A method, system, device and readable storage medium for generating three-dimensional reconstruction model

Abstract

The present application discloses a method, system, device and readable storage medium for generating a three-dimensional reconstruction model, the method comprising: calculating the current signed distance function value from each spatial unit to the surface of the first three-dimensional reconstruction model; decomposing and weightedly fusion each current signed distance function value with the initial signed distance function value of the spatial unit in different directions, and updating the initial signed distance function value to the obtained fusion result; determining whether the current data frame is the last frame; if so, establishing a second three-dimensional reconstruction model according to each initial signed distance function value. The present application decomposes and weightedly fuses the current signed distance function value with the initial signed distance function value, realizes a more accurate recording and expression of the reconstruction process data, solves the problem of the accuracy of the reconstruction process data recording, solves the problem of the smoothing of the reconstruction result details caused by the data fusion process, better retains the reconstructed geometric details, and improves the accuracy of the three-dimensional reconstruction model.

Description

Technical Field

[0001] The present application relates to the field of three-dimensional reconstruction models, and in particular to a method, system, device and readable storage medium for generating a three-dimensional reconstruction model. Background Art

[0002] 3D reconstruction is a technology that uses data collected by various sensors to build a digital model of a real scene. It involves the intersection of multiple disciplines such as sensing, computer vision, and computer graphics, and has a wide range of applications in many fields such as robotics, autonomous driving, VR / AR, etc. Different sensors collect data in different formats and have different corresponding reconstruction methods, and their reconstructed models have different representation formats and application scenarios. Generally speaking, being able to reconstruct a dense model of a scene in real time and stably is the goal that people pursue for 3D reconstruction technology.

[0003] Existing 3D reconstruction data representation formats include point clouds, voxels, and implicit representations. Implicit representation defines data in 3D space in the form of a function, and defines the shape to be expressed based on the isosurface of the function. Signed distance function (SDF) is a common implicit representation method for 3D data. It represents 3D data by defining the distance from a certain position in the spatial field to the nearest object surface. However, this representation method will cause the geometric details of the reconstructed object to be lost, resulting in a relatively low accuracy of the reconstructed model.

[0004] Therefore, how to improve the accuracy of the three-dimensional reconstruction model is a technical problem that technicians in this field currently need to solve. Summary of the invention

[0005] The purpose of this application is to provide a method, system, device and readable storage medium for generating a three-dimensional reconstruction model, so as to improve the accuracy of the three-dimensional reconstruction model.

[0006] In order to solve the above technical problems, the present application provides a method for generating a three-dimensional reconstruction model, the method comprising:

[0007] Reconstructing the input data frame to obtain a first three-dimensional reconstructed model;

[0008] Acquire a spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, and calculate a current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model;

[0009] Decomposing each of the current signed distance function values ​​and the initial signed distance function value of the spatial unit in different directions and weightedly fusing them in different directions, and updating the initial signed distance function value to the obtained fusion result;

[0010] Determine whether the current data frame is the last frame;

[0011] If not, returning to the step of reconstructing the input data frame to obtain the first three-dimensional reconstructed model;

[0012] If yes, a second three-dimensional reconstruction model is established according to each of the initial signed distance function values.

[0013] Optionally, the decomposing and weighting each of the current signed distance function values ​​and the initial signed distance function value of the spatial unit in different directions comprises:

[0014] Selecting at least two preset directions in a preset coordinate system, and decomposing the current signed distance function value into the preset directions to obtain corresponding current function value components;

[0015] Decomposing the initial signed distance function value into the preset direction to obtain corresponding initial function value components;

[0016] The current function value component and the initial function value component in the same preset direction are weightedly fused to obtain the fusion result.

[0017] Optionally, establishing a second three-dimensional reconstruction model according to each of the initial signed distance function values ​​includes:

[0018] According to a preset synthesis strategy, distance information from the spatial unit to the second three-dimensional reconstruction model is calculated according to the initial signed distance function value, and reconstruction of the second three-dimensional reconstruction model is completed according to the distance information.

