CAD model construction method based on point cloud data domain reconstruction

Abstract

The invention relates to a CAD model construction method based on point cloud data domain reconstruction, and the method comprises the following steps: extracting the prior information of an ideal CAD model of a to-be-constructed model; obtaining surface point cloud data of the actually manufactured workpiece through three-dimensional scanning equipment; registering the three-dimensional point cloud data with the surface point cloud data of the ideal CAD model; dividing the actual point cloud data into point cloud data belonging to different features according to the curvature change of the point cloud data, and extracting the point cloud data belonging to different regions through point feature template matching; fitting the point cloud data of different areas in the step S4 into a feature curved surface by using an NURBS curved surface; and the feature curved surfaces are connected and recombined according to the same feature adjacent connection relation of the ideal CAD model, and the CAD model based on the three-dimensional actual point cloud data is obtained. According to the method, the modeling efficiency and precision are improved in combination with the prior information of the ideal CAD model.

Description

Technical Field

[0001] The invention belongs to the field of computer-aided design (CAD), and in particular is a method for constructing a CAD model based on regional reconstruction of point cloud data of three-dimensional scanning. Background Art

[0002] In the field of industrial manufacturing, CAD models are the most commonly used and universal models. In order to create a CAD model that reflects the actual shape of the workpiece, it is usually necessary to spend a lot of time and effort based on the measurement data and use manual modeling technology to create the CAD model, but the modeling accuracy is difficult to guarantee. Summary of the invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a CAD model construction method based on domain reconstruction of three-dimensional scanning point cloud data, and to improve the modeling efficiency and accuracy by combining the prior information of the ideal CAD model. The point cloud data acquired by the three-dimensional scanning device is divided into multiple regions, and the point cloud data is reconstructed for each region and converted into a NURBS surface in the CAD model. The CAD models reconstructed in each region are fused and reorganized to create a complete CAD model. In this process, the prior information of the ideal CAD model is combined to improve the modeling efficiency and accuracy.

[0004] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a CAD model construction method based on point cloud data domain reconstruction, comprising the following steps:

[0005] Step S1: for an ideal CAD model of a model to be constructed, extract the prior information of the ideal CAD model, obtain a point feature template by uniform grid sampling on the surface of the ideal CAD model, and discretize the ideal CAD model into point cloud data;

[0006] Step S2: obtaining surface point cloud data of the actual manufactured workpiece through a three-dimensional scanning device, performing a data preprocessing operation, and obtaining three-dimensional point cloud data;

[0007] Step S3: registering the three-dimensional point cloud data obtained in step S2 with the surface point cloud data of the ideal CAD model to obtain actual point cloud data in the same coordinate system as the ideal CAD model;

[0008] Step S4: using the region growing method to segment the actual point cloud data obtained in S3 into point cloud data belonging to different features according to the curvature change of the point cloud data; matching the point cloud data belonging to different regions with the point feature template of the ideal CAD model in S1 according to the position in the coordinate system, and extracting the point cloud data belonging to different regions using the point feature template of the ideal CAD model;

[0009] Step S5: Using NURBS surfaces to fit the point cloud data of different regions in step S4 into characteristic surfaces;

[0010] Step S6: The feature surfaces obtained by fitting the point cloud data in step S5 are connected and reorganized according to the same feature adjacent connection relationship as the ideal CAD model in S1 to obtain a CAD model based on the three-dimensional actual point cloud data.

[0011] Described step S1 comprises the following steps:

[0012] Step S1.1: Extracting prior information of the ideal CAD model: Identify the geometric features of the ideal CAD model by a feature recognition method based on geometric information, and record the position, geometric properties, and Brep boundary of each feature;

[0013] Step S1.2: uniformly sampling grid points on the surface of the ideal CAD model according to different features to obtain multiple grid point feature templates of the ideal CAD models according to the features; using the Monte Carlo method to sample the entire surface of the ideal CAD model to discretize the ideal CAD model into point cloud data.

