Random geometric modeling three-dimensional data set generation method and device, medium and product

Abstract

The invention discloses a random geometric modeling three-dimensional data set generation method and device, a medium and a product, and relates to the technical field of data set generation, and the method comprises the steps: for any three-dimensional geometry under the current cycle index, randomly generating depth, space parameters and size within a set value range, generating an initial three-dimensional geometry in the empty three-dimensional scene; randomly generating a noise texture and a target energy value; applying the noise textures to the corresponding initial three-dimensional geometries to obtain an initial three-dimensional scene; in the initial three-dimensional scene, setting a fluorescent lamp light source with a target energy value to obtain a target three-dimensional scene; and rendering the target three-dimensional scene to obtain an image data set, adding the image data set into the initial data set to obtain a target data set, and performing the next cycle until the maximum cycle index is reached to obtain a random geometric modeling three-dimensional data set. According to the invention, rapid generation of the stochastic geometric modeling three-dimensional data set is realized.

Description

Technical Field

[0001] The present application relates to the technical field of dataset generation, and particularly to a method, device, medium and product for generating a three-dimensional dataset by random geometric modeling. Background Art

[0002] In recent years, the research and application in the field of images have received extensive attention from people. For example, through the research of image recognition, common functions such as face-swiping payment, clocking in to work, and commodity classification are realized; semantic segmentation is used to improve the diagnostic efficiency of medical diseases; pedestrians and vehicles are accurately segmented to promote the development of driverless; through the research of computational holography, the effect of naked-eye true 3D stereoscopic display is realized, etc.

[0003] With the progress of hardware technology and the significant improvement of the computing power of computing platforms, researchers are increasingly inclined to use neural networks to solve the above common problems in the field of images. A large amount of three-dimensional datasets are required as inputs when training neural network models. Common image training sets include NYU, MIT-CGH-4K, etc. A suitable dataset is extremely important for the training of network models. In order to approximate the limit performance of the network model and achieve a better training effect, the training set used for model training should be as rich and comprehensive as possible. However, the data in the dataset is limited and cannot provide all possible training data, which results in the fact that the effect obtained by the trained network model on the test set is often not as good as that on the training set.

[0004] Therefore, a method for generating an artificial synthetic three-dimensional dataset that can be quickly generated by a computer, does not rely on external assets, has infinite possibilities and is easy to control is needed. Summary of the Invention

[0005] The purpose of the present application is to provide a method, device, medium and product for generating a three-dimensional dataset by random geometric modeling, which realizes the rapid generation of a three-dimensional dataset by random geometric modeling.

[0006] To achieve the above purpose, the present application provides the following solutions:

[0007] In a first aspect, the present application provides a method for generating a three-dimensional dataset by random geometric modeling, including:

[0008] Create an empty three-dimensional scene at the current loop count in Blender software;

[0009] Randomly generate floating-point numbers of multiple three-dimensional geometric bodies at the current loop count within a set numerical range, and respectively determine the depths of the corresponding three-dimensional geometric bodies based on the floating-point numbers of the three-dimensional geometric bodies at the current loop count and a uniform depth distribution fitting function;

[0010] Randomly generate the spatial parameters, dimensions, and noise textures of each three-dimensional geometric body at the current loop count; the spatial parameters include: position, rotation angle, and scaling ratio;

[0011] In an empty three-dimensional scene at the current loop count, generate multiple initial three-dimensional geometric bodies at the current loop count based on the depth, spatial parameters, and dimensions of each three-dimensional geometric body at the current loop count;

[0012] Apply the noise texture of each three-dimensional geometric body at the current loop count to the corresponding initial three-dimensional geometric body respectively to obtain multiple target three-dimensional geometric bodies at the current loop count, thereby obtaining an initial three-dimensional scene at the current loop count containing multiple target three-dimensional geometric bodies;

[0013] Randomly generate the energy value of the daylight light source at the current loop count to obtain the target energy value at the current loop count, and in the initial three-dimensional scene at the current loop count, set the daylight light source with the target energy value at the current loop count to obtain the target three-dimensional scene at the current loop count;

