Plane segmentation and tree optimization model reconstruction method, device and equipment and medium

Abstract

The present application relates to a kind of plane segmentation and tree optimization model reconstruction method, device, equipment and medium, belong to three-dimensional modeling technical field, wherein, the method includes obtaining the point cloud data to be reconstructed;Supporting plane extraction is carried out to the point cloud data to be reconstructed, and the plane set containing multiple supporting planes is obtained;Tree-shaped space division structure is constructed based on the plane set, and the recursive segmentation of bounding box is carried out through the tree-shaped space division structure and tree simplification, and the convex cell set is obtained;Through the convex cell set, the three-dimensional surface grid model reconstructed by the three-dimensional point cloud data to be reconstructed is generated.Effectively improve the stability and reliability of three-dimensional reconstruction, realize the stable and efficient reconstruction in plane dominant scene, and then improve the precision and efficiency of three-dimensional model reconstruction.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, and in particular to a method, apparatus, equipment and medium for model reconstruction using planar segmentation and tree optimization. Background Technology

[0002] In fields such as digital city construction, digital preservation of cultural relics, and building information modeling, 3D model reconstruction is a crucial foundation of data. Lightweight, topologically complete 3D geometric meshes are the core carriers of 3D models, directly determining their engineering practicality and computational efficiency. Current 3D models primarily acquire point cloud data through 3D scanning; however, due to factors such as noise, occlusion, uneven density, and complex geometric structures, these models often exhibit problems like redundant faces, topological breaks, and irregular boundaries, severely restricting their engineering applications and widespread value.

[0003] Traditional 3D model reconstruction mainly relies on global fitting and surface reconstruction methods, which are inefficient and lack robustness, making it difficult to meet the high-precision and lightweight requirements of real-world scenarios. While mesh repair methods based on point cloud fitting (such as Poisson reconstruction and the moving cube algorithm) can achieve basic reconstruction, the repair effect is heavily dependent on the quality of the point cloud. For complex geometric structures (such as polyhedral edges and regular support surfaces) or noisy point clouds, errors such as redundant faces, topological breaks, and blurred boundaries often occur, making it impossible to generate regular and compact 3D models. Furthermore, traditional methods lack effective utilization of prior knowledge of the scene plane, and the spatial partitioning method is fixed and singular, making it difficult to adaptively simplify redundant structures. The final model is large in size and structurally chaotic, making it unsuitable for direct use in subsequent engineering applications.

[0004] In existing 3D reconstruction techniques based on planar segmentation and spatial partitioning, both traditional RANSAC-based planar extraction methods and fixed-mesh spatial partitioning models suffer from a fundamental limitation: a lack of adaptive understanding and optimization capabilities for scene geometry. Traditional methods achieve planar detection through point-by-point fitting and global thresholding, which is susceptible to noise interference and lacks robustness. Existing spatial partitioning models often employ fixed structures such as octrees and regular meshes, failing to achieve adaptive recursive segmentation based on the distribution of supporting planes, resulting in high cell redundancy. Both methods are data-driven and lack effective integration of planar priors and tree structures, leading to incomplete planar extraction, unreasonable spatial partitioning, and chaotic topological structures in the reconstruction results, making it difficult to balance model accuracy and lightweight requirements. Summary of the Invention

[0005] In view of this, it is necessary to provide a model reconstruction method, apparatus, equipment and medium based on planar segmentation and tree optimization to solve the problem that 3D model reconstruction cannot simultaneously meet the requirements of model accuracy and lightweight design.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a model reconstruction method based on planar segmentation and tree optimization, comprising: Acquire the point cloud data to be reconstructed; The point cloud data to be reconstructed is subjected to support plane extraction to obtain a set of planes containing multiple support planes; Based on the planar set, a tree-shaped spatial partitioning structure is constructed, and the bounding box is recursively segmented and simplified through the tree-shaped spatial partitioning structure to obtain a set of convex cells; The 3D surface mesh model to be reconstructed from the 3D point cloud data is generated using the set of convex cells.