[0014] Described step S2 comprises the following steps:

[0015] Step S2.1: using a three-dimensional scanning device to collect data on the surface morphology of the actual manufactured workpiece to obtain surface point cloud data of the actual manufactured workpiece;

[0016] Step S2.2: Processing the actual point cloud data obtained by scanning with the three-dimensional scanning device in step S2.1, and filtering out noise points of the point cloud data using a radius filter or a statistical filter method;

[0017] Step S2.3: For the actual point cloud data obtained in step S2.2, use the random sampling method to simplify the data; use the Gaussian filtering method to smooth the data to improve the accuracy of the three-dimensional point cloud data.

[0018] Described step S3 comprises the following steps:

[0019] Step S3.1: First, perform a rough registration of the actual point cloud data and the ideal point cloud data: use the local point pair feature descriptor to determine the initial coordinate transformation matrix through the random consistency sampling method to complete the rough registration of the point cloud data collected by the 3D scanning device;

[0020] Step S3.2: Perform precise registration of the point cloud data collected by three-dimensional scanning: Use the iterative closest point algorithm to calculate the coordinate transformation matrix of the point cloud to obtain precisely registered three-dimensional scanning point cloud data.

[0021] Described step S4 comprises the following steps:

[0022] Step S4.1: Calculate the curvature of the point cloud data based on the relationship between the point and the neighboring points;

[0023] Step S4.2: Randomly specify an initial seed point, specify a growth criterion and the number of iterations, perform regional growth segmentation of the point cloud data, and obtain point cloud data belonging to different regions segmented according to the curvature change;

[0024] Step S4.3: Match the point cloud data belonging to different regions obtained in step S4.2 with the point feature template of the ideal CAD model in step S1.2 according to the coordinate position, and mark the point cloud data of different regions as the feature attributes of the corresponding ideal CAD model;

[0025] Step S4.4: Use the point feature template of the ideal CAD model to extract point cloud data belonging to different areas.

[0026] Described step S4.1 comprises the following steps:

[0027] Step S4.1.1: For each point P, define a neighborhood with P as the center;

[0028] Step S4.1.2: For each point Pi in the neighborhood, calculate its relative coordinates relative to point P, and then construct the covariance matrix C;

[0029] Step S4.1.3: For each point P, based on the covariance matrix C, the principal component analysis method is used to calculate the normal vector of point P, and then the curvature of the point cloud data is obtained.

[0030] The step S6 includes the following steps:

[0031] Step S6.1: Find the adjacent connection relationship of features from the Brep boundary information of the ideal CAD model prior information in step S1.1, and record the adjacent features of each feature;

[0032] Step S6.2: The NURBS surface fitted in step S5.1 is connected and reorganized according to the same feature adjacent connection relationship as the corresponding ideal CAD model in step S6.1 to obtain a CAD model based on the three-dimensional scanning point cloud data.

[0033] The CAD model construction system based on point cloud data domain reconstruction includes:

[0034] The ideal model data acquisition module is used to extract the prior information of the ideal CAD model of the model to be constructed, obtain the point feature template by uniform grid sampling on the surface of the ideal CAD model, and discretize the ideal CAD model into point cloud data;

[0035] The workpiece three-dimensional point cloud data acquisition module is used to obtain the surface point cloud data of the actual manufactured workpiece through a three-dimensional scanning device, perform data preprocessing operations, and obtain three-dimensional point cloud data;

[0036] The point cloud data registration module is used to register the three-dimensional point cloud data with the surface point cloud data of the ideal CAD model to obtain the actual point cloud data in the same coordinate system as the ideal CAD model;

[0037] The point cloud data segmentation module is used to use the regional growing method to segment the actual point cloud data into point cloud data with different features according to the curvature change of the point cloud data; according to the position in the coordinate system, the point cloud data belonging to different regions are matched with the point feature template of the ideal CAD model, and the point cloud data belonging to different regions are extracted using the point feature template of the ideal CAD model;

[0038] The model reconstruction module is used to fit the point cloud data of different areas into feature surfaces using NURBS surfaces; the feature surfaces obtained by fitting the point cloud data are connected and reorganized according to the same feature adjacent connection relationship of the ideal CAD model to obtain a CAD model based on the three-dimensional actual point cloud data.

[0039] A CAD model construction device based on point cloud data domain reconstruction includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the CAD model construction method based on point cloud data domain reconstruction when executing the computer program.

[0040] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the CAD model construction method based on point cloud data domain reconstruction is implemented.