[0014] Set an orthographic camera for rendering the target three-dimensional scene at the current loop count to obtain an image data group at the current loop count, and add the image data group at the current loop count to the initial data set at the current loop count to obtain the target data set at the current loop count; the image data group includes: depth image and intensity image, the initial data set at the current loop count is the target data set in the previous loop count, and the initial data set in the initial loop count is an empty set;

[0015] Determine whether the current loop count has reached the maximum loop count;

[0016] If so, determine the target data set at the current loop count as the random geometric modeling three-dimensional data set;

[0017] If not, update the current loop count to the next loop count and return "Create an empty three-dimensional scene at the current loop count in Blender software" until the maximum loop count is reached to obtain the random geometric modeling three-dimensional data set.

[0018] Optionally, the set numerical range is [0, 1].

[0019] Optionally, the uniform depth distribution fitting function includes:

[0020]

[0021] Among them, P(r) is the depth of the three-dimensional geometric body; r is the floating-point number of the three-dimensional geometric body.

[0022] Optionally, the process of determining the uniform depth distribution fitting function includes:

[0023] Construct a cumulative distribution function; the cumulative distribution function is:

[0024]

[0025] where C(n) is the value of the cumulative distribution function of the nth depth plane within the line of sight of the orthographic camera; N is the total number of depth planes within the line of sight of the orthographic camera; i and k are the layer numbers of the depth planes within the line of sight of the orthographic camera;

[0026] Obtain the fitting inverse function of the cumulative distribution function by value fitting, and determine the fitting inverse function as the uniform depth distribution fitting function.

[0027] Optionally, the process of generating the noise texture of any three-dimensional geometric body includes:

[0028] Randomly generate the initial number of height pixels of the noise image of the three-dimensional geometric body;

[0029] Determine the initial number of width pixels of the noise image of the three-dimensional geometric body based on the initial number of height pixels;

[0030] Generate an initial noise image based on the initial number of height pixels and the initial number of width pixels of the noise image;

[0031] Interpolate and enlarge the initial noise image according to the preset number of height pixels and the preset number of width pixels to obtain the target noise image of the three-dimensional geometric body;

[0032] Determine the target noise image of the three-dimensional geometric body as the noise texture of the three-dimensional geometric body.

[0033] Optionally, randomly generating the initial number of height pixels of the noise image of the three-dimensional geometric body includes:

[0034] Randomly generate the power number of the noise image of the three-dimensional geometric body;

[0035] Use the initial height pixel number calculation formula to calculate the initial number of height pixels of the noise image of the three-dimensional geometric body according to the power number; the initial height pixel number calculation formula is:

[0036] H noise =2 m ;

[0037] where H noise is the initial number of height pixels; m is the power number.

[0038] Optionally, the noise texture is a low-frequency noise texture, a medium-frequency noise texture, or a high-frequency noise texture;

[0039] When the noise texture is a low-frequency noise texture, the range of values of the power exponent is [1, 2];

[0040] When the noise texture is a medium-frequency noise texture, the range of values of the power exponent is [5, 6];

[0041] When the noise texture is a high-frequency noise texture, the range of values of the power exponent is [9, 10].

[0042] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the method for generating a random geometric modeling three-dimensional data set described in any one of the above.

[0043] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for generating a random geometric modeling three-dimensional data set described in any one of the above is implemented.

[0044] In a fourth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method for generating a random geometric modeling three-dimensional data set described in any one of the above is implemented.

[0045] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application:

[0046] The present application discloses a method, device, medium, and product for generating a random geometric modeling three-dimensional data set. By determining the depth of the corresponding three-dimensional geometric body based on the uniform depth distribution fitting function, the uniformity of the pixels occupied by different depths can be ensured; randomly generating the spatial parameters, sizes, and noise textures of each three-dimensional geometric body makes the obtained three-dimensional geometric bodies and the target three-dimensional scene richer and more flexible, without relying on pre-existing three-dimensional models or external assets, and realizes the rapid generation of a random geometric modeling three-dimensional data set. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0048] Figure 1 It is a schematic flowchart of the method for generating a random geometric modeling three-dimensional data set provided by an embodiment of the present application;