[0007] In one possible implementation, before acquiring the point cloud data to be reconstructed, the method further includes: Acquire initial 3D point cloud data, and estimate the normal vector of the 3D point cloud data to be reconstructed using a hybrid domain search strategy to obtain the estimation result; The point cloud density of the three-dimensional point cloud data to be reconstructed is adjusted based on the estimation results to obtain the point cloud data to be reconstructed.

[0008] In one possible implementation, the step of extracting support planes from the point cloud data to be reconstructed to obtain a set of planes containing multiple support planes includes: An initial planar model corresponding to the point cloud data to be reconstructed is constructed using a preset planar detection algorithm; Multiple supporting planes are obtained by filtering and extracting from the initial plane model, and a plane set is constructed based on each supporting plane.

[0009] In one possible implementation, the step of filtering and extracting multiple supporting planes through the initial planar model, and constructing a plane set based on each supporting plane, includes: Multiple initial planes are obtained by extracting the initial plane model; Each initial plane is selected according to a preset selection strategy to obtain multiple supporting planes. The preset selection strategy sets selection parameters based on the point cloud noise level and gradually eliminates each initial plane based on the selection parameters until the proportion of the remaining point cloud in each initial plane is lower than a preset threshold. The selection parameters include a distance threshold, a minimum number of internal points, and the number of iterations. Construct a set of planes based on each supporting plane.

[0010] In one possible implementation, a tree-shaped spatial partitioning structure is constructed based on the planar set, and the bounding boxes are recursively segmented and simplified using the tree-shaped spatial partitioning structure to obtain a set of convex cells, including: A tree-shaped spatial partitioning structure is constructed based on the aforementioned planar set, and the initial bounding box is determined. The initial bounding box is used as the root node to generate multiple initial convex cells corresponding to multiple child nodes; The tree-shaped spatial partitioning structure is optimized and the initial convex cells are merged to obtain a set of convex cells.

[0011] In one possible implementation, optimizing the tree-like spatial partitioning structure and merging the initial convex cells to obtain a set of convex cells includes: The optimized tree structure is obtained by optimizing the tree space partitioning structure using a preset cost function. Based on the optimized tree structure, the groups of cells to be optimized in each initial cell are merged to obtain a convex cell set, wherein the groups of cells to be optimized are combinations of initial cells that satisfy at least one of the following: redundancy, volume less than a preset volume threshold, or similar geometric features.

[0012] In one possible implementation, generating the 3D surface mesh model reconstructed from the 3D point cloud data to be reconstructed using the convex cell set includes: Extract multiple convex hull vertices from the convex cell set; The point cloud data to be input is generated based on the convex hull vertices, and a three-dimensional surface mesh model corresponding to the three-dimensional point cloud data to be reconstructed is generated through the point cloud data to be input and the target energy function.

[0013] Secondly, the present invention also provides a model reconstruction device for planar segmentation and tree optimization, comprising: an acquisition module, an extraction module, a construction module, and a generation module; the acquisition module is used to acquire point cloud data to be reconstructed; the extraction module is used to extract supporting planes from the point cloud data to be reconstructed to obtain a set of planes containing multiple supporting planes; the construction module is used to construct a tree spatial partitioning structure based on the set of planes, and to perform recursive segmentation and tree simplification of bounding boxes through the tree spatial partitioning structure to obtain a set of convex cells; the generation module is used to generate a three-dimensional surface mesh model reconstructed from the three-dimensional point cloud data to be reconstructed through the set of convex cells.

[0014] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the planar segmentation and tree optimization model reconstruction method described in any of the above implementations.

[0015] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps in the model reconstruction method of planar segmentation and tree optimization described in any of the above implementations.

