Method, device and equipment for post-processing of facemesh structured light point cloud data

Abstract

The application discloses a kind of face scanning structured light point cloud data post-processing method, device and equipment, it is related to three-dimensional vision measurement and point cloud data processing technical field, the method includes obtaining the multiple frame point cloud data of target measurement object and each frame point cloud data obtained is carried out outlier processing, multiple frame point cloud data after outlier processing is converted to the same coordinate system;Fusion is converted to the same coordinate system under multiple frame point cloud data, obtain after fusion point cloud data;Normal vector estimation is carried out to after fusion point cloud data, geometric feature identification, and noise reduction processing and sampling optimization, complete the post-processing of point cloud data.The application can effectively improve the processing efficiency of face scanning structured light point cloud data.

Description

Technical Field

[0001] This application relates to the field of 3D vision measurement and point cloud data processing technology, specifically to a post-processing method, apparatus, and equipment for surface-scanned structured light point cloud data. Background Technology

[0002] With the rapid development of 3D vision measurement technology, surface scanning structured light 3D measurement technology has been widely used in industrial quality control, reverse engineering, and cultural heritage protection due to its advantages such as high precision, high efficiency, and non-contact operation. This technology projects a specific grating pattern onto the surface of the object being measured, captures the deformed grating image using a camera, and obtains dense point cloud data of the object's surface through phase calculation and 3D reconstruction algorithms.

[0003] However, in practical applications, the raw point cloud data acquired by the area scanning structured light system faces the following technical challenges: due to factors such as ambient light interference, differences in surface reflection characteristics, and occlusion, there are a large number of outliers and noise points in the raw point cloud. Errors in the measurement system itself and imperfect calibration lead to systematic biases in the point cloud data. Traditional noise reduction methods often over-smooth the data, resulting in the loss of important geometric feature details. The noise patterns unique to structured light point clouds require targeted processing strategies.

[0004] Industrial inspection and other applications demand high real-time performance. Existing post-processing workflows are time-consuming, fragmented, and lack seamless integration between stages, as well as dedicated algorithms specific to the characteristics of area-scanned structured light data. Therefore, developing a highly efficient post-processing method specifically tailored to the characteristics of area-scanned structured light point cloud data, which can significantly improve processing efficiency while maintaining processing accuracy, and meet the needs of practical industrial applications, has become an urgent problem to be solved. Summary of the Invention

[0005] This application provides a post-processing method, apparatus, and device for area-scanned structured light point cloud data, which can effectively improve the processing efficiency of area-scanned structured light point cloud data.

[0006] In a first aspect, embodiments of this application provide a post-processing method for area-scanned structured light point cloud data, the post-processing method for area-scanned structured light point cloud data including: Acquire multi-frame point cloud data of the target measurement object and perform outlier removal processing on each frame of point cloud data. Then, transform the multi-frame point cloud data after outlier removal to the same coordinate system. Multiple frames of point cloud data, transformed to the same coordinate system, are fused to obtain fused point cloud data. The fused point cloud data undergoes normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization to complete post-processing of the point cloud data.

[0007] In conjunction with the first aspect, in one implementation, the step of acquiring multi-frame point cloud data of the target measurement object and performing outlier removal processing on each acquired frame of point cloud data specifically includes: The target object is scanned by a surface scanning sensor to obtain multiple frames of point cloud data; Based on the plane fitting method, outlier removal is performed on each frame of point cloud data to identify and remove noise points.

[0008] In conjunction with the first aspect, in one implementation, the outlier removal process is performed on each frame of point cloud data obtained based on the plane fitting method. Specifically, the outlier removal process for the current frame of point cloud data includes: A KDTree search structure is established for the current frame point cloud data. The current point in the current frame point cloud data is recorded as the target point. Plane fitting is performed on all points in the neighborhood of the target point to obtain the fitting plane. The average distance from all points in the neighborhood of the target point to the fitted plane, and the distance from the target point to the fitted plane are calculated. For each point in the target point's neighborhood, perform a nearest neighbor search and then sequentially determine the neighborhood of each point in the target point's neighborhood. If the neighborhood of the current point in the target point's neighborhood contains the target point, then determine that the current point and the target point are neighboring mutually beneficial points in the target point's neighborhood, and obtain the total number of neighboring mutually beneficial points corresponding to the target point. Based on the total number of neighboring mutually beneficial points corresponding to the target point, and the calculated distance information corresponding to the target point, the criterion value of the target point is calculated:

[0009] in, The criterion value representing the target point, when If the value exceeds the outlier threshold, the target point is considered a noise point and is removed; otherwise, no action is taken. This represents the distance from the target point to the fitted plane. This represents the average distance from all points in the neighborhood of the target point to the fitted plane. This indicates the total number of neighboring reciprocal points. This represents the total number of points in the target area.