[0019] Optionally, the calculating the current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model includes:

[0020] A point on the surface of the first three-dimensional reconstructed model that is closest to the spatial unit is determined as a reconstruction point, and the current signed distance function value is calculated using the coordinates of the reconstruction point and the spatial unit.

[0021] Optionally, the calculating the current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model includes:

[0022] When the data frame is a depth map, transforming the spatial unit into the depth map to obtain corresponding depth information;

[0023] The depth information is used as a current signed distance function value from the spatial unit to the surface of the first three-dimensional reconstructed model.

[0024] Optionally, the data frame is specifically image data;

[0025] The step of reconstructing the input data frame to obtain a first three-dimensional reconstruction model includes:

[0026] The visual three-dimensional reconstruction component is called to reconstruct the image data to obtain the first three-dimensional reconstruction model.

[0027] Optionally, before obtaining the spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, the method further includes:

[0028] The space is dynamically divided according to the position of the data frame to obtain each of the space units.

[0029] The present application also provides a system for generating a three-dimensional reconstruction model, the system comprising:

[0030] A first reconstruction module, used to reconstruct the input data frame to obtain a first three-dimensional reconstruction model;

[0031] an acquisition and calculation module, used for acquiring a spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, and calculating a current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model;

[0032] A fusion and update module, used for performing directional decomposition and directional weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit, and updating the initial signed distance function value to the obtained fusion result;

[0033] A judgment module, used to judge whether the current data frame is the last frame;

[0034] A return module, configured to return to the first reconstruction module to execute the step of reconstructing the input data frame to obtain a first three-dimensional reconstructed model if the current data frame is not the last frame;

[0035] The second reconstruction module is used to establish a second three-dimensional reconstruction model according to each of the initial signed distance function values ​​if the current data frame is the last frame.

[0036] The present application also provides a three-dimensional reconstruction model generation device, the three-dimensional reconstruction model generation device comprising:

[0037] Memory for storing computer programs;

[0038] A processor is used to implement the steps of the method for generating a three-dimensional reconstruction model as described in any one of the above items when executing the computer program.

[0039] The present application also provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating a three-dimensional reconstruction model as described in any one of the above items are implemented.

[0040] The method for generating a three-dimensional reconstruction model provided in the present application includes: reconstructing an input data frame to obtain a first three-dimensional reconstruction model; obtaining a spatial unit within a preset range corresponding to each point in the first three-dimensional reconstruction model, and calculating a current signed distance function value from each spatial unit to the surface of the first three-dimensional reconstruction model; performing directional decomposition and directional weighted fusion on each current signed distance function value and an initial signed distance function value of the spatial unit, and updating the initial signed distance function value to the obtained fusion result; determining whether the current data frame is the last frame; if not, returning to execute the step of reconstructing the input data frame to obtain the first three-dimensional reconstruction model; if so, establishing a second three-dimensional reconstruction model according to each initial signed distance function value.

[0041] The technical solution provided by the present application achieves more accurate recording and expression of reconstruction process data by decomposing each current signed distance function value and the initial signed distance function value of the spatial unit in different directions, and decomposing the current signed distance function value of a single direction into components under multiple fixed directions, thereby solving the problem of accuracy of reconstruction process data recording; at the same time, each current signed distance function value and the initial signed distance function value of the spatial unit are weightedly fused in different directions, thereby solving the problem of smoothing of reconstruction result details caused by the data fusion process, better preserving the reconstructed geometric details, and greatly improving the accuracy of the three-dimensional reconstruction model. The present application also provides a system, device and readable storage medium for generating a three-dimensional reconstruction model, which have the above-mentioned beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0043] Figure 1 A flowchart of a method for generating a three-dimensional reconstruction model provided in an embodiment of the present application;

[0044] Figure 2 An example diagram of a distance representation in different directions provided in an embodiment of the present application;

[0045] Figure 3 A structural diagram of a system for generating a three-dimensional reconstruction model provided in an embodiment of the present application;

[0046] Figure 4 A structural diagram of a three-dimensional reconstruction model generation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] The core of this application is to provide a method, system, device and readable storage medium for generating a three-dimensional reconstruction model, so as to improve the accuracy of the three-dimensional reconstruction model.

