Farmland simulation terrain generation method, device, medium and system

Abstract

The invention provides a method, a device, a medium and a system for generating a simulated farmland terrain. Poisson curved surface reconstruction processing is respectively carried out on the real farmland point cloud and all the test random point clouds once to obtain a plurality of test farmland terrains and a plurality of corresponding test flatness, so that the construction of the farmland terrains has the advantages of high usability, no need of modeling experience, low use cost, high efficiency and time cost reduction; therefore, compared with the existing scheme, the method has the advantages of low time cost and high efficiency, and the problems of low construction efficiency and high time cost in the construction process of the farmland simulation terrain in the existing scheme are solved.

Description

Technical Field

[0001] The present application relates to the technical field of farmland simulation, and in particular to a method, device, medium and system for generating farmland simulation terrain. Background Art

[0002] There are relatively few solutions for simulating unstructured road scenes such as farmland. There are currently two main technical solutions: one is artificial 3D modeling, and the other is 3D reconstruction based on real terrain data.

[0003] The manual 3D modeling solution is to use 3D modeling software (such as 3Dsmax, Blender, CATIA, etc.), which requires personnel with modeling experience to draw models manually in the modeling software. In the drawing, the real terrain is referenced, and the concave and convex feeling of the farmland terrain is simulated through manual rotation, translation and stretching, and finally the required terrain surface is drawn.

[0004] A three-dimensional reconstruction solution based on real terrain data. Existing patents are more inclined to use existing data for three-dimensional reconstruction and digital twins. For example, in the surveying and mapping field, satellites are used to scan the ground elevation data to reconstruct the earth's surface. In the field of virtual reality, lidar or cameras are used to capture point cloud data of real objects, form an initial grid based on the point cloud data, and then reconstruct the model surface according to their respective optimization algorithms to finally form the required model. Although this solution can reconstruct a realistic model, it cannot be generalized. This disadvantage is reflected in farmland, that is, each scan and reconstruction can only generate one terrain model, and the surface of the model is fixed and cannot be applied to different scenes.

[0005] That is, manually using 3D modeling software to construct the terrain model from scratch, which requires modelers to have more modeling experience and invest a lot of time cost.

[0006] In summary, the existing solutions have the problems of low construction efficiency and high time cost in the process of constructing farmland simulation terrain. Summary of the invention

[0007] The main purpose of the present application is to provide a method, device, medium and system for generating farmland simulation terrain, so as to at least solve the problem of low construction efficiency and high time cost in the process of constructing farmland simulation terrain in existing solutions.

[0008] In order to achieve the above object, according to one aspect of the present application, a method for generating a simulated farmland terrain is provided, the method comprising:

[0009] Acquire current farmland data, wherein the current farmland data includes current terrain length, current terrain width and current terrain flatness;

[0010] The current terrain flatness is processed by using a neural network model technology to obtain an optimal height and an optimal number of clouds;

[0011] Using random point cloud generation technology, based on the optimal height, the optimal point cloud quantity, the current terrain length and the current terrain width, a current random point cloud is generated;

[0012] The current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain a final farmland simulation terrain, and the final farmland simulation terrain is imported into a simulation platform to simulate a preset scene.

[0013] Optionally, the current terrain flatness is processed using a neural network model technology to obtain an optimal height and an optimal number of clouds, including:

[0014] Using the current terrain flatness as an input of an optimal neural network model so that the optimal neural network model processes the current terrain flatness;

[0015] Obtaining an output of the optimal neural network model;

[0016] The output of the optimal neural network model is determined as the optimal height and the optimal number of point clouds.

[0017] Optionally, before using the neural network model technology to process the current terrain flatness to obtain the optimal height and the optimal number of clouds, the method further includes:

[0018] Acquire multiple test random point clouds and real farmland point clouds, wherein the real farmland point clouds are point clouds obtained by scanning the real farmland with a data acquisition device;

[0019] Perform a Poisson surface reconstruction process on the real farmland point cloud and all the test random point clouds respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses;

[0020] According to the size of the test flatness, a training data set is screened out from all the test farmland terrains, the training data set includes multiple groups of training data, each group of the training data includes a test height, a number of test point clouds and the corresponding test flatness;

[0021] The initial neural network model is iteratively trained using the training data set, and the current neural network model is determined to be the optimal neural network model at least when the number of iterations reaches a preset maximum number of iterations.

[0022] Optionally, obtain multiple test random point clouds, including:

[0023] Acquire multiple groups of test data, each group of test data includes an initial test farmland length, an initial test farmland width, an initial test height, and an initial test point cloud quantity;

[0024] The random point cloud generation technology is used to process the multiple groups of test data respectively to generate the multiple test random point clouds.

[0025] Optionally, obtain a real farmland point cloud, including:

[0026] Acquire a real scan image acquired by the data acquisition device;

[0027] The real scanned image is processed by using Colmap sparse point cloud building technology to generate the real farmland point cloud.

[0028] Optionally, according to the size of the test flatness, a training data set is screened out from all the test farmland terrains, including:

[0029] When the test flatness is within a preset range, determining to select the training data set from the test farmland terrain corresponding to the test flatness;

[0030] When the test flatness is not within the preset range, it is determined not to filter the training data set from the test farmland terrain corresponding to the test flatness.