[0010] In conjunction with the first aspect, in one implementation, when processing outlier removal from the current frame point cloud data, the method further includes: The outlier removal process is repeated a preset number of times on the current frame point cloud data. Points in the neighborhood of the noise points detected during the current outlier removal process are marked. During the next outlier removal process, only the marked points are checked to see if they are noise points.

[0011] In conjunction with the first aspect, in one implementation, the step of converting the processed multi-frame point cloud data to the same coordinate system specifically includes: Based on the relative positional relationship, the attitude matrix of each point cloud data relative to the reference coordinate system is obtained; Based on the attitude matrix, the multi-frame point cloud data is transformed to the reference coordinate system.

[0012] In conjunction with the first aspect, in one implementation, the fusion of multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data specifically includes: Establish a spatial search structure octree, set the minimum cube grid according to the resolution of single frame point cloud data, and insert multiple frames of point cloud data into the spatial search structure octree. Check in turn whether there is a point in each of the smallest cubes. If there is no point in the current smallest cube, do nothing. If there is a point in the current smallest cube, average all the points in the current smallest cube and calculate the centroid position. Use the centroid to replace all the points in the current smallest cube as the unique point of the current smallest cube. Based on the unique point of all the smallest cubes, the point cloud data of multiple frames is fused to obtain the fused point cloud data.

[0013] In conjunction with the first aspect, in one implementation, the step of performing normal vector estimation, geometric feature recognition, noise reduction processing, and sampling optimization on the fused point cloud data, wherein the normal vector estimation and geometric feature recognition specifically include: Based on PCA estimation, the eigenvalue and eigenvector of each point in the fused point cloud data are calculated. The normal vector of the fused point cloud data is obtained based on the point corresponding to the minimum eigenvalue. Calculate the normal angles of the current point and its neighboring points in the fused point cloud data based on the normal vectors, and calculate the standard deviation of the angles based on the normal angles of all neighboring points of the current point in the fused point cloud data. The calculated angle standard deviation is compared with the angle standard deviation threshold. If it is greater than the threshold, the current point in the fused point cloud data is determined to be a feature point.

[0014] In conjunction with the first aspect, in one implementation, the step of performing normal vector estimation, geometric feature recognition, noise reduction processing, and sampling optimization on the fused point cloud data, wherein the noise reduction processing and sampling optimization specifically include: Based on the established spatial search structure octree, the minimum cube grid is set again according to the preset multiple resolution of the single frame point cloud data. For the newly set minimum cube, determine whether there are feature points in the current minimum cube. If not, use the center point of the current minimum cube to replace all points within the current minimum cube. If they exist, do not process them. Search for neighborhood points using the replaced points in the current smallest cube, use Gaussian filtering or moving least squares method to smooth the point cloud for denoising, recalculate the position of the points, and complete the denoising process and sampling optimization.

[0015] Secondly, embodiments of this application provide a post-processing device for area-scanned structured light point cloud data, the post-processing device for area-scanned structured light point cloud data comprising: The conversion module is used to acquire multi-frame point cloud data of the target measurement object and perform outlier removal processing on each frame of point cloud data, and convert the multi-frame point cloud data after outlier removal to the same coordinate system. The fusion module is used to fuse multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data. The processing module is used to perform normal vector estimation, geometric feature recognition, noise reduction and sampling optimization on the fused point cloud data to complete the post-processing of the point cloud data.

[0016] Thirdly, embodiments of this application provide a post-processing device for area-scanned structured light point cloud data. The post-processing device for area-scanned structured light point cloud data includes a processor, a memory, and a post-processing program for area-scanned structured light point cloud data stored in the memory and executable by the processor. When the post-processing program for area-scanned structured light point cloud data is executed by the processor, it implements the steps of the post-processing method for area-scanned structured light point cloud data described above.