[0041] The present invention has the following beneficial effects and advantages:

[0042] 1. The present invention can improve the efficiency of CAD models based on measured point cloud data, and save a lot of time in creating CAD models by manual modeling technology;

[0043] 2. The present invention divides the point cloud data into multiple areas and reconstructs NURBS surface features respectively to improve the accuracy of the reconstructed features.

[0044] 3. The present invention introduces the prior information of the ideal CAD model through point cloud registration during the modeling process, and combines the prior information of the ideal CAD model in the process of point cloud segmentation and NURBS surface fitting point selection to improve the modeling efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1This is the overall process of the CAD model construction method based on domain reconstruction of three-dimensional scanning point cloud data of the present invention;

[0046] Figure 2 is a principle block diagram of step S3 in an embodiment of the present invention;

[0047] Figure 3 It is a principle block diagram of step S4 in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments, but the scope of protection claimed in the present invention is not limited to the following specific embodiments.

[0049] Example

[0050] This embodiment discloses a CAD model construction method based on domain reconstruction of 3D scanning point cloud data

[0051] like Figure 1 As shown, the CAD model construction method based on domain reconstruction of three-dimensional scanning point cloud data includes the following steps:

[0052] Step S1: For the ideal CAD model of the model to be constructed, extract the prior information of the ideal CAD model. First, mark all the features of the ideal CAD model, record the position, geometric properties and Brep boundary information of each feature; perform uniform grid sampling on the surface of the ideal CAD model according to different features, and obtain several grid point feature templates of the ideal CAD model according to the features; sample the entire surface of the ideal CAD model, and discretize the ideal CAD model into ideal point cloud data;

[0053] Step S2: Obtain surface point cloud data of the actual manufactured workpiece through a three-dimensional scanning device, perform data preprocessing operations (point cloud data denoising, data simplification, and smoothing) to obtain high-precision three-dimensional actual point cloud data.

[0054] Step S3: registering the three-dimensional point cloud data obtained in S2 with the surface point cloud data of the ideal CAD model to obtain actual point cloud data in the same coordinate system as the ideal CAD model;

[0055] Step S4: Use the region growing method to divide the point cloud data obtained in S3 into point cloud data belonging to different features according to the curvature change of the point cloud data; according to the position in the coordinate system, match the point cloud data belonging to different regions with the point feature template of the ideal CAD model in S1, and use the point feature template of the ideal CAD model to extract the point cloud data belonging to different regions.

[0056] Step S5: Use NURBS surfaces to fit the point cloud data of different regions in step S4 into characteristic surfaces.

[0057] Step S6: The feature surfaces obtained by fitting the point cloud data in S5 are connected and reorganized according to the same feature adjacent connection relationship as the ideal CAD model in S1 to obtain a CAD model based on the three-dimensional actual point cloud data.

[0058] Specifically, the step S1 includes the following steps:

[0059] Step S1.1: Extracting the prior information of the ideal CAD model. The geometric features of the ideal CAD model are identified by a feature recognition method based on geometric information, and the position, geometric properties, and Brep boundary of each feature are recorded.

[0060] Specifically, the step S1.1 includes the following steps:

[0061] Step S1.1.1: Use feature recognition methods based on geometric information such as geometric shape and curvature to identify and mark the geometric features of the ideal CAD model. These features include holes, planes, cylinders, etc.

[0062] Step S1.1.2: After the features are identified, the position, geometric attributes and Brep boundary information of each feature are recorded. The position information includes the three-dimensional coordinates of the feature, the geometric attributes include the shape, size, angle, etc. of the feature, and the Brep boundary information includes the parameterized representation of the curve around the feature.

[0063] Step S1.2: uniformly sampling grid points on the surface of the ideal CAD model according to different features to obtain several grid point feature templates of the ideal CAD model according to the features; using the Monte Carlo method to sample the entire surface of the ideal CAD model to discretize the ideal CAD model into point cloud data.

[0064] Specifically, the step S1.2 includes the following steps:

[0065] Step S1.2.1: For each identified feature, uniform grid point sampling is performed according to different features. In each feature area, a uniformly distributed grid point set is generated so that the point cloud data can be mapped to these grid points in subsequent steps.