[0049] Figure 2Schematic diagram for rendering a target three-dimensional scene by setting an orthographic camera

[0050] Figure 3 Schematic diagram of a three-dimensional dataset for random geometry modeling

[0051] Figure 4 Schematic diagram of an intensity image with a power of 1 in a three-dimensional dataset for random geometry modeling

[0052] Figure 5 Schematic diagram of a depth image with a power of 2 in a three-dimensional dataset for random geometry modeling

[0053] Figure 6 Schematic diagram of an intensity image with a power of 5 in a three-dimensional dataset for random geometry modeling

[0054] Figure 7 Schematic diagram of a depth image with a power of 6 in a three-dimensional dataset for random geometry modeling

[0055] Figure 8 Schematic diagram of an intensity image with a power of 9 in a three-dimensional dataset for random geometry modeling

[0056] Figure 9 Schematic diagram of a depth image with a power of 10 in a three-dimensional dataset for random geometry modeling

[0057] Figure 10 Schematic diagram of the structure of a computer device provided by an embodiment of the present application Detailed implementation manners

[0058] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0059] The purpose of the present application is to provide a method, device, medium and product for generating a three-dimensional dataset for random geometry modeling, aiming to achieve the rapid generation of a three-dimensional dataset for random geometry modeling.

[0060] To make the above objects, features and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0061] In an exemplary embodiment, as Figure 1 shown, the method for generating a three-dimensional dataset for random geometry modeling in this embodiment includes:

[0062] Step 01: Create an empty 3D scene in the Blender software for the current loop count.

[0063] Step 02: Randomly generate floating-point numbers of multiple 3D geometries for the current loop count within the set numerical range, and respectively determine the depths of the corresponding 3D geometries based on the floating-point numbers of each 3D geometry for the current loop count and the uniform depth distribution fitting function.

[0064] As an alternative implementation, the set numerical range is [0, 1].

[0065] As an alternative implementation, the uniform depth distribution fitting function includes:

[0066]

[0067] where P(r) is the depth of the 3D geometry; r is the floating-point number of the 3D geometry.

[0068] As an alternative implementation, the determination process of the uniform depth distribution fitting function in Step 02 includes:

[0069] Step 021: Construct the cumulative distribution function; the cumulative distribution function is:

[0070]

[0071] where C(n) is the value of the cumulative distribution function of the nth depth plane within the line of sight of the orthographic camera; N is the total number of depth planes within the line of sight of the orthographic camera; i and k are the layer numbers of the depth planes within the line of sight of the orthographic camera.

[0072] Specifically, the construction process of the cumulative distribution function includes:

[0073] Let a series of variables z 0 , z 1 , …, z N respectively represent the 0th to Nth depth planes within the line of sight of the orthographic camera. When converting a continuous geometry scene to a multi-depth scene, the pixels of the 3D geometry on the (n - 1)th depth plane z n-1 and the nth depth plane z n within the line of sight of the orthographic camera are projected and compressed to z n-1 . To achieve uniform depth distribution, theoretically, the pixels of the projected 3D geometry on each depth layer should have equal statistical proportions, that is:

[0074]

[0075] where p znis the probability density function value of the nth depth plane within the line of sight of the orthographic camera.

[0076] However, since there may be occlusions in the three-dimensional geometries at different positions, it is necessary to calculate the corresponding occlusion probabilities. The pixel occupancy probabilities of the projections of the three-dimensional geometries distributed at z 0 , z 1 , …, z N can be respectively expressed as harmonic series 1. Further calculation gives the probability density function considering occlusion, which is used to determine the depth of objects in the scene. The probability density function considering occlusion is defined as:

[0077]

[0078] where p(n) is the probability density function value considering occlusion of the nth depth plane within the line of sight of the orthographic camera.

[0079] Furthermore, cumulative integration operation is performed on the probability density function considering occlusion to obtain the cumulative distribution function.

[0080] Step 022: Obtain the fitting inverse function of the cumulative distribution function through value fitting, and determine the fitting inverse function as the uniform depth distribution fitting function.