[0016] The beneficial effects of this invention are as follows: The model reconstruction method based on planar segmentation and tree optimization provided by this invention first acquires the point cloud data to be reconstructed and extracts supporting planes from the point cloud data, thereby obtaining a set of planes containing multiple supporting planes. Then, a tree spatial partitioning structure is introduced to adaptively partition the plane set. Through robust planar segmentation and adaptive spatial partitioning, lightweighting and topological regularization of the model are achieved. After obtaining the set of convex cells, a 3D surface mesh model reconstructed from the point cloud data is generated based on the set of convex cells. After the tree spatial partitioning step in the model reconstruction process, recursive segmentation and tree simplification are performed. While preserving the core structure of the model to the greatest extent, tree simplification reduces the storage and computational burden of redundant patches, effectively improving the stability and reliability of 3D reconstruction, achieving stable and efficient reconstruction in plane-dominated scenes, and thus improving the accuracy and efficiency of model reconstruction based on planar segmentation and tree optimization.

[0017] The technical effects achieved by the second and third aspects mentioned above are similar to those achieved by the corresponding technical means in the first aspect, and will not be repeated here. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the steps of a model reconstruction method for planar segmentation and tree optimization provided in an embodiment of this application; Figure 2 A flowchart illustrating another step of the planar segmentation and tree optimization model reconstruction method provided in an embodiment of this application; Figure 3 A schematic diagram of the overall technical route of the model reconstruction method for planar segmentation and tree optimization provided in an embodiment of this application; Figure 4 A functional block diagram of a planar segmentation and tree optimization model reconstruction device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] Before demonstrating the embodiments, the following terms will be explained.

[0024] Point cloud data refers to a set of vectors in a three-dimensional coordinate system. Scanned data is recorded in the form of points, each containing three-dimensional coordinates, or information such as color and reflectivity.

[0025] RANSAC Algorithm: RANSAC (RANdom Sampling Consensus) is a robust estimation method for estimating mathematical model parameters from a sample set containing outliers.

[0026] KDTree: short for kd-tree, is a tree-like data structure that stores instance points in k-dimensional space for fast retrieval. It is primarily used for searching key data in multi-dimensional space (e.g., range search and nearest neighbor search).

[0027] Tree-like spatial partitioning structure: recursively divides space into smaller sub-regions, forming a hierarchical tree structure, thereby accelerating spatial queries (such as range queries, nearest neighbor searches, etc.).

[0028] Convex cell: A convex cell is the smallest geometric unit in 3D space with a convex shape, commonly found in point clouds, mesh partitioning, and spatial meshing. A line connecting any two points within a cell lies entirely within the cell. The smallest basic block after spatial partitioning is a convex cell.

[0029] Bounding box: A simple, regular geometric structure that encloses a complex object (point cloud data, model, etc.).

[0030] BSP Tree Structure: BSP (Binary Space Partitioning) is a 3D graphics processing technique that recursively partitions space to form a binary tree structure. It uses the Cartesian plane equation to recursively divide 3D space and generate a node structure containing partition plane information, which is used for collision detection, view frustum culling, and scene management.

[0031] This invention provides a model reconstruction method, apparatus, electronic device, and storage medium for planar segmentation and tree optimization, which will be described below.

[0032] Figure 1 A schematic flowchart of an embodiment of the planar segmentation and tree optimization model reconstruction method provided by the present invention is shown below. Figure 1 As shown, the model reconstruction methods of planar segmentation and tree optimization include: S101. Obtain the point cloud data to be reconstructed.

[0033] It should be noted that the point cloud data to be reconstructed is the original initial 3D point cloud data that needs to be reconstructed into a 3D model, which has been processed by steps such as normal vector estimation and preprocessing.

[0034] In some embodiments, before the step of acquiring the point cloud data to be reconstructed, the method further includes: acquiring initial three-dimensional point cloud data, and estimating the normal vector of the three-dimensional point cloud data to be reconstructed using a hybrid neighborhood search strategy to obtain an estimation result; adjusting the point cloud density of the three-dimensional point cloud data to be reconstructed based on the estimation result to obtain the point cloud data to be reconstructed.

[0035] Understandably, the initial 3D point cloud data is first acquired, and the normal vector of the initial 3D point cloud data is estimated to obtain the estimation result.