[0066] Step S1.2.2: Use the Monte Carlo method or other random sampling methods to sample the entire surface of the ideal CAD model and discretize it into point cloud data for subsequent registration of point cloud data collected by a three-dimensional scanning device.

[0067] Described step S2 comprises the following steps:

[0068] Step S2.1: using a three-dimensional scanning device to collect data on the surface morphology of the actual manufactured workpiece to obtain surface point cloud data of the actual manufactured workpiece;

[0069] Step S2.2: Processing the actual point cloud data obtained by scanning with the three-dimensional scanning device in S2.1, and filtering out noise points of the point cloud data using radius filtering and statistical filtering methods;

[0070] Specifically, the step S2.2 includes the following steps:

[0071] Step S2.2.1: Set the filter radius and threshold, and use radius filtering to remove isolated noise points in the point cloud data;

[0072] Step S2.2.2: Set the mean and variance of the filter and use the statistical filtering method to further remove outliers in the point cloud data;

[0073] Step S2.3: For the point cloud data obtained in S2.2, use the random sampling method to simplify the data; use the Gaussian filtering method to smooth the data to improve the accuracy of the three-dimensional point cloud data.

[0074] Specifically, the step S2.3 includes the following steps:

[0075] Step S2.3.1: Set the sampling percentage according to the point cloud data density obtained by discretizing the ideal CAD model in S1, and use the random sampling method to select a part of the points to reduce the data volume of the actual point cloud data.

[0076] Step S2.3.2: Use Gaussian filtering to smooth the point cloud data, eliminate irregularities caused by scanning equipment or environmental noise, reduce spikes and potholes, and improve the overall accuracy of the point cloud data.

[0077] The actual point cloud data in S2 is registered with the ideal point cloud data obtained by discretizing the ideal CAD model in S1.2, and the step S3 includes the following steps:

[0078] Step S3.1: First, a rough registration is performed between the actual point cloud data and the ideal point cloud data. The initial coordinate transformation matrix is ​​determined using a local point pair feature descriptor combined with a random consistency sampling method to complete the rough registration of the point cloud data collected by the 3D scanning device.

[0079] like Figure 2 As shown, specifically, the step S3.1 includes the following steps:

[0080] Step S3.1.1: Calculate the local point pair feature descriptors of all points in the actual point cloud data in S2 and the ideal point cloud data in S1 respectively.

[0081] Step S3.1.2: using a random consistency sampling method to identify point pairs having a corresponding relationship between the actual point cloud data and the ideal point cloud data;

[0082] Step S3.1.3: Based on the point pairs obtained by the random consistency sampling method in step S3.1.2, calculate the coordinate transformation matrix from the actual point cloud data to the ideal point cloud data, and apply the matrix transformation to the actual point cloud data, which is transformed from the coordinate system of the scanning device to the same coordinate system as the ideal point cloud data.

[0083] Step S3.2: Perform precise registration of the point cloud data collected by three-dimensional scanning, and use the iterative closest point (ICP) algorithm to calculate the coordinate transformation matrix of the point cloud precise registration to obtain precisely registered three-dimensional scanning point cloud data.

[0084] Specifically, the step S3.2 includes the following steps:

[0085] Step S3.2.1: Initialize the ICP algorithm and set the convergence condition: the error is less than a certain threshold or the number of iterations reaches a predetermined value;

[0086] Step S3.2.2: For each point in the actual point cloud data, find the point in the ideal point cloud closest to it and establish a point pair matching relationship;

[0087] Step S3.2.3: Use Euclidean distance to measure the distance error between each point pair;

[0088] Step S3.2.4: Assign weights to each point pair using a Gaussian weight function based on the error value;

[0089] Step S3.2.5: Use numerical optimization methods to minimize the error between weighted point pairs and adjust the coordinate transformation matrix between the point pairs;

[0090] Step S3.2.6: Check whether the convergence condition is met. If the convergence condition is not met, execute step S3.2.7. If the convergence condition is met, execute step S3.2.8.

[0091] Step S3.2.7: Continue iterating and update the coordinate transformation matrix;

[0092] Step S3.2.8: Obtain the final coordinate transformation matrix, apply the matrix to the actual point cloud data, and accurately align it to the ideal point cloud data.