[0017] The beneficial effects of the technical solutions provided in this application include: (1) It has significant advantages in the synergy between noise suppression and data fusion: Traditional methods usually process denoising, registration and fusion as independent steps in sequence, which leads to the accumulation of errors between each step. This application uses an innovative iterative hierarchical outlier detection mechanism to first accurately identify multi-level outliers using local geometric criteria based on plane fitting and mutual nearest neighbors, and then combines a neighborhood labeling strategy to achieve efficient iterative removal. This high-precision denoising process in the pre-process fundamentally reduces the cumulative error amplification effect caused by noise points when fusing multiple frames of point cloud data in the subsequent process, and provides a high-quality data foundation for the subsequent process. Compared with conventional statistical filtering or radius filtering methods, this application is more targeted and thorough in identifying outliers while maintaining the integrity of the feature structure. (2) A breakthrough in the integration of data fusion and feature extraction processes: In existing methods, point cloud fusion often adopts simple voxel lattice centroidalization or direct splicing. Normal estimation and feature recognition are usually treated as independent post-processing steps, resulting in fragmented processes and computational redundancy. This application creatively couples octree multi-resolution fusion with feature point recognition based on normal distribution. In the fusion stage, the original details of the data are preserved through the fine octree structure of the first resolution. In the feature recognition stage, PCA normal estimation and normal angle standard deviation analysis are innovatively combined to effectively distinguish between real geometric feature points and curvature anomalies caused by noise. This overcomes the shortcomings of the traditional single curvature threshold method in terms of reliability in noisy environments and achieves a balance between data compression and feature preservation. (3) Improved efficiency in parallel processing of simplification, noise reduction and feature preservation: Traditional methods usually treat point cloud simplification (downsampling) and smoothing noise reduction as two independent and sequential steps, which is not only inefficient, but also prone to losing features during simplification or excessive smoothing during noise reduction. This application proposes a synchronous simplification and noise reduction method based on dual-resolution octree. This method uses a second-resolution grid to intelligently aggregate and replace non-feature regions, which effectively reduces the amount of data. At the same time, only the points that need to be adjusted (such as the new points after replacement) are subjected to neighborhood-based bilateral filtering or moving least squares smoothing, and the regions where the identified feature points are located are protected from the smoothing operation. This integrated strategy of identification-protection-processing completes simplification, feature-preserving noise reduction and point cloud optimization in a single traversal, which greatly improves the processing efficiency and fundamentally solves the contradiction between simplification and noise reduction, and between noise reduction and feature preservation. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the post-processing method for surface-scanned structured light point cloud data in this application. Figure 2 This is a schematic diagram of planar fitting; Figure 3 This is a schematic diagram of outlier detection; Figure 4 For simplification and noise reduction purposes, please refer to the diagram. Figure 5 This is a schematic diagram of the functional modules of the post-processing device for surface-scanned structured light point cloud data in this application. Figure 6 This is a schematic diagram of the hardware structure of the post-processing device for the surface-scanned structured light point cloud data of this application. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0021] In a first aspect, embodiments of this application provide a post-processing method for area-scanned structured light point cloud data, applicable to high-precision 3D point cloud data processing and optimization in fields such as industrial inspection, reverse engineering, cultural relic digitization, and medical imaging. This method addresses the current problems of isolated processing flows, low efficiency, and difficulty in maintaining feature integrity. By constructing a fully optimized processing flow with deep collaboration between links, it uses plane fitting and the principle of mutual nearest neighbor to remove outliers based on weighted criteria, selects the denoising level according to the noise level, iterates denoising, and balances effectiveness and speed. At the same time, it optimizes the feature point recognition algorithm to improve recognition accuracy, and combines downsampling and denoising processing to improve processing efficiency.

[0022] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the post-processing method for surface-scanned structured light point cloud data in this application. Figure 1 As shown, the post-processing method for area scan structured light point cloud data includes: S1: Acquire multi-frame point cloud data of the target object and perform outlier removal processing on each frame of point cloud data, and transform the multi-frame point cloud data after outlier removal to the same coordinate system. S2: Fuse multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data; S3: Perform normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization on the fused point cloud data to complete the post-processing of the point cloud data.