[0036] Specifically, normal vector estimation employs a KDTree hybrid neighborhood search strategy. It fits the normal vector through the geometric relationships of points within a local neighborhood, while simultaneously performing normal vector consistency orientation processing to ensure a unified normal vector direction. The radius and number of nearest neighbors in the neighborhood search can be adaptively adjusted according to the point cloud density. Finally, the initial 3D point cloud data undergoes preprocessing to obtain the point cloud data to be reconstructed, which is then used for subsequent model reconstruction. This ensures a unified normal vector direction in the point cloud data, preventing plane recognition errors, subsequent segmentation errors, and model orientation errors caused by inconsistent normal vectors in the subsequent point cloud data.

[0037] S102. Extract the supporting planes from the point cloud data to be reconstructed to obtain a set of planes containing multiple supporting planes.

[0038] In the specific implementation, after obtaining the point cloud data to be reconstructed, robust plane detection and plane segmentation are introduced. First, multiple supporting planes are extracted, and then the supporting planes are combined into a plane set. The supporting plane can be the ground or a tabletop, or other planes that can be used as a reference to normalize the scene coordinates. This embodiment does not limit this.

[0039] In some embodiments, extracting supporting planes from the point cloud data to be reconstructed to obtain a plane set containing multiple supporting planes includes: constructing an initial plane model corresponding to the point cloud data to be reconstructed using a preset plane detection algorithm; filtering and extracting multiple supporting planes using the initial plane model; and constructing a plane set based on each supporting plane.

[0040] It should be noted that a preset plane detection algorithm is first introduced to construct an initial plane model. The preset plane prediction algorithm is the RANSAC algorithm, which can be used to construct the initial plane model.

[0041] Specifically, RANSAC plane detection constructs an initial plane model by randomly sampling three points. The distance from a point to the plane is calculated using the following formula:

[0042] in, a, b, c, d For planar model parameters, x, y, z For point cloud coordinates, d The perpendicular distance from the point to the plane.

[0043] Understandably, after constructing the initial planar model, all supporting planes in the point cloud data are selected using the constructed initial planar model, and the supporting planes are aggregated into a planar set. The planar model obtained in this way can filter out the optimal planar model with the most interior points, completely avoiding the interference of outliers on the fitting.

[0044] In some embodiments, multiple supporting planes are obtained by filtering and extracting the initial plane model, and a plane set is constructed based on each supporting plane. This includes: extracting multiple initial planes by the initial plane model; filtering each initial plane according to a preset filtering strategy to obtain multiple supporting planes, wherein the preset filtering strategy sets filtering parameters based on the point cloud noise level, and gradually eliminates each initial plane based on the filtering parameters until the proportion of the remaining point cloud in each initial plane is lower than a preset threshold, the filtering parameters including a distance threshold, a minimum number of interior points, and an iteration count; and constructing a plane set based on each supporting plane.

[0045] In the specific implementation, the process begins with plane extraction to obtain multiple initial planes, i.e., counting all planes. Then, supporting planes are selected. RANSAC plane detection employs a stepwise elimination strategy to filter supporting planes. The selection parameters can be dynamically adjusted based on the point cloud noise level, including but not limited to distance thresholds, minimum number of interior points, and number of iterations. Segmentation stops when the proportion of remaining point cloud points falls below a preset threshold, ultimately retaining the core supporting planes of the scene. This plane segmentation method then determines the supporting planes, thereby improving the accuracy and precision of subsequent 3D model reconstruction.

[0046] S103. Construct a tree-shaped spatial partitioning structure based on the planar set, and perform recursive segmentation and tree simplification of the bounding box through the tree-shaped spatial partitioning structure to obtain a set of convex cells.

[0047] It should be noted that after obtaining the planar set, a tree-shaped spatial partitioning structure is constructed to achieve adaptive tree partitioning, which can improve the reconstruction accuracy and efficiency in complex scenarios, and also improve the wide applicability of the reconstruction method.

[0048] S104. Generate a three-dimensional surface mesh model reconstructed from the three-dimensional point cloud data to be reconstructed using the convex cell set.