[0093] like Figure 3 As shown, the point cloud data obtained in S3 is segmented, and the step S4 includes the following steps:

[0094] Step S4.1: Calculate the curvature of the point cloud data based on the relationship between the point and the neighboring points;

[0095] Specifically, the step S4.1 includes the following steps:

[0096] Step S4.1.1: For each point P, a neighborhood is defined with P as the center;

[0097] Step S4.1.2: For each point Pi in the neighborhood, calculate its relative coordinates relative to point P, and then construct the covariance matrix C;

[0098] Step S4.1.3: For each point P, based on the covariance matrix C, use the principal component analysis (PCA) method to calculate the normal vector;

[0099] Step S4.2: Randomly specify an initial seed point, specify a growth criterion and the number of iterations, perform regional growth segmentation of the point cloud data, and obtain point cloud data belonging to different regions segmented according to the curvature change;

[0100] Specifically, the step S4.2 includes the following steps:

[0101] Step S4.2.1: Randomly select a seed point from the point cloud data as the initial point of segmentation, and give a curvature threshold as the growth criterion;

[0102] Step S4.2.2: For each seed point, calculate the curvature value of the seed point as the curvature reference value of the region;

[0103] Step S4.2.3: Check the neighboring points around the seed point and calculate their curvature values. If the curvature of the neighboring points is similar to that of the seed point (satisfies the growth criterion), the neighboring points are assigned to the same region and marked as processed.

[0104] Step S4.2.4: Repeat step S4.2.3 until there are no more neighboring points that satisfy the growing criterion, or the region no longer grows.

[0105] Step S4.2.5: Select an unprocessed point that is not assigned to any region as the next seed point, and repeat steps S4.2.2, S4.2.3, and S4.2.4 until all points are assigned to a region. Finally, a series of point cloud datasets belonging to different regions are obtained. Each region has similar curvature features, which represent the local geometric properties of the region.

[0106] Step S4.3: Match the point cloud data belonging to different areas obtained in S4.2 with the point feature template of the ideal CAD model in step S1.2 according to the coordinate position, and mark the point cloud data of different areas as the feature attributes (holes, planes, cylinders, etc.) of the corresponding ideal CAD model.

[0107] Specifically, the step S4.3 includes the following steps:

[0108] Step S4.3.1: Perform the following operations for the point cloud data area divided in step S4.2: traverse each area, use the coordinate position relationship to determine the point feature template that best matches the current area, and based on the matching results, mark the current point as the feature attribute of the corresponding ideal CAD model, such as a hole, plane, cylinder, etc.

[0109] Step S4.3.2: Repeat step S4.3.1 until each region is assigned a corresponding characteristic attribute.

[0110] Step S4.4: extracting point cloud data belonging to different regions using the point feature template of the ideal CAD model;

[0111] Specifically, the step S4.4 includes the following steps:

[0112] Step S4.4.1: For each point feature template, perform the following operations: traverse each point in the point feature template, and for each point, calculate its nearest neighbor point in the point cloud data belonging to different features matched in step S4.3.1; use the distance between points to determine which point in the point feature template best matches the current point; replace the point in the current point feature template with the nearest neighbor point in the point cloud data belonging to different features matched, and obtain the point feature template of the actual point cloud data;

[0113] Step S4.4.2: Repeat step S4.4.1 until the actual point cloud is replaced for each point feature template.

[0114] Surface fitting is performed on the domain point cloud data obtained in S4, and the step S5 includes the following steps:

[0115] Step S5.1: Determine the data nodes and control points according to the point cloud data extracted in S4.4, and use the NURBS surface for fitting.

