i represents the coordinates of the i-th 3D point cloud captured by the camera, q i Indicates p i Coordinates in the world coordinate system; S12: Standardize the coordinates: in: They are p i ,q i The standardized results of S13: The normalized coordinates are used to form matrices P and Q respectively: S14: Calculate the transformation matrices A and B: A=P T Q B=PQ T S15: Calculate the N eigenvalues ​​and eigenvectors of the transformation matrix A and B respectively, and sort the eigenvectors in descending order of the eigenvalues, where the N eigenvectors of the transformation matrix A are: (α1，α2，…，α i ，…，α N ) in: α i represents the i-th eigenvector in the transformation matrix A; The N eigenvectors of the transformation matrix B are: (β1，β2，…，β i ，…，β N ) in: β i represents the i-th eigenvector in the transformation matrix B; S16: Construct the eigenvectors into matrix forms α and β respectively: α＝[α1，α2，…，α i ，…，α N ] β＝[β1，β2，…，β i ，…，β N ] S17: Calculate the rotation matrix R between the current point cloud coordinate system and the world coordinate system, and the translation matrix H: R＝αβ T H=Q-RP The calculated rotation matrix R and translation matrix H are used to transform the three-dimensional point cloud coordinates in the three-dimensional point cloud data into the world coordinate system.

4. A method for correcting and processing digital point clouds during the entire construction process as claimed in claim 1, characterized in that: The three-dimensional point cloud data enhancement model is constructed in step S2, including: The enhancement process based on the 3D point cloud data enhancement model is as follows: S21: taking the three-dimensional point cloud data Data in the world coordinate system as input, wherein Data includes D groups of three-dimensional point cloud coordinates in the world coordinate system and color values ​​corresponding to the coordinates; S22: The input layer samples the D groups of 3D point cloud coordinates at resolutions of D / 2, D / 4, and D / 8 to obtain three groups of sampling results, where the sampling process for the D / 2 resolution is as follows: S221: Select a coordinate from the D group of three-dimensional point cloud coordinates as a starting point, construct a sampling point set, and add the starting point to the sampling point set; S222: Calculate the distance from the external coordinates of the sampling point set to the starting point, and select the coordinate with the largest distance to add to the sampling point set; S223: Calculate the shortest distance from the external coordinates of the sampling point set to any coordinates in the sampling point set, select the coordinate corresponding to the maximum shortest distance and add it to the sampling point set; S224: Repeat steps S222 to S223 until the number of sampling points in the sampling point set reaches S23: The multi-layer encoding layer uses a multi-layer perceptron to increase the dimension of each set of sampling results until the dimension of each sampling coordinate reaches 128; the maximum pooling layer is used to extract the feature map with the highest dimension of each set of sampling results. The dimension of each feature map is 128, where the feature map feature under D / 2 resolution sampling is D / 2 for: feature D / 2 =Pooling(MLE(sample D / 2 (1))||...||MLE(sample D / 2 (D / 2))) in: || represents the concatenation of the processing results of the multi-layer perceptron; MLE(sample D / 2 (D / 2)) represents the dimension-raising result of the D / 2th sampling result under D / 2 resolution sampling, and MLE represents multi-layer perceptron; Pooling (·) represents the pooling operation. For the sampling result under D / 2 resolution sampling, the dimensionality-enhanced 128×3 dimensionality-enhanced result MLE (sample D / 2 (1))||...||MLE(sample D / 2 (D / 2)) pooled to 128 dimensions to obtain the feature map feature D / 2 ; S24: The multi-level coding layer uses multiple independent convolution layers to perform convolution processing on the feature maps of the three sets of sampling results, and assigns different weights to the convolution processing results of the three sets of feature maps, and concatenates the weighted convolution processing results to obtain multi-level coding features; S25: The decoding point cloud enhancement layer contains multiple multi-layer perceptrons with different parameters. The multi-layer perceptron is used to reconstruct and upsample the multi-level coding features to obtain the missing point cloud coordinates at different resolutions, and the missing point cloud coordinates at different resolutions are combined with the original point cloud coordinates to form a set of completed and enhanced three-dimensional point cloud coordinates, where the color value of the missing point cloud coordinate is the average color value of the adjacent three-dimensional point cloud coordinates, to obtain enhanced three-dimensional point cloud data.

5. A method for correcting and processing digital point cloud in the whole construction process as claimed in claim 4, characterized in that: In the step S3, the optimization objective function of the three-dimensional point cloud data enhancement model is determined and optimization training is performed, including: Determine the optimization objective function of the three-dimensional point cloud data enhancement model, and optimize the training of the three-dimensional point cloud data enhancement model based on the optimization objective function to obtain the trained model parameters, wherein the model parameters to be trained are the multi-layer perceptron parameters in the multi-level encoding layer and the convolution layer parameters θ1, and the multi-layer perceptron parameters θ2 in the decoding point cloud enhancement layer, wherein θ1 and θ2 are parameter vectors; Obtain a set of 3D point cloud data with missing point cloud data, and obtain a set of real 3D point cloud coordinates of the missing point cloud data. The determined optimization objective function is: in: Ω represents the real 3D point cloud coordinate set of missing point cloud data, num(Ω) represents the number of missing point cloud data, and num represents the number of missing point clouds obtained by the 3D point cloud data enhancement model; loc c Represents the true 3D point cloud coordinates of the cth missing point cloud data; Represents the distance loc output by the 3D point cloud data enhancement model c The nearest missing point cloud coordinates; dis(·) represents the distance calculation formula; ε represents the proportional weight, which is set to 0.2; Based on the optimization objective function, the Adam optimizer is used to train the 3D point cloud data enhancement model to obtain the trained model parameters, and then the optimal 3D point cloud data enhancement model is constructed based on the trained model parameters.

6. A method for correcting and processing digital point clouds during the entire construction process as claimed in claim 1, characterized in that: In the step S4, the three-dimensional point cloud data after coordinate transformation is enhanced by using the optimal three-dimensional point cloud data enhancement model, including: The three-dimensional point cloud data after the coordinate transformation in step S2 is input into the optimal three-dimensional point cloud data enhancement model, and the optimal three-dimensional point cloud data enhancement model outputs enhanced three-dimensional point cloud data, wherein the enhanced three-dimensional point cloud data set is: {r m |m∈[1,M]} in: r m It represents the mth 3D point cloud data output by the optimal point cloud enhancement model. The 3D point cloud data includes 3D point cloud coordinates and 3D point cloud color values. M represents the number of 3D point cloud data output by the optimal point cloud enhancement model.

Images