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] 3D reconstruction is a technology that uses data collected by various sensors to build a digital model of a real scene. It involves the intersection of multiple disciplines such as sensing, computer vision, and computer graphics, and has a wide range of applications in many fields such as robotics, autonomous driving, VR / AR, etc. Different sensors collect data in different formats and have different corresponding reconstruction methods, and their reconstructed models have different representation formats and application scenarios. Generally speaking, being able to reconstruct a dense model of a scene in real time and stably is the goal that people pursue for 3D reconstruction technology.

[0050] Common three-dimensional reconstruction data representation formats include point cloud, voxel and implicit representation. Point cloud is an irregular data representation method. There is no topological relationship between the points. It is generally used to represent sparse field data in space and is difficult to represent complex geometric shapes or large-scale dense data. Voxel is the division of space into grid units. Each grid unit can take different states according to its position relationship with the object in the reconstructed scene, thereby realizing the expression of the scene. The expressive power of voxel representation depends on the resolution of space division. The finer the spatial grid division, the stronger the shape expression ability and the larger the storage space required. Therefore, it is not suitable for a more detailed representation of large-scale scenes. Implicit representation defines the data in three-dimensional space in the form of a function, and defines the shape to be expressed according to the isosurface of the function. Implicit representation can theoretically achieve data expression capabilities with infinite resolution and can generate continuous surface models. Signed distance function is a common implicit representation method for three-dimensional data. It represents three-dimensional data by defining the distance from a certain position in the spatial field to the nearest object surface. Three-dimensional reconstruction is generally performed based on serialized data collected by sensors, and different frame data need to be fused to obtain a complete reconstruction result. The signed distance function usually stores a signed distance and a weight value at each data position. The reconstruction result is obtained by weighted fusion of the signed distances obtained by reconstruction of different frames. This direct weighted fusion of distances will cause the loss of geometric details of the reconstructed object itself and cause the reconstructed surface to be too smooth. Therefore, this application provides a method for generating a three-dimensional reconstruction model to solve the above problem.

[0051] Please refer to Figure 1 , Figure 1 A flowchart of a method for generating a three-dimensional reconstruction model provided in an embodiment of the present application.

[0052] It specifically includes the following steps:

[0053] S101: Reconstruct the input data frame to obtain a first three-dimensional reconstruction model;

[0054] The first 3D reconstruction model mentioned here can specifically be a 3D point cloud. This step can perform different operations for different input data types. For example, when the input is a point cloud, this step can perform no operation.

[0055] In a specific embodiment, the data frame may be image data. On this basis, the input data frame is reconstructed to obtain the first three-dimensional reconstructed model, which may be:

[0056] The visual three-dimensional reconstruction component is called to reconstruct the image data to obtain a first three-dimensional reconstruction model.

[0057] S102: Acquire a spatial unit within a preset range corresponding to each point in the first 3D reconstructed model, and calculate a current signed distance function value from each spatial unit to the surface of the first 3D reconstructed model;

[0058] After obtaining the first 3D reconstructed model, the spatial unit within a certain distance range before and after the observation direction of each point in the space is calculated, wherein the determination of the preset range may be related to the error of the sensor acquisition data, or may be determined according to factors such as the size of the spatial unit. For example, in a specific embodiment, before obtaining the spatial unit within the preset range corresponding to each point in the first 3D reconstructed model, the following steps may be performed to complete the determination of the preset range:

[0059] The acquisition error of the data frame is obtained, and the size of the preset range is determined according to the acquisition error.