[0116] Specifically, the step S5.1 includes the following steps:

[0117] Step S5.1.1: Determine the data nodes and control points for constructing the NURBS surface according to the point feature template of the actual CAD model extracted in step S4.4.1;

[0118] Step S5.1.2: Initialize the NURBS surface using the data nodes and initial control points;

[0119] Step S5.1.3: Use a numerical optimization method to adjust the position of the control points to minimize the distance error between the point cloud data and the NURBS surface, and generate the NURBS surface by connecting the control points;

[0120] Step S5.1.4: Repeat steps S5.1.1 to S5.1.3 to construct a corresponding NURBS surface for each point feature template of the actual point cloud data;

[0121] The NURBS surface obtained in step S5 is reorganized according to the Brep boundary information of the ideal CAD model in step S1.1, and step S6 includes the following steps:

[0122] Step S6.1: Find the adjacent connection relationship of the features from the Brep boundary information of the ideal CAD model in S1.1, and record the adjacent features of each feature;

[0123] Specifically, the step S6.1 includes the following steps:

[0124] Step S6.1.1: traverse each feature in the ideal CAD model, and for each feature, check its Brep boundary information obtained in step S1.1, and find and determine the adjacent features by checking whether the boundary curves share endpoints or intersect.

[0125] Step S6.1.2: Use an association list to store and record the type of each feature (eg, hole, plane, cylinder, etc.) and its adjacent features.

[0126] Step S6.1.3: Repeat the above steps S6.1.1 to S6.1.2 until the information of adjacent features is recorded for each feature.

[0127] Step S6.2: The NURBS surface fitted in step S5.1 is connected and reorganized according to the same feature adjacent connection relationship as the corresponding ideal CAD model in step S6.1 to obtain a CAD model based on the three-dimensional scanning point cloud data.

[0128] Step S6.2.1: traverse the neighbor relationship of each feature of the ideal CAD model, and for each pair of adjacent features, determine which boundaries are shared between them;

[0129] Step S6.2.2: For features with shared boundaries, their NURBS surfaces are merged along the shared boundaries to ensure a smooth transition between the surfaces;

[0130] Step S6.2.3: Repeat the above process until each pair of adjacent features is connected through a shared boundary. By connecting and reassembling the NURBS surfaces, a complete CAD model is obtained.

Claims

1. A CAD model construction method based on point cloud data domain reconstruction, characterized in that: The following steps are involved: Step S1: for an ideal CAD model of a model to be constructed, extract the prior information of the ideal CAD model, obtain a point feature template by uniform grid sampling on the surface of the ideal CAD model, and discretize the ideal CAD model into point cloud data; Step S2: obtaining surface point cloud data of the actual manufactured workpiece through a three-dimensional scanning device, performing a data preprocessing operation, and obtaining three-dimensional point cloud data; Step S3: registering the three-dimensional point cloud data obtained in step S2 with the surface point cloud data of the ideal CAD model to obtain actual point cloud data in the same coordinate system as the ideal CAD model; Step S4: using the region growing method to segment the actual point cloud data obtained in S3 into point cloud data belonging to different features according to the curvature change of the point cloud data; matching the point cloud data belonging to different regions with the point feature template of the ideal CAD model in S1 according to the position in the coordinate system, and extracting the point cloud data belonging to different regions using the point feature template of the ideal CAD model; Step S5: Using NURBS surfaces to fit the point cloud data of different regions in step S4 into characteristic surfaces; Step S6: The feature surfaces obtained by fitting the point cloud data in step S5 are connected and reorganized according to the same feature adjacent connection relationship as the ideal CAD model in S1 to obtain a CAD model based on the three-dimensional actual point cloud data.

2. The CAD model construction method based on point cloud data domain reconstruction according to claim 1 is characterized in that: The step S1 comprises the following steps: Step S1.1: Extracting prior information of the ideal CAD model: Identify the geometric features of the ideal CAD model by a feature recognition method based on geometric information, and record the position, geometric properties, and Brep boundary of each feature; Step S1.2: uniformly sampling grid points on the surface of the ideal CAD model according to different features to obtain multiple grid point feature templates of the ideal CAD models according to the features; using the Monte Carlo method to sample the entire surface of the ideal CAD model to discretize the ideal CAD model into point cloud data.

3. The CAD model construction method based on point cloud data domain reconstruction according to claim 1 is characterized in that: Described step S2 comprises the following steps: Step S2.1: using a three-dimensional scanning device to collect data on the surface morphology of the actual manufactured workpiece to obtain surface point cloud data of the actual manufactured workpiece; Step S2.2: Processing the actual point cloud data obtained by scanning with the three-dimensional scanning device in step S2.1, and filtering out noise points of the point cloud data using a radius filter or a statistical filter method; Step S2.3: For the actual point cloud data obtained in step S2.2, use the random sampling method to simplify the data; use the Gaussian filtering method to smooth the data to improve the accuracy of the three-dimensional point cloud data.