[0023] Specifically, firstly, multiple frames of raw point cloud data of the target object are acquired, and each frame is processed to remove outliers and filter out noise. Secondly, the outlier-removed multi-frame point cloud data is transformed to the same coordinate system using a high-precision registration algorithm. Then, the registered multi-frame point cloud data is fused to generate a single, high-density 3D point cloud model, i.e., the fused point cloud data. Next, the normal vectors of the fused point cloud data are uniformly estimated, and geometric features are identified and extracted based on this. Finally, feature-preserving denoising and adaptive sampling optimization are performed on the point cloud data to smooth noise while preserving the sharp edges and detailed features of the model to the maximum extent. This application significantly improves overall processing efficiency by integrating discrete processing steps into a tightly connected collaborative process. By introducing a feature-preserving mechanism in the backend fusion, normal estimation, and sampling optimization, the contradiction between denoising and detail loss is effectively resolved. This application is effectively applicable to industrial online inspection, reverse engineering of precision parts, and other application scenarios that require high accuracy and efficiency.

[0024] Furthermore, in one embodiment, acquiring multiple frames of point cloud data of the target measurement object and performing outlier removal processing on each acquired frame of point cloud data specifically includes: S101: Scans the target object using a surface scanning sensor to obtain multiple frames of point cloud data; S102: Based on the plane fitting method, outlier removal is performed on each frame of point cloud data to identify and remove noise points.

[0025] Specifically, the target object is scanned by a surface scanning sensor to acquire multiple frames of raw point cloud data. Assuming there are n frames of point cloud data, due to noise or errors during the scanning process, outlier removal is performed on these n frames of point cloud data. Outlier removal mainly identifies points that are far from the correct surface. If these points are not removed, cumulative error noise will be introduced when fusing these n frames of point cloud data, affecting the fusion quality and leading to an increase in measurement error.

[0026] Specifically, based on a plane fitting method, outlier removal is performed on each frame of point cloud data. The outlier removal process for the current frame of point cloud data specifically includes: S1021: Establish a KDTree search structure for the current frame point cloud data, record the current point in the current frame point cloud data as the target point, and perform plane fitting on all points in the neighborhood of the target point to obtain the fitting plane. Specifically, taking the outlier removal process of a single frame point cloud data as an example, first, a KDTree search structure is established for the current frame point cloud data. The purpose is to quickly find the neighborhood points. For ease of description, the current point in the current frame point cloud data is recorded as the target point. Then, the neighborhood of the target point is found, that is, the nearest neighbor point of the target point. Then, plane fitting is performed on all points in the neighborhood of the target point to obtain the fitting plane. S1022: Calculate the average distance from all points in the neighborhood of the target point to the fitting plane, and the distance from the target point to the fitting plane; Specifically, the distances from all points in the neighborhood of the target point to the fitting plane are calculated, and the average distance is obtained by averaging the calculated distances. At the same time, the distance from the target point to the fitting plane is also required. S1023: Perform nearest neighbor search on each point in the target point's neighborhood, and then judge the neighborhood of each point in the target point's neighborhood in turn. If the neighborhood of the current point in the target point's neighborhood contains the target point, then determine that the current point and the target point in the target point's neighborhood are neighboring mutually beneficial points, and obtain the total number of neighboring mutually beneficial points corresponding to the target point. Specifically, a nearest neighbor search is also performed on each point in the target point's neighborhood. If the current point's neighborhood contains the target point, it means that the current point and the target point are neighboring mutually beneficial. If the current point's neighborhood does not contain the target point, it means that the current point and the target point are not neighboring mutually beneficial. S1024: Calculate the criterion value of the target point based on the total number of neighboring mutually beneficial points corresponding to the target point and the calculated distance information corresponding to the target point:

[0027] in, The criterion value representing the target point. Used to determine whether a target point is a noise point, when If the value is greater than the outlier threshold, the target point is considered a noise point (outlier) and is removed; otherwise, no action is taken (i.e., the target point is not a noise point). This represents the distance from the target point to the fitted plane. This represents the average distance from all points in the neighborhood of the target point to the fitted plane. This indicates the total number of neighboring reciprocal points. This represents the total number of points in the target area.