[0060] In a specific embodiment, before obtaining the spatial unit within the preset range corresponding to each point in the first 3D reconstructed model, the following steps may be performed to complete the determination of the preset range:

[0061] The size of the spatial unit is obtained, and the size of the preset range is determined according to the size of the spatial unit.

[0062] The spatial unit mentioned here may be obtained by dividing the space in advance, or may be dynamically divided according to the position of each frame input point cloud. For example, in a specific embodiment, before obtaining the spatial unit within a preset range corresponding to each point in the first 3D reconstructed model, the following steps may be performed to complete the division of each spatial unit:

[0063] The space is dynamically divided according to the position of the data frame to obtain each space unit.

[0064] In a specific embodiment, the calculation of the current signed distance function value from each spatial unit to the surface of the first three-dimensional reconstructed model mentioned here may specifically be:

[0065] The point on the surface of the first three-dimensional reconstructed model that is closest to the spatial unit is determined as a reconstruction point, and the coordinates of the reconstruction point and the spatial unit are used to calculate the current signed distance function value.

[0066] When the data frame is a depth map, the calculation of the current signed distance function value from each spatial unit to the surface of the first 3D reconstructed model mentioned here may also be specifically:

[0067] When the data frame is a depth map, the spatial unit is transformed into the depth map to obtain the corresponding depth information;

[0068] The depth information is taken as the current signed distance function value from the spatial unit to the surface of the first three-dimensional reconstructed model.

[0069] The current signed distance function value has a sign, and when the spatial unit is located on the side of the reconstruction point close to the observation position, the sign is positive; when the spatial unit is located on the side of the reconstruction point far from the observation position, the sign is negative, that is, the above embodiment includes:

[0070] (1) Calculate the current signed distance function value from a certain spatial unit to the reconstructed object surface. The position of the spatial unit can be represented by the position of its center or a vertex, or determined by other rules. The reconstructed object surface can be represented by the coordinates of the reconstruction point. In this case, the current signed distance function value can be calculated using the coordinates of the reconstruction point and the spatial unit. When the input data has a depth map, the spatial unit can also be transformed into the depth map of the current frame to find its corresponding depth information, and use the depth as the depth position of the reconstructed object surface.

[0071] (2) The calculation of the current signed distance function value can be the straight-line distance between two points, or the distance obtained by projecting the distance to the direction of the camera optical axis. In this case, a certain filtering effect can be achieved on the errors existing in the input data itself.

[0072] (3) The current signed distance function value calculated above is transformed into a unified world coordinate system. The transformation process requires the camera's pose data, which is generally solved during the reconstruction process or input with the input data. The unified world coordinate system can select the local coordinate system corresponding to the first frame data, or be determined by other methods, which is not specifically limited in this application.

[0073] S103: performing direction-wise decomposition and direction-wise weighted fusion on each current signed distance function value and the initial signed distance function value of the spatial unit, and updating the initial signed distance function value to the obtained fusion result;

[0074] In this step, the purpose of directional decomposition and directional weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit is to decompose the current signed distance function value of a single direction into distance components in multiple fixed directions, so as to achieve more accurate recording and expression of the reconstruction process data, thereby solving the problem of accuracy of data recording in the three-dimensional reconstruction process; at the same time, directional weighted fusion is used to solve the problem of smoothing of reconstruction result details caused by the data fusion process, better retain the reconstructed geometric details, and improve the accuracy of the three-dimensional reconstruction model.

[0075] In a specific embodiment, the above-mentioned decomposition and weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit in different directions may be:

[0076] Selecting at least two preset directions in a preset coordinate system, and decomposing the current signed distance function value into the preset directions to obtain corresponding current function value components;

[0077] Decomposing the initial signed distance function value into preset directions to obtain corresponding initial function value components;

[0078] The current function value component and the initial function value component in the same preset direction are weightedly fused to obtain a fusion result.