[0081] Step 03: Randomly generate the spatial parameters, sizes, and noise textures of the three-dimensional geometries in the current loop; the spatial parameters include: position, rotation angle, and scaling ratio.

[0082] Specifically, when randomly generating the position (x, y), the value range of the x-axis component of the position is [-1.92, 1.92], and the value range of the y-axis component of the position is [-1.08, 1.08]. The scaling ratio can also be set according to actual needs. For the three-dimensional geometries of cube, regular tetrahedron, and sphere respectively, the scaling ratios of the three are set to 1.61199:3:1. This scaling ratio can make the volumes of the three basically the same, better ensuring the uniformity of depth distribution. When actually assigning values, the sizes of the three-dimensional geometries should not be too large, which needs to be less than half of the target three-dimensional scene to be rendered, nor too small, which will cause too many models and reduce the rendering efficiency.

[0083] As an alternative implementation, in step 03, the generation process of the noise texture of any three-dimensional geometry includes:

[0084] Step 031: Randomly generate the initial number of height pixels of the noise image of the three-dimensional geometry.

[0085] As an alternative implementation, step 031 includes:

[0086] Step 0311: Randomly generate the power number of the noise image of the three-dimensional geometric body.

[0087] Step 0312: Use the initial height pixel number calculation formula to calculate the initial height pixel number of the noise image of the three-dimensional geometric body according to the power number; the initial height pixel number calculation formula is:

[0088] H noise = 2 m .

[0089] where H noise is the initial height pixel number; m is the power number.

[0090] Step 032: Determine the initial width pixel number of the noise image of the three-dimensional geometric body based on the initial height pixel number.

[0091] Step 033: Generate an initial noise image based on the initial height pixel number and the initial width pixel number of the noise image.

[0092] Step 034: Interpolate and enlarge the initial noise image according to the preset height pixel number and the preset width pixel number to obtain the target noise image of the three-dimensional geometric body.

[0093] Step 035: Determine the target noise image of the three-dimensional geometric body as the noise texture of the three-dimensional geometric body.

[0094] As an optional implementation manner, the noise texture is a low-frequency noise texture, a medium-frequency noise texture, or a high-frequency noise texture.

[0095] When the noise texture is a low-frequency noise texture, the value range of the power number is [1, 2].

[0096] When the noise texture is a medium-frequency noise texture, the value range of the power number is [5, 6].

[0097] When the noise texture is a high-frequency noise texture, the value range of the power number is [9, 10].

[0098] Step 04: In the empty three-dimensional scene at the current loop count, generate multiple initial three-dimensional geometric bodies at the current loop count based on the depth, spatial parameters, and sizes of the three-dimensional geometric bodies at the current loop count.

[0099] Step 05: Apply the noise texture of each three-dimensional geometric body at the current loop count to the corresponding initial three-dimensional geometric body respectively to obtain multiple target three-dimensional geometric bodies at the current loop count, thereby obtaining the initial three-dimensional scene at the current loop count containing multiple target three-dimensional geometric bodies.

[0100] Step 06: Randomly generate the energy value of the daylight light source at the current loop count to obtain the target energy value at the current loop count, and in the initial three-dimensional scene at the current loop count, set the daylight light source with the target energy value at the current loop count to obtain the target three-dimensional scene at the current loop count.

[0101] Specifically, when randomly generating the target energy value, the value range is [5, 20]. If the target energy value is too small or too large, it will cause more low-frequency components and reduce the generalization ability of the random geometric modeling three-dimensional data set.

[0102] Step 07: Set an orthographic camera to render the target three-dimensional scene at the current loop count to obtain the image data group at the current loop count, and add the image data group at the current loop count to the initial data set at the current loop count to obtain the target data set at the current loop count.

[0103] Among them, the image data group includes: depth image and intensity image. The initial data set at the current loop count is the target data set in the previous loop count, and the initial data set in the initial loop count is an empty set.