[0028] It should be noted that since outlier removal is required for each frame of point cloud data, and the outlier removal method is the same for each frame of point cloud data, the above steps S1021~S1024 can be used to achieve outlier removal for each frame of point cloud data.

[0029] The following combination Figure 2 A detailed explanation of plane fitting is provided. Figure 2 In the diagram, orange dots represent target points, green dots represent neighborhood points of the target points, and H represents the fitted plane obtained by plane fitting. The calculated criterion value of the target point is compared with the outlier threshold to determine whether the target point is a noise point. The outlier threshold can be between 0 and 1, and the specific value is determined by the attributes of the point cloud reconstructed by the sensor itself.

[0030] Furthermore, the outlier removal process for the current frame point cloud data also includes: The outlier removal process is repeated a preset number of times on the current frame point cloud data. Points in the neighborhood of the noise points detected during the current outlier removal process are marked. During the next outlier removal process, only the marked points are checked to see if they are noise points.

[0031] Specifically, outliers are typically clustered together. When the most influential outlier is removed, the remaining outliers may not be identified. Therefore, iterative processing is required. Outliers usually occur in groups. Recording the neighborhood points of the first identified outlier and marking them allows subsequent processing to focus on detecting only these points, significantly improving detection efficiency. For example, see... Figure 3 As shown, red dots represent target points, and yellow and green dots represent neighborhood points of the target points. If a red dot is identified as an outlier, the green dots in the neighborhood are marked. In the next outlier removal process, only green dots are identified.

[0032] Based on the noise level, the outliers identified in the first outlier removal process are recorded as first-level outliers, and so on, with the second outlier being the second-level outlier. Typically, the maximum number of iterations does not exceed five. Following the above processing flow, the outlier removal process for multi-frame point cloud data is completed, resulting in high-quality multi-frame point cloud data to be fused.

[0033] Furthermore, in one embodiment, the processed multi-frame point cloud data is transformed to the same coordinate system, specifically including: S111: Based on the relative positional relationship, obtain the attitude matrix of each point cloud data relative to the reference coordinate system; S112: Based on the attitude matrix, transform the multi-frame point cloud data to the reference coordinate system.

[0034] Specifically, for multi-frame point cloud data after outlier removal, their coordinate systems are not in the same coordinate system. Therefore, they are transformed to the same coordinate system. Depending on the measurement method, there are axis rotation measurement, marker point measurement, etc. The attitude matrix of each frame of point cloud data relative to the reference coordinate system can be obtained according to the actual relative position relationship. All these multi-frame point cloud data are transformed to the reference coordinate system to complete the data stitching.

[0035] Furthermore, in one embodiment, fusing multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data specifically includes: S201: Establish a spatial search structure octree, set the minimum cube grid according to the resolution of single frame point cloud data, and insert multiple frames of point cloud data into the spatial search structure octree. S202: Sequentially determine whether there is a point in each minimum cube. If there is no point in the current minimum cube, do not process it. If there is a point in the current minimum cube, average the weights of all points in the current minimum cube and calculate the centroid position. Use the centroid to replace all points in the current minimum cube as the unique point of the current minimum cube. S203: Based on the unique point of all the smallest cubes, fuse the point cloud data of multiple frames to obtain the fused point cloud data.

[0036] Specifically, for multi-frame point cloud data transformed to the same coordinate system, a spatial search structure octree is first established. A minimum cube grid is set based on the resolution of a single-frame point cloud data (also called the first resolution). All multi-frame point cloud data are inserted into the spatial search structure octree. The point cloud data inserted into the spatial search structure octree only needs to ensure that it has a unique coordinate within the minimum cube grid. If there are no points in the current minimum cube, it proves that this is not a measurement location, and this minimum cube does not need to be managed. If it is a measurement location, that is, there are points in the current minimum cube, assuming there are m points in the current minimum cube, only the average weight of these m points needs to be calculated to determine its centroid position. This centroid is then used to replace all points in the current minimum cube. Therefore, all minimum cubes are ultimately in a state with or without unique points, completing the fusion of multi-frame point cloud data and obtaining the initial state point cloud after fusion, i.e., the fused point cloud data.