[0079] The preset directions mentioned here can be the three coordinate axis directions of the world coordinate system, or can be specified in other ways, and once determined, they remain fixed during the entire reconstruction process. Here, the three coordinate axis directions X, Y, and Z of the world coordinate system are used as an example to illustrate. The X, Y, and Z directions in the world coordinate system are selected, and the current signed distance function values ​​of each spatial unit calculated are decomposed into these directions, which can specifically include:

[0080] (1) After the direction decomposition of the current signed distance function value, it can be represented by different variables, or it can be represented by one variable. When it is represented by one variable, different positions of the variable are used to represent different direction values. Here, one variable is used as an example for explanation.

[0081] Please refer to Figure 2 , Figure 2This is an example diagram of a directional distance representation provided in an embodiment of the present application. In this embodiment, a 32-bit variable is used to represent the distance values ​​in the X, Y, and Z directions, where the first two digits represent the sign of the distance, the first ten digits represent the distance in the X direction, the middle ten digits represent the distance in the Y direction, and the last ten digits represent the distance in the Z direction.

[0082] It should be noted that the number of digits of the variable and the representation method of each direction can also be determined according to other rules. Figure 2 In the example, X, Y, and Z do not store symbol information separately, and each direction shares the same symbol information, or each direction may store separate symbol information.

[0083] (2) In addition to saving the distance information between itself and the reconstructed surface in the above manner, each spatial unit also needs to save the corresponding weight information for the subsequent fusion process.

[0084] (3) For each spatial unit, the reconstructed data of the new frame needs to be fused into the reconstructed overall result. The fusion process is performed in different directions according to the above-mentioned directional distance representation method.

[0085] Here we take the X direction as an example. When fusion is performed, the distance sdf_x in the X direction must be read out first. Figure 2 When the representation method shown represents the distance sdf, the bit operation of reading sdf_x from sdf is sdf_x=(sdf&1072693248)>>20, the bit operation floating point type is 32 bits, and it is split into three 10 bits. The X-direction distance calculated by the current data frame is recorded as sdf_x', and the two are weighted fusion to obtain the fused X-direction distance sdf_x"=sdf_x'+w*sdf_x. Among them, w is the weight corresponding to sdf_x, and the weight corresponding to sdf_x' is 1. The update process needs to update the weight at the same time, which can be updated in the way of w'=w+1. The above weighted fusion and weight update can also be performed using other strategies.

[0086] Write the updated X-direction distance into sdf, and the corresponding bit operation is sdf'=(sdf&1048575)|(sdf_x"<<20). Save the updated distance sdf' and weight w' as the new sdf and w in the spatial unit.

[0087] The fusion in other directions is similar to the above process. The bit operation of reading sdf_y from sdf is sdf_y=(sdf&1047552)>>10, and the bit operation of writing sdf_y into sdf is sdf'=(sdf&1072694271)|(sdf_y<<10). The bit operation of reading sdf_z from sdf is sdf_z=sdf&1023, and the bit operation of writing sdf_z into sdf is sdf'=(sdf&1073740800)|sdf_z.

[0088] S104: Determine whether the current data frame is the last frame;

[0089] If not, return to step S101;

[0090] If yes, proceed to step S105;

[0091] When the current data frame is not the last frame, it indicates that there are still data frames to be input, and the process returns to step S101 to continue updating the signed distance function value of the spatial unit according to the input data frame;

[0092] When the current data frame is the last frame, it indicates that no data frame needs to be input at this time, and step S105 is executed at this time to establish a second 3D reconstruction model according to each initial signed distance function value, thereby obtaining an optimal 3D reconstruction model.

[0093] S105: Establishing a second three-dimensional reconstruction model according to each initial signed distance function value.