[0049] It should be understood that after obtaining the set of convex cells, the point cloud data in the set of convex cells is processed as input point cloud data, and then the model is reconstructed to obtain a three-dimensional surface mesh model. At the same time, the reconstructed three-dimensional surface mesh model is also topologically optimized, which can repair the holes in the model boundary and remove duplicate faces.

[0050] In some embodiments, generating a 3D surface mesh model reconstructed from the 3D point cloud data to be reconstructed using the set of convex cells includes: extracting multiple convex hull vertices from the set of convex cells; generating input point cloud data based on the convex hull vertices; and generating a 3D surface mesh model corresponding to the 3D point cloud data to be reconstructed using the input point cloud data and a target energy function.

[0051] It should be noted that convex hull fitting is performed on each convex cell and all faces are merged, and redundancy defects are eliminated through topology optimization.

[0052] Specifically, in this embodiment, topology optimization is performed through convex hull fitting. The convex hull fitting uses the ConvexHull algorithm to extract the convex hull vertices of the point set within the cell. The number of convex hull vertices can be dynamically adjusted according to the cell geometry, and this embodiment does not limit this.

[0053] It should be understood that if the number of extracted vertices is less than the preset value, it will be supplemented by filling with the centroids of the point set within the cell. The centroid calculation formula is as follows:

[0054] in, C For the point set's center of mass, n The number of points within the cell. x i， y i， z i For the cell's first i The coordinates of the points.

[0055] In this way, the number of convex hull vertices is guaranteed to be no less than the preset value, thereby ensuring the accuracy of the reconstructed 3D model.

[0056] It should be noted that topology optimization includes overlapping patch removal and boundary hole repair. Patch merging is achieved by minimizing energy, and the objective energy function expression is as follows:

[0057]

[0058]

[0059] in, E Let be the target energy function. E data For data fitting terms, E smooth For smoothing terms, λ These are adjustable weighting coefficients. E data In the calculation formula, p i The first in the original point cloud i One point, f j To reconstruct the first on the patch model j Each point / sampling point. In E smooth In the calculation formula, NTo reconstruct the set of all adjacent face pairs in the model, i and j Each represents a pair of face pieces. n j and n k These are the unit normal vectors of adjacent faces. The final output is a lightweight 3D surface mesh model with a complete topological structure and no redundancy defects.

[0060] Compared with existing technologies, the planar segmentation and tree-based optimization model reconstruction method provided in this embodiment first acquires the point cloud data to be reconstructed and extracts supporting planes from the point cloud data to obtain a set of planes containing multiple supporting planes. Then, a tree-based spatial partitioning structure is introduced to adaptively partition the plane set. Through robust planar segmentation and adaptive spatial partitioning, lightweighting and topological regularization of the model are achieved. After obtaining the set of convex cells, a 3D surface mesh model reconstructed from the point cloud data is generated based on the set of convex cells. After the tree-based spatial partitioning step in the model reconstruction process, recursive segmentation and tree simplification are performed. While preserving the core structure of the model to the greatest extent, tree simplification reduces the storage and computational burden of redundant patches, effectively improving the stability and reliability of 3D reconstruction. This enables stable and efficient reconstruction in plane-dominated scenes, thereby improving the accuracy and efficiency of 3D model reconstruction.

[0061] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the steps of an embodiment of the model reconstruction method for plane segmentation and tree optimization according to this application. The order of the steps in this flowchart can be changed, and some steps can be omitted, depending on different requirements. This model reconstruction method for plane segmentation and tree optimization can be applied to the aforementioned model reconstruction apparatus for plane segmentation and tree optimization, but is not limited thereto, and the embodiments of this application do not limit it in this regard.

[0062] This embodiment is a further improvement on the aforementioned embodiment. The main improvement lies in the following: In this embodiment, the steps of constructing a tree-shaped spatial partitioning structure based on the planar set, and recursively segmenting and simplifying the bounding box using the tree-shaped spatial partitioning structure to obtain a set of convex cells include: constructing a tree-shaped spatial partitioning structure to determine the initial bounding box, and using the initial bounding box as the root node to determine the initial convex cells as child nodes, and further optimizing and simplifying the tree structure to obtain a more concise set of convex cells. Through this method, it is possible to...