[0104] Specifically, the schematic diagram of setting an orthographic camera to render the target three-dimensional scene is as Figure 2 shown. The orthographic multiple of the orthographic camera is 3.84. The orthographic multiple is the width of the depth image and the intensity image in the rendered image data group. According to the 4K ratio, the height of the depth image and the intensity image can be obtained as 2.16. The position of the orthographic camera is set to (0, 0, 1) to ensure that it will not be blocked or penetrated by the three-dimensional geometry.

[0105] The rendering resolution is 3840×2160 to obtain the rendered image data. Both the depth image and the intensity image in the rendered image data group are presented in RGB mode.

[0106] Step 08: Determine whether the current loop count has reached the maximum loop count.

[0107] Step 09: If so, determine the target data set at the current loop count as the random geometric modeling three-dimensional data set.

[0108] Step 10: If not, update the current loop count to the next loop count and return to Step 01 until the maximum loop count is reached to obtain the random geometric modeling three-dimensional data set.

[0109] The random geometric modeling three-dimensional data set is as Figure 3As shown. In the 3D dataset of random geometric modeling, the intensity images with power number 1, the depth images with power number 2, the intensity images with power number 5, the depth images with power number 6, the intensity images with power number 9, and the depth images with power number 10 are as Figures 4 - 9 shown.

[0110] This application introduces a uniform depth distribution fitting function to control the depth of 3D geometric bodies, ensuring the uniformity of pixels at different depths, so that the depth distribution law of 3D geometric bodies in the target 3D scene can be better used for neural network training, preventing the neural network from being biased towards any frequently occurring depth and producing adverse results at those sparse depths when there is an uneven pixel depth distribution. The 3D dataset of random geometric modeling has more detailed information such as angles, edges, occlusions, etc. and background information of light at the same time. Compared with traditional datasets, the image information it contains is richer, and the distribution of high-frequency and low-frequency components in the spectrum can also be flexibly adjusted. It allows for the flexible design of 3D scenes by setting parameters without relying on pre-existing 3D models or external assets. The automation of the entire dataset generation is achieved through the Python script of Blender, including the random generation of 3D geometric bodies to the rendering and saving of images, and the entire process is automatically executed by the program. This automated control greatly reduces the time and effort required for human intervention, and through parameterized settings, users can easily adjust the characteristics of the generated 3D geometric bodies and 3D scenes. These parameter settings make the generated geometric modeling scenes not only have rich diversity but also can flexibly adjust the scene complexity and geometric body attributes according to specific requirements. By randomly generating the positions, sizes, and rotation angles of 3D geometric bodies, the diversity and unpredictability of the 3D dataset of random geometric modeling are ensured, which is particularly important for the training of deep learning models. The application of random noise textures and the setting of random positions and rotation angles make each rendered target 3D scene unique, thus avoiding the homogenization problem caused by repetitive data in traditional methods.

[0111] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the method for generating a 3D dataset of random geometric modeling.

[0112] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the method for generating a 3D dataset of random geometric modeling is implemented.

[0113] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method for generating a 3D dataset of random geometric modeling is implemented.

[0114] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal, and its internal structure diagram may be as shown in Figure 10 . The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes a method for generating a three-dimensional dataset of random geometric modeling.