4. The CAD model construction method based on point cloud data domain reconstruction according to claim 1 is characterized in that: Described step S3 comprises the following steps: Step S3.1: First, perform a rough registration of the actual point cloud data and the ideal point cloud data: use the local point pair feature descriptor to determine the initial coordinate transformation matrix through the random consistency sampling method to complete the rough registration of the point cloud data collected by the 3D scanning device; Step S3.2: Perform precise registration of the point cloud data collected by three-dimensional scanning: Use the iterative closest point algorithm to calculate the coordinate transformation matrix of the point cloud to obtain precisely registered three-dimensional scanning point cloud data.

5. The CAD model construction method based on point cloud data domain reconstruction according to claim 1, characterized in that: Described step S4 comprises the following steps: Step S4.1: Calculate the curvature of the point cloud data based on the relationship between the point and the neighboring points; Step S4.2: Randomly specify an initial seed point, specify a growth criterion and the number of iterations, perform regional growth segmentation of the point cloud data, and obtain point cloud data belonging to different regions segmented according to the curvature change; Step S4.3: Match the point cloud data belonging to different regions obtained in step S4.2 with the point feature template of the ideal CAD model in step S1.2 according to the coordinate position, and mark the point cloud data of different regions as the feature attributes of the corresponding ideal CAD model; Step S4.4: Use the point feature template of the ideal CAD model to extract point cloud data belonging to different areas.

6. The CAD model construction method based on point cloud data domain reconstruction according to claim 5 is characterized in that: Described step S4.1 comprises the following steps: Step S4.1.1: For each point P, define a neighborhood with P as the center; Step S4.1.2: For each point Pi in the neighborhood, calculate its relative coordinates relative to point P, and then construct the covariance matrix C; Step S4.1.3: For each point P, based on the covariance matrix C, the principal component analysis method is used to calculate the normal vector of point P, and then the curvature of the point cloud data is obtained.

7. The CAD model construction method based on point cloud data domain reconstruction according to claim 1, characterized in that: The step S6 includes the following steps: Step S6.1: Find the adjacent connection relationship of features from the Brep boundary information of the ideal CAD model prior information in step S1.1, and record the adjacent features of each feature; Step S6.2: The NURBS surface fitted in step S5.1 is connected and reorganized according to the same feature adjacent connection relationship as the corresponding ideal CAD model in step S6.1 to obtain a CAD model based on the three-dimensional scanning point cloud data.

8. A CAD model construction system based on point cloud data domain reconstruction, characterized in that: include: The ideal model data acquisition module is used to extract the prior information of the ideal CAD model of the model to be constructed, obtain the point feature template by uniform grid sampling on the surface of the ideal CAD model, and discretize the ideal CAD model into point cloud data; The workpiece three-dimensional point cloud data acquisition module is used to obtain the surface point cloud data of the actual manufactured workpiece through a three-dimensional scanning device, perform data preprocessing operations, and obtain three-dimensional point cloud data; The point cloud data registration module is used to register the three-dimensional point cloud data with the surface point cloud data of the ideal CAD model to obtain the actual point cloud data in the same coordinate system as the ideal CAD model; The point cloud data segmentation module is used to use the regional growing method to segment the actual point cloud data into point cloud data with different features according to the curvature change of the point cloud data; according to the position in the coordinate system, the point cloud data belonging to different regions are matched with the point feature template of the ideal CAD model, and the point cloud data belonging to different regions are extracted using the point feature template of the ideal CAD model; The model reconstruction module is used to fit the point cloud data of different areas into feature surfaces using NURBS surfaces; the feature surfaces obtained by fitting the point cloud data are connected and reorganized according to the same feature adjacent connection relationship of the ideal CAD model to obtain a CAD model based on the three-dimensional actual point cloud data.

9. A CAD model construction device based on point cloud data domain reconstruction, characterized in that: It comprises a memory and a processor; the memory is used to store a computer program; the processor is used to implement the CAD model construction method based on point cloud data domain reconstruction as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the CAD model construction method based on point cloud data domain reconstruction as described in any one of claims 1 to 7 is implemented.

Images