[0063] The specific process of this embodiment is as follows: Figure 2 As shown, it includes the following steps: S201. Construct a tree-shaped spatial partitioning structure based on the planar set and determine the initial bounding box.

[0064] It should be noted that a tree-shaped spatial partitioning structure is constructed based on the aforementioned planar set, which in turn yields an initial bounding box. Then, recursively segmenting and simplifying the tree based on the initial bounding box results in a Tree Simplified Partition convex cell set.

[0065] S202. Using the initial bounding box as the root node, generate multiple initial convex cells corresponding to multiple child nodes.

[0066] It should be understood that the tree space partitioning in this embodiment adopts a BSP tree structure, with the initial bounding box surrounding the point cloud as the root node, and child nodes generated by recursively bisecting the space through the supporting plane, with each node corresponding to a convex cell.

[0067] S203. Optimize the tree-shaped spatial partitioning structure and merge the initial convex cells to obtain a set of convex cells.

[0068] In practice, after obtaining the BSP tree structure, the tree structure needs to be optimized before subsequent operations are performed. This can reduce the model size and data volume, while preventing noisy data from affecting the reconstructed model.

[0069] In some embodiments, optimizing the tree spatial partitioning structure and merging each initial convex cell to obtain a set of convex cells includes: optimizing the tree spatial partitioning structure through a preset cost function to obtain an optimized tree structure; merging the groups of cells to be optimized in each initial cell based on the optimized tree structure to obtain a set of convex cells, wherein the groups of cells to be optimized are combinations of initial cells that satisfy at least one of the following: redundancy, volume less than a preset volume threshold, or similar geometric features.

[0070] It should be noted that, firstly, a preset cost function is introduced to optimize the parameters and the tree structure of the tree space partitioning structure, resulting in an optimized tree structure. Then, during the tree simplification process using the preset cost function, redundant, excessively small, or geometrically similar cells are merged.

[0071] Specifically, the preset cost function expression is as follows:

[0072] in, For cells c The number of boundary patches, complexity(c) For cells c The geometric complexity of interior point clouds, α , β These are adjustable weighting coefficients.

[0073] It is worth noting that, such as Figure 3The diagram shown illustrates the complete implementation route of the planar segmentation and tree optimization model reconstruction method in this embodiment. Figure 3 As shown, the original 3D point cloud data is first preprocessed using the original parameter configuration, and normal vector estimation is performed, specifically through KDTree neighborhood search and normal vector orientation. After preprocessing, the RANSAC algorithm is used for plane extraction, achieving robust planar model calculation and screening of supporting planes. Then, BSP tree space partitioning is used to recursively segment the space, followed by tree simplification using a pre-defined cost function. The simplified tree is then subjected to convex hull fitting and topology optimization, and an energy function is introduced for minimization and repair, resulting in the final reconstructed lightweight 3D mesh model. The 3D mesh model reconstructed in this way can achieve adaptive recursive segmentation based on the distribution of supporting planes. Through robust planar segmentation and adaptive space partitioning, the model achieves lightweighting and topology regularization, improving the accuracy, efficiency, and practicality of reconstruction in complex scenes.

[0074] Compared with the prior art, the model reconstruction method of planar segmentation and tree optimization provided in this embodiment first determines the initial BSP tree structure and the initial bounding box, then obtains multiple initial convex cells with the initial bounding box as the root node, and then optimizes the tree space partitioning structure to achieve the technical effect of preventing subsequent model redundancy and the occurrence of redundant structures.

[0075] Based on the same idea as the planar segmentation and tree optimization model reconstruction method in the above embodiments, this application also provides a planar segmentation and tree optimization model reconstruction apparatus, which can be used to perform the above-described planar segmentation and tree optimization model reconstruction method. For ease of explanation, the structural schematic diagram of the planar segmentation and tree optimization model reconstruction apparatus embodiment only shows the parts related to the embodiments of this application. Those skilled in the art will understand that the illustrated structure does not constitute a limitation on the apparatus, and it may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0076] like Figure 4 As shown, the model reconstruction apparatus 40 for planar segmentation and tree optimization includes an acquisition module 401, an extraction module 402, a construction module 403, and a generation module 404. In some embodiments, the above modules can be programmable software instructions stored in memory and executable by a processor. It is understood that in other embodiments, the above modules can also be program instructions or firmware embedded in the processor.