[0115] Those skilled in the art can understand that Figure 10 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0116] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0117] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include Read-Only Memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0118] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0119] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0120] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for generating a three-dimensional data set for random geometric modeling, characterized in that: The method for generating the random geometric modeling three-dimensional data set includes: Create an empty 3D scene at the current loop number in Blender software; Randomly generate floating point numbers of multiple three-dimensional geometric bodies at the current number of cycles within a set value range, and determine the depth of the corresponding three-dimensional geometric body based on the floating point number of each three-dimensional geometric body at the current number of cycles and a uniform depth distribution fitting function; Randomly generate the spatial parameters, size and noise texture of each three-dimensional geometric body under the current number of cycles; the spatial parameters include: position, rotation angle and scaling; In the empty three-dimensional scene at the current loop number, based on the depth, spatial parameters and size of each three-dimensional geometric body at the current loop number, a plurality of initial three-dimensional geometric bodies at the current loop number are generated; Apply the noise texture of each three-dimensional geometric body at the current loop number to the corresponding initial three-dimensional geometric body respectively, to obtain multiple target three-dimensional geometric bodies at the current loop number, thereby obtaining an initial three-dimensional scene at the current loop number containing multiple target three-dimensional geometric bodies; Randomly generate an energy value of a fluorescent lamp light source at the current cycle number to obtain a target energy value at the current cycle number, and set a fluorescent lamp light source with the target energy value at the current cycle number in an initial three-dimensional scene at the current cycle number to obtain a target three-dimensional scene at the current cycle number; An orthogonal camera is set to render the target three-dimensional scene at the current cycle number to obtain an image data group at the current cycle number, and the image data group at the current cycle number is added to the initial data set at the current cycle number to obtain the target data set at the current cycle number; the image data group includes: a depth image and an intensity image, the initial data set at the current cycle number is the target data set in the previous cycle number, and the initial data set in the initial cycle number is an empty set; Determine whether the current number of loops has reached the maximum number of loops; If so, the target data set at the current number of cycles is determined as a random geometric modeling three-dimensional data set; If not, the current loop number is updated to the next loop number, and the method "create an empty 3D scene under the current loop number in Blender software" is returned until the maximum loop number is reached to obtain a random geometric modeling 3D data set.

2. The method for generating a three-dimensional data set for random geometric modeling according to claim 1, characterized in that: The setting value range is [0, 1].

3. The method for generating a three-dimensional data set for random geometric modeling according to claim 1, characterized in that: The uniform depth distribution fitting function comprises: Wherein, P(r) is the depth of the three-dimensional geometric body; r is the floating point number of the three-dimensional geometric body.

4. The method for generating a three-dimensional data set for random geometric modeling according to claim 3, characterized in that: The process of determining the uniform depth distribution fitting function includes: Construct a cumulative distribution function; the cumulative distribution function is: Wherein, C(n) is the cumulative distribution function value of the nth depth plane in the line of sight of the orthogonal camera; N is the total number of depth planes in the line of sight of the orthogonal camera; i and k are both the layer numbers of the depth plane in the line of sight of the orthogonal camera; The inverse fitting function of the cumulative distribution function is obtained by value fitting, and the inverse fitting function is determined as the uniform depth distribution fitting function.

5. The method for generating a three-dimensional data set for random geometric modeling according to claim 1, characterized in that: The process of generating noise texture of any 3D geometry includes: The initial height pixel number of the noise image of the randomly generated 3D geometry; Determine the initial width pixel number of the noise image of the three-dimensional geometric body based on the initial height pixel number; Generate an initial noise image based on an initial height pixel number and an initial width pixel number of the noise image; The initial noise image is interpolated and amplified according to a preset number of height pixels and a preset number of width pixels to obtain a target noise image of a three-dimensional geometric body; A target noise image of the three-dimensional geometry is determined as a noise texture of the three-dimensional geometry.

6. The method for generating a three-dimensional data set for random geometric modeling according to claim 5, characterized in that: The number of initial height pixels of a randomly generated noise image of a 3D geometry, including: The power number of the noise image of the randomly generated 3D geometry; The initial height pixel number calculation formula is used to calculate the initial height pixel number of the noise image of the three-dimensional geometric body according to the power number; the initial height pixel number calculation formula is: H noise ＝2 m ； Among them, H noise is the initial height pixel number; m is the power number.

7. The method for generating a three-dimensional data set for random geometric modeling according to claim 6, characterized in that: The noise texture is a low-frequency noise texture, a medium-frequency noise texture or a high-frequency noise texture; When the noise texture is a low-frequency noise texture, the power value range is [1, 2]; When the noise texture is a medium-frequency noise texture, the power value range is [5, 6]; When the noise texture is a high-frequency noise texture, the power number ranges from [9, 10].

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for generating a three-dimensional data set for random geometric modeling as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a three-dimensional data set for random geometric modeling as described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for generating a three-dimensional data set for random geometric modeling as described in any one of claims 1 to 7 is implemented.

Images