[0094] When establishing the second 3D reconstruction model based on each initial signed distance function value, the distance information from a certain space unit to the reconstructed object surface is needed. This distance information can be obtained by synthesizing the distances in various directions, or by adopting a certain strategy to select the distance in a certain direction as the distance of the space unit, or by transforming the distances in various directions according to other strategies as the distance from the space unit to the reconstructed object surface. When the distance needs to be used for calculations and other operations, use Figure 2 The calculation of the whole number in the example representation is equivalent to performing the same processing operation on the distances in the X, Y, and Z directions respectively, thereby achieving efficient representation and processing.

[0095] That is, in a specific embodiment, the second three-dimensional reconstruction model is established according to each initial signed distance function value, which can be specifically:

[0096] According to a preset synthesis strategy, distance information from the space unit to the second three-dimensional reconstruction model is calculated according to the initial signed distance function value, and reconstruction of the second three-dimensional reconstruction model is completed according to the distance information.

[0097] Based on the above technical solution, a method for generating a three-dimensional reconstruction model provided by the present application achieves more accurate recording and expression of reconstruction process data by decomposing each current signed distance function value and the initial signed distance function value of the spatial unit in different directions, and decomposing the current signed distance function value of a single direction into components in multiple fixed directions, thereby solving the problem of accuracy of reconstruction process data recording; at the same time, each current signed distance function value and the initial signed distance function value of the spatial unit are weightedly fused in different directions, thereby solving the problem of smoothing of reconstruction result details caused by the data fusion process, better retaining the reconstructed geometric details, and greatly improving the accuracy of the three-dimensional reconstruction model.

[0098] Please refer to Figure 3 , Figure 3 A structural diagram of a system for generating a three-dimensional reconstruction model provided in an embodiment of the present application.

[0099] The system may include:

[0100] A first reconstruction module 100 is used to reconstruct the input data frame to obtain a first three-dimensional reconstruction model;

[0101] The acquisition and calculation module 200 is used to acquire the spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, and calculate the current signed distance function value from each spatial unit to the surface of the first three-dimensional reconstructed model;

[0102] A fusion and update module 300 is used to perform directional decomposition and directional weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit, and update the initial signed distance function value to the obtained fusion result;

[0103] The judging module 400 is used to judge whether the current data frame is the last frame;

[0104] The return module 500 is used to return to the first reconstruction module to perform the step of reconstructing the input data frame to obtain a first three-dimensional reconstruction model if the current data frame is not the last frame;

[0105] The second reconstruction module 600 is used to establish a second 3D reconstruction model according to each initial signed distance function value if the current data frame is the last frame.

[0106] Since the embodiments of the system part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the system part, which will not be repeated here.

[0107] Please refer to Figure 4 , Figure 4 A structural diagram of a three-dimensional reconstruction model generation device provided in an embodiment of the present application.

[0108] The three-dimensional reconstruction model generation device 700 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 722 (for example, one or more processors) and a memory 732, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 742 or data 744. Among them, the memory 732 and the storage medium 730 can be temporary storage or permanent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the device. Furthermore, the processor 722 can be configured to communicate with the storage medium 730 to execute a series of instruction operations in the storage medium 730 on the three-dimensional reconstruction model generation device 700.

[0109] The three-dimensional reconstruction model generating device 700 may also include one or more power supplies 727, one or more wired or wireless network interfaces 750, one or more input and output interfaces 758, and / or one or more operating systems 741, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0110] Above Figure 1 The steps in the method for generating a three-dimensional reconstruction model are performed by a three-dimensional reconstruction model generating device based on the Figure 4 The structure shown is implemented.

[0111] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0112] In the several embodiments provided in the present application, it should be understood that the disclosed devices, equipment and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0113] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0114] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0115] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a function calling device, or a network device, etc.) to execute all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program code.