[0037] Furthermore, in one embodiment, the fused point cloud data undergoes normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization. Specifically, the normal vector estimation and geometric feature recognition include: S301: Based on PCA (Principal Component Analysis) estimation, the eigenvalues ​​and eigenvectors of each point in the fused point cloud data are calculated. The normal vector of the fused point cloud data is obtained based on the point corresponding to the minimum eigenvalue. Specifically, since the initial fused point cloud data does not have accurate normal vectors, it is necessary to perform initial normal vector estimation. By using fast PCA estimation, the eigenvalues ​​and eigenvectors of each point in the fused point cloud data are calculated, and the smallest eigenvalue corresponds to its normal vector. The original estimated normal vector may be affected by noise, so anisotropic filtering can be used to filter the normal vector again to obtain more accurate normal vector information. S302: Calculate the normal angles of the current point and its neighborhood points in the fused point cloud data based on the normal vectors, and calculate the standard deviation of the angles based on the normal angles of all neighborhood points of the current point in the fused point cloud data; S303: Compare the calculated angle standard deviation with the angle standard deviation threshold. If it is greater than the threshold, then the current point in the fused point cloud data is determined to be a feature point.

[0038] Specifically, although the curvature of each point can be calculated using eigenvalues, the actual curvature is affected by point cloud noise. Therefore, both feature points and noise points have large curvatures, making it difficult to distinguish between them. Based on this consideration, the normal vector information obtained above is used to calculate the normal angle between the current point and its neighborhood points in the fused point cloud data. Then, the standard deviation of the angles is calculated based on the normal angles of all neighborhood points. Feature points are further filtered based on the standard deviation of the angles. The threshold value of the standard deviation of the angles ranges from 1 to 5. For points with large curvature in the fused point cloud data, their corresponding standard deviations are judged. If the standard deviation of the angles is greater than the standard deviation threshold, the current point is considered a feature point.

[0039] Furthermore, in one embodiment, the fused point cloud data undergoes normal vector estimation, geometric feature recognition, denoising, and sampling optimization. Specifically, the denoising and sampling optimization include: S311: Based on the established spatial search structure octree, the minimum cube grid is set again according to the preset multiple resolution of the single frame point cloud data resolution. S312: For the smallest cube obtained by resetting, determine whether there are feature points in the current smallest cube. If not, use the center point of the current smallest cube to replace all points within the range of the current smallest cube. If they exist, do not process them. S313: Search for neighborhood points using the replaced points in the current smallest cube, use Gaussian filtering or moving least squares method to smooth the point cloud for denoising, recalculate the position of the points, and complete the denoising process and sampling optimization.

[0040] The following combination Figure 4 The noise reduction process and sampling optimization are explained in detail.

[0041] The fused point cloud data obtained above, which has already distinguished feature points from non-feature points, needs to be simplified and denoised next. In this application, a new processing method is adopted to process simplification and denoising in parallel, thereby improving processing efficiency.

[0042] In the steps described above, an octree spatial search structure has been established, allowing for rapid acquisition of neighborhood information. The first resolution mentioned above is the original point cloud resolution. To quickly simplify the point cloud data, the simplification value is reduced to a maximum of 1 / 4 of the original value. Therefore, a second resolution needs to be set, which is twice the spacing of the first resolution. See also... Figure 4 As shown, assuming that the fused point cloud data are all gray point clouds, after the above feature recognition, the blue points are the feature points. In the smallest cube obtained at the second resolution, if there are no feature points, the center of the smallest cube obtained at the second resolution needs to be used to replace the four points in the smallest cube divided at the first resolution. If there are feature points in the smallest cube obtained at the second resolution, the points in the cube are not changed.

[0043] The point cloud data has been simplified through the above partitioning. Simultaneously, the data is smoothed and denoised. Neighborhood points are searched using the replaced points, and Gaussian filtering or moving least squares methods are applied to smooth the point cloud and denoise it. The point positions are recalculated, resulting in the red points, which represent the final processed points. This process yields a simplified point cloud information that retains detailed features. Finally, after meshing, object reconstruction and subsequent measurements can be completed.

[0044] The post-processing method for area scan structured light point cloud data in this application integrates multiple steps such as denoising, registration, fusion, normal estimation, feature recognition, simplification and noise reduction by designing a complete post-processing workflow with interlocking and collaborative optimization. It achieves comprehensive improvement over existing methods in terms of processing accuracy, computational efficiency and feature preservation capability, and is especially suitable for industrial area scan structured light measurement scenarios with stringent requirements for efficiency and quality.