[0077] The acquisition module 401 is used to acquire the point cloud data to be reconstructed; The extraction module 402 is used to extract the supporting planes from the point cloud data to be reconstructed, and obtain a set of planes containing multiple supporting planes. The construction module 403 is used to construct a tree-shaped spatial partitioning structure based on the planar set, and to perform recursive segmentation and tree simplification of bounding boxes through the tree-shaped spatial partitioning structure to obtain a set of convex cells; The generation module 404 is used to generate a three-dimensional surface mesh model reconstructed from the three-dimensional point cloud data to be reconstructed through the convex cell set.

[0078] The planar segmentation and tree optimization model reconstruction device 40 provided in the above embodiments can realize the technical solutions described in the above planar segmentation and tree optimization model reconstruction method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above planar segmentation and tree optimization model reconstruction method embodiments, which will not be repeated here.

[0079] Please refer to Figure 5 , Figure 5 This is a schematic diagram of an embodiment of the electronic device of this application. In this embodiment of the invention, the electronic device 500 includes a processor 501, a memory 502, and a display 503. Figure 5 Only some components of the electronic device 500 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0080] In some embodiments, processor 501 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 502 or process data, such as the planar segmentation and tree optimization model reconstruction method of the present invention.

[0081] In some embodiments, processor 501 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 501 may be local or remote. In some embodiments, processor 501 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0082] In some embodiments, memory 502 may be an internal storage unit of electronic device 500, such as a hard disk or memory of electronic device 500. In other embodiments, memory 502 may also be an external storage device of electronic device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 500.

[0083] Furthermore, the memory 502 may include both internal storage units of the electronic device 500 and external storage devices. The memory 502 is used to store application software and various types of data installed on the electronic device 500.

[0084] In some embodiments, display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 503 is used to display information from electronic device 500 and to display visual user applications. Components 501-503 of electronic device 500 communicate with each other via a system bus.

[0085] In one embodiment, when processor 501 executes the planar segmentation and tree optimization model reconstruction program in memory 502, the following steps can be implemented: Acquire the point cloud data to be reconstructed; The point cloud data to be reconstructed is subjected to support plane extraction to obtain a set of planes containing multiple support planes; Based on the planar set, a tree-shaped spatial partitioning structure is constructed, and the bounding box is recursively segmented and simplified through the tree-shaped spatial partitioning structure to obtain a set of convex cells; The 3D surface mesh model to be reconstructed from the 3D point cloud data is generated using the set of convex cells.

[0086] It should be understood that when the processor 501 executes the model reconstruction program for plane segmentation and tree optimization in the memory 502, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0087] Furthermore, this embodiment of the invention does not specifically limit the type of electronic device 500 mentioned. Electronic device 500 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the invention, electronic device 500 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0088] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions of the planar segmentation and tree optimization model reconstruction methods provided in the above-described method embodiments.

[0089] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0090] The foregoing has provided a detailed description of the planar segmentation and tree optimization model reconstruction method, apparatus, electronic device, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A model reconstruction method of planar segmentation and tree optimization, characterized in that, include: Acquire the point cloud data to be reconstructed; The point cloud data to be reconstructed is subjected to support plane extraction to obtain a set of planes containing multiple support planes; Based on the planar set, a tree-shaped spatial partitioning structure is constructed, and the bounding box is recursively segmented and simplified through the tree-shaped spatial partitioning structure to obtain a set of convex cells; The 3D surface mesh model to be reconstructed from the 3D point cloud data is generated using the set of convex cells.