[0116] The above is a detailed introduction to the method, system, device and readable storage medium for generating a three-dimensional reconstruction model provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

[0117] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

Claims

1. A method for generating a three-dimensional reconstruction model, characterized in that: include: Reconstructing the input data frame to obtain a first three-dimensional reconstruction model; wherein the data frame is image data; Acquire a spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, and calculate a current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model; Decomposing each of the current signed distance function values ​​and the initial signed distance function value of the spatial unit in different directions and weightedly fusing them in different directions, and updating the initial signed distance function value to the obtained fusion result; Determine whether the current data frame is the last frame; If not, returning to the step of reconstructing the input data frame to obtain the first three-dimensional reconstructed model; If yes, establishing a second three-dimensional reconstruction model according to each of the initial signed distance function values; The step of performing directional decomposition and directional weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit comprises: Select at least two preset directions under a preset coordinate system, and decompose the current signed distance function value into the preset directions to obtain corresponding current function value components; decompose the initial signed distance function value into the preset directions to obtain corresponding initial function value components; perform weighted fusion of the current function value components and the initial function value components in the same preset direction to obtain the fusion result.

2. The method according to claim 1, characterized in that: Establishing a second three-dimensional reconstruction model according to each of the initial signed distance function values ​​includes: According to a preset synthesis strategy, distance information from the spatial unit to the second three-dimensional reconstructed model is calculated according to the initial signed distance function value, and reconstruction of the second three-dimensional reconstructed model is completed according to the distance information.

3. The method according to claim 1, characterized in that: The calculating of the current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model comprises: A point on the surface of the first three-dimensional reconstructed model that is closest to the spatial unit is determined as a reconstruction point, and the current signed distance function value is calculated using the coordinates of the reconstruction point and the spatial unit.

4. The method according to claim 1, characterized in that: The calculating of the current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model comprises: When the data frame is a depth map, transforming the spatial unit into the depth map to obtain corresponding depth information; The depth information is used as a current signed distance function value from the spatial unit to the surface of the first three-dimensional reconstructed model.

5. The method according to claim 1, characterized in that The step of reconstructing the input data frame to obtain a first three-dimensional reconstruction model includes: The visual three-dimensional reconstruction component is called to reconstruct the image data to obtain the first three-dimensional reconstruction model.

6. The method according to claim 1, characterized in that Before obtaining the spatial unit within a preset range corresponding to each point in the first three-dimensional reconstruction model, the method further includes: The space is dynamically divided according to the position of the data frame to obtain each of the space units.

7. A system for generating a three-dimensional reconstruction model, characterized in that: include: A first reconstruction module, used to reconstruct an input data frame to obtain a first three-dimensional reconstruction model; wherein the data frame is image data; an acquisition and calculation module, used for acquiring a spatial unit within a preset range corresponding to each point in the first three-dimensional reconstructed model, and calculating a current signed distance function value from each of the spatial units to the surface of the first three-dimensional reconstructed model; A fusion and update module, used for performing directional decomposition and directional weighted fusion of each current signed distance function value and the initial signed distance function value of the spatial unit, and updating the initial signed distance function value to the obtained fusion result; A judgment module, used to judge whether the current data frame is the last frame; A return module, configured to return to the first reconstruction module to execute the step of reconstructing the input data frame to obtain a first three-dimensional reconstructed model if the current data frame is not the last frame; A second reconstruction module, configured to establish a second three-dimensional reconstruction model according to each of the initial signed distance function values ​​if the current data frame is the last frame; The fusion and update module is specifically used for: Select at least two preset directions under a preset coordinate system, and decompose the current signed distance function value into the preset directions to obtain corresponding current function value components; decompose the initial signed distance function value into the preset directions to obtain corresponding initial function value components; perform weighted fusion of the current function value components and the initial function value components in the same preset direction to obtain the fusion result.

8. A three-dimensional reconstruction model generation device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the method for generating a three-dimensional reconstructed model as claimed in any one of claims 1 to 6 when executing the computer program.

9. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for generating a three-dimensional reconstruction model according to any one of claims 1 to 6 are implemented.

Images