[0045] Secondly, embodiments of this application also provide a post-processing device for area-scanned structured light point cloud data.

[0046] In one embodiment, reference is made to Figure 5 , Figure 5 This is a schematic diagram of the functional modules of the post-processing device for surface-scanned structured light point cloud data in this application. Figure 5 As shown, the post-processing device for area scan structured light point cloud data includes: a conversion module, a fusion module, and a processing module.

[0047] The conversion module acquires multiple frames of point cloud data of the target object and performs outlier removal on each frame, transforming the outlier-removed multiple frames of point cloud data to the same coordinate system. The fusion module fuses the multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data. The processing module performs normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization on the fused point cloud data to complete the post-processing of the point cloud data.

[0048] The functions of each module in the post-processing device for structured light point cloud data correspond to the steps in the post-processing method embodiment for structured light point cloud data. Their functions and implementation processes will not be described in detail here.

[0049] Thirdly, embodiments of this application provide a post-processing device for area scan structured light point cloud data. The post-processing device for area scan structured light point cloud data can be a personal computer (PC), laptop computer, server, or other device with data processing capabilities.

[0050] Reference Figure 6 , Figure 6 This is a schematic diagram of the hardware structure of the post-processing device for area-scanned structured light point cloud data involved in the embodiments of this application. In the embodiments of this application, the post-processing device for area-scanned structured light point cloud data may include a processor, a memory, a communication interface, and a communication bus.

[0051] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0052] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal devices within the post-processing equipment for area-scanned structured light point cloud data, as well as interfaces used for interconnecting the post-processing equipment with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0053] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0054] The processor can be a general-purpose processor, which can call the post-processing program for the area-scanned structured light point cloud data stored in the memory and execute the post-processing method for the area-scanned structured light point cloud data provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the post-processing program for the area-scanned structured light point cloud data is called can be referred to the various embodiments of the post-processing method for the area-scanned structured light point cloud data in this application, and will not be repeated here.

[0055] Those skilled in the art will understand that Figure 6 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0056] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0057] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0058] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0059] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0060] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0061] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A post-processing method for area-scanned structured light point cloud data, characterized in that, The post-processing method for the area scan structured light point cloud data includes: Acquire multi-frame point cloud data of the target measurement object and perform outlier removal processing on each frame of point cloud data. Then, transform the multi-frame point cloud data after outlier removal to the same coordinate system. Multiple frames of point cloud data, transformed to the same coordinate system, are fused to obtain fused point cloud data. The fused point cloud data undergoes normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization to complete post-processing of the point cloud data.

2. The post-processing method for area-scanned structured light point cloud data as described in claim 1, characterized in that, The process of acquiring multi-frame point cloud data of the target object and removing outliers from each frame of point cloud data specifically includes: The target object is scanned by a surface scanning sensor to obtain multiple frames of point cloud data; Based on the plane fitting method, outlier removal is performed on each frame of point cloud data to identify and remove noise points.

3. The post-processing method for area-scanned structured light point cloud data as described in claim 2, characterized in that, The planar fitting method performs outlier removal processing on each frame of point cloud data. Specifically, the outlier removal processing for the current frame of point cloud data includes: A KDTree search structure is established for the current frame point cloud data. The current point in the current frame point cloud data is recorded as the target point. Plane fitting is performed on all points in the neighborhood of the target point to obtain the fitting plane. The average distance from all points in the neighborhood of the target point to the fitted plane, and the distance from the target point to the fitted plane are calculated. For each point in the target point's neighborhood, perform a nearest neighbor search and then sequentially determine the neighborhood of each point in the target point's neighborhood. If the neighborhood of the current point in the target point's neighborhood contains the target point, then determine that the current point and the target point are neighboring mutually beneficial points in the target point's neighborhood, and obtain the total number of neighboring mutually beneficial points corresponding to the target point. Based on the total number of neighboring mutually beneficial points corresponding to the target point, and the calculated distance information corresponding to the target point, the criterion value of the target point is calculated: in, The criterion value representing the target point, when If the value exceeds the outlier threshold, the target point is considered a noise point and is removed; otherwise, no action is taken. This represents the distance from the target point to the fitted plane. This represents the average distance from all points in the neighborhood of the target point to the fitted plane. This indicates the total number of neighboring reciprocal points. This represents the total number of points in the target area.