2. The planar partitioning and tree-optimizing model reconstruction method of claim 1, wherein, Before acquiring the point cloud data to be reconstructed, the process also includes: Acquire initial 3D point cloud data, and estimate the normal vector of the 3D point cloud data to be reconstructed using a hybrid domain search strategy to obtain the estimation result; The point cloud density of the three-dimensional point cloud data to be reconstructed is adjusted based on the estimation results to obtain the point cloud data to be reconstructed.

3. The model reconstruction method based on planar segmentation and tree optimization according to claim 1, characterized in that, The step of extracting support planes from the point cloud data to be reconstructed to obtain a set of planes containing multiple support planes includes: An initial planar model corresponding to the point cloud data to be reconstructed is constructed using a preset planar detection algorithm; Multiple supporting planes are obtained by filtering and extracting from the initial plane model, and a plane set is constructed based on each supporting plane.

4. The model reconstruction method of planar segmentation and tree optimization according to claim 3, characterized in that, The process of filtering and extracting multiple supporting planes through the initial planar model, and constructing a plane set based on each supporting plane, includes: Multiple initial planes are obtained by extracting the initial plane model; Each initial plane is selected according to a preset selection strategy to obtain multiple supporting planes. The preset selection strategy sets selection parameters based on the point cloud noise level and gradually eliminates each initial plane based on the selection parameters until the proportion of the remaining point cloud in each initial plane is lower than a preset threshold. The selection parameters include a distance threshold, a minimum number of internal points, and the number of iterations. Construct a set of planes based on each supporting plane.

5. The planar partitioning and tree-optimized model reconstruction method of claim 1, wherein, The process involves constructing a tree-like spatial partitioning structure based on the planar set, and then recursively segmenting and simplifying the bounding boxes using this tree-like spatial partitioning structure to obtain a set of convex cells, including: A tree-shaped spatial partitioning structure is constructed based on the aforementioned planar set, and the initial bounding box is determined. The initial bounding box is used as the root node to generate multiple initial convex cells corresponding to multiple child nodes; The tree-shaped spatial partitioning structure is optimized and the initial convex cells are merged to obtain a set of convex cells.

6. The planar partitioning and tree optimizing model reconstruction method of claim 5, wherein, The optimization of the tree-like spatial partitioning structure and the merging of each initial convex cell result in a set of convex cells, including: The optimized tree structure is obtained by optimizing the tree space partitioning structure using a preset cost function. Based on the optimized tree structure, the groups of cells to be optimized in each initial cell are merged to obtain a convex cell set, wherein the groups of cells to be optimized are combinations of initial cells that satisfy at least one of the following: redundancy, volume less than a preset volume threshold, or similar geometric features.

7. The planar partitioning and tree-optimized model reconstruction method of claim 1, wherein, The step of generating a 3D surface mesh model from the 3D point cloud data to be reconstructed using the convex cell set includes: Extract multiple convex hull vertices from the convex cell set; The point cloud data to be input is generated based on the convex hull vertices, and a three-dimensional surface mesh model corresponding to the three-dimensional point cloud data to be reconstructed is generated through the point cloud data to be input and the target energy function.

8. A model reconstruction device for planar segmentation and tree optimization, characterized in that, include: Acquire modules, extract modules, build modules, and generate modules; The acquisition module is used to acquire the point cloud data to be reconstructed; The extraction module is used to extract the supporting planes from the point cloud data to be reconstructed, and obtain a set of planes containing multiple supporting planes; The construction module is used to construct a tree-shaped spatial partitioning structure based on the planar set, and to perform recursive segmentation and tree simplification of bounding boxes through the tree-shaped spatial partitioning structure to obtain a set of convex cells; The generation module is used to generate a three-dimensional surface mesh model reconstructed from the three-dimensional point cloud data to be reconstructed using the convex cell set. 9.An electronic device comprising a processor and a memory, wherein, The memory is used to store instructions, and the processor is used to call the instructions in the memory to cause the electronic device to execute the planar segmentation and tree optimization model reconstruction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed on an electronic device, cause the electronic device to perform the planar segmentation and tree optimization model reconstruction method as described in any one of claims 1 to 7.

Images