4. The post-processing method for area-scanned structured light point cloud data as described in claim 3, characterized in that, The outlier removal process for the current frame point cloud data also includes: The outlier removal process is repeated a preset number of times on the current frame point cloud data. Points in the neighborhood of the noise points detected during the current outlier removal process are marked. During the next outlier removal process, only the marked points are checked to see if they are noise points.

5. The post-processing method for area-scanned structured light point cloud data as described in claim 1, characterized in that, The process of converting the processed multi-frame point cloud data to the same coordinate system specifically includes: Based on the relative positional relationship, the attitude matrix of each point cloud data relative to the reference coordinate system is obtained; Based on the attitude matrix, the multi-frame point cloud data is transformed to the reference coordinate system.

6. The post-processing method for area-scanned structured light point cloud data as described in claim 1, characterized in that, The process of fusing and transforming multiple frames of point cloud data into the same coordinate system to obtain fused point cloud data specifically includes: Establish a spatial search structure octree, set the minimum cube grid according to the resolution of single frame point cloud data, and insert multiple frames of point cloud data into the spatial search structure octree. Check in turn whether there is a point in each of the smallest cubes. If there is no point in the current smallest cube, do nothing. If there is a point in the current smallest cube, average all the points in the current smallest cube and calculate the centroid position. Use the centroid to replace all the points in the current smallest cube as the unique point of the current smallest cube. Based on the unique point of all the smallest cubes, the point cloud data of multiple frames is fused to obtain the fused point cloud data.

7. The post-processing method for area-scanned structured light point cloud data as described in claim 6, characterized in that, The process of performing normal vector estimation, geometric feature recognition, noise reduction, and sampling optimization on the fused point cloud data includes, specifically, normal vector estimation and geometric feature recognition: Based on PCA estimation, the eigenvalue and eigenvector of each point in the fused point cloud data are calculated. The normal vector of the fused point cloud data is obtained based on the point corresponding to the minimum eigenvalue. Calculate the normal angles of the current point and its neighboring points in the fused point cloud data based on the normal vectors, and calculate the standard deviation of the angles based on the normal angles of all neighboring points of the current point in the fused point cloud data. The calculated angle standard deviation is compared with the angle standard deviation threshold. If it is greater than the threshold, the current point in the fused point cloud data is determined to be a feature point.

8. The post-processing method for area-scanned structured light point cloud data as described in claim 7, characterized in that, The process of estimating normal vectors, identifying geometric features, and performing noise reduction and sampling optimization on the fused point cloud data includes, specifically, the noise reduction and sampling optimization processes: Based on the established spatial search structure octree, the minimum cube grid is set again according to the preset multiple resolution of the single frame point cloud data. For the newly set minimum cube, determine whether there are feature points in the current minimum cube. If not, use the center point of the current minimum cube to replace all points within the current minimum cube. If they exist, do not process them. Search for neighborhood points using the replaced points in the current smallest cube, use Gaussian filtering or moving least squares method to smooth the point cloud for denoising, recalculate the position of the points, and complete the denoising process and sampling optimization.

9. A post-processing device for area-scanned structured light point cloud data, characterized in that, The post-processing device for the area scan structured light point cloud data includes: The conversion module is used to acquire multi-frame point cloud data of the target measurement object and perform outlier removal processing on each frame of point cloud data, and convert the multi-frame point cloud data after outlier removal to the same coordinate system. The fusion module is used to fuse multiple frames of point cloud data transformed to the same coordinate system to obtain fused point cloud data. The processing module is used to perform normal vector estimation, geometric feature recognition, noise reduction and sampling optimization on the fused point cloud data to complete the post-processing of the point cloud data.

10. A post-processing device for area-scanned structured light point cloud data, characterized in that, The post-processing device for area scan structured light point cloud data includes a processor, a memory, and a post-processing program for area scan structured light point cloud data stored in the memory and executable by the processor. When the post-processing program for area scan structured light point cloud data is executed by the processor, it implements the steps of the post-processing method for area scan structured light point cloud data as described in any one of claims 1 to 8.

Images