[0031] Optionally, the current random point cloud is subjected to Poisson surface reconstruction and texture rendering in sequence to obtain a final farmland simulation terrain, including:

[0032] After texture rendering is performed on the result obtained by the Poisson surface reconstruction process, the result obtained by the Poisson surface reconstruction process is optimized by using Open3D to smooth the mesh curvature of the result obtained by the Poisson surface reconstruction process, and bridge the cross section of the result obtained by the Poisson surface reconstruction process to obtain an intermediate farmland simulation surface;

[0033] The intermediate farmland simulation surface is texture rendered to obtain the final farmland simulation terrain.

[0034] According to another aspect of the present application, a device for generating a simulated farmland terrain is provided, the device comprising:

[0035] A first acquisition unit, used to acquire current farmland data, wherein the current farmland data includes current terrain length, current terrain width and current terrain flatness;

[0036] A first processing unit is used to process the current terrain flatness using a neural network model technology to obtain an optimal height and an optimal number of clouds;

[0037] A second processing unit is used to generate a current random point cloud based on the optimal height, the optimal point cloud quantity, the current terrain length and the current terrain width by using a random point cloud generation technology;

[0038] The third processing unit is used to perform Poisson surface reconstruction and texture rendering on the current random point cloud in sequence to obtain a final farmland simulation terrain, and import the final farmland simulation terrain into a simulation platform to simulate a preset scene.

[0039] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods for generating farmland simulation terrain.

[0040] According to another aspect of the present application, a system for generating a farmland simulation terrain is provided, the system comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing any one of the methods for generating a farmland simulation terrain.

[0041] By applying the technical solution of the present application, the real farmland point cloud and all the test random point clouds are subjected to Poisson surface reconstruction processing respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses, so that the construction of the farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is adopted to generate the current random point cloud based on the optimal height, the optimal number of point clouds, the current terrain length and the current terrain width, and the current random point cloud is subjected to Poisson surface reconstruction processing and texture rendering processing in turn to obtain the final farmland simulation terrain, and the final farmland simulation terrain is imported into the simulation platform for simulation of the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem of low construction efficiency and high time cost in the process of constructing the farmland simulation terrain in the existing scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 A schematic diagram of a flow chart of a method for generating a simulated farmland terrain according to an embodiment of the present application is shown;

[0044] Figure 2 A schematic diagram of a process for obtaining an optimal height and an optimal number of clouds according to an embodiment of the present application is shown;

[0045] Figure 3 A schematic diagram of a process for determining an optimal neural network model according to an embodiment of the present application is shown;

[0046] Figure 4 A schematic diagram of a process for selecting a training data set from all test farmland terrains according to the size of the test flatness provided in an embodiment of the present application is shown;

[0047] Figure 5 A schematic diagram of a process of generating another method for simulated farmland terrain according to an embodiment of the present application is shown;

[0048] Figure 6 A structural block diagram of a device for generating simulated farmland terrain according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0049] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0050] In order to enable those skilled in the art to better understand the solution of the present application, 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 only 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 should fall within the scope of protection of the present application.

[0051] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0052] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0053] Colmap sparse point cloud building technology is a technology based on camera positioning and 3D reconstruction algorithm, which is used to extract sparse 3D point cloud models from a set of images. By extracting and matching features from images, as well as positioning and estimating the camera's posture, Colmap can reconstruct the 3D point cloud model in the scene.

[0054] Open3D is an open source C++ library for processing 3D data. It provides a range of functions, including point cloud processing, 3D reconstruction, geometry processing, visualization, etc. Open3D can be used to process 3D data obtained from various sensors and devices, such as laser scanning, RGB-D cameras, etc. It is widely used in computer vision, robotics, virtual reality and other fields.

[0055] ReLU (Rectified Linear Unit) is an activation function commonly used in neural networks. It outputs the input value when the input is greater than 0 and outputs 0 when it is less than or equal to 0.

[0056] The SGD algorithm (Stochastic Gradient Descent) is an optimization algorithm used to train neural networks. It minimizes the loss function by continuously iteratively updating model parameters.

[0057] Regularization technology is a method used to prevent overfitting. By adding regularization terms to the loss function, the model complexity is penalized, making the model simpler and more generalizable.

[0058] MSE (mean squared error) is a commonly used loss function used to evaluate the difference between the predicted value and the actual value of the model on the training set. The calculation method is the average of the squares of the difference between the predicted value and the actual value.

[0059] 3ds Max is a professional 3D modeling, animation and rendering software. It is widely used in film and television, games, architecture and industrial design, and can help users create high-quality 3D models, animations and visual effects. 3ds Max provides powerful modeling, animation, rendering and special effects tools, allowing users to realize various complex creative ideas.

[0060] As introduced in the background technology, a three-dimensional reconstruction solution based on real terrain data. Existing patents currently prefer to use existing data for three-dimensional reconstruction and digital twins. For example, in the surveying and mapping field, satellites are used to scan the ground elevation data to reconstruct the earth's surface. In the field of virtual reality, lidar or cameras are used to capture point cloud data of real objects, form an initial grid based on the point cloud data, and then reconstruct the model surface according to their respective optimization algorithms to finally form the required model. Although this solution can reconstruct a realistic model, it cannot be generalized. This disadvantage is reflected in farmland, that is, each scan and reconstruction can only generate one terrain model, and the surface of the model is fixed and cannot be applied to different scenes. In order to solve the problem that the existing solution has low construction efficiency and high time cost in the process of constructing farmland simulation terrain, the embodiments of the present application provide a method, device, medium and system for generating farmland simulation terrain.

[0061] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0062] In this embodiment, a method for generating a simulated farmland terrain is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0063] Figure 1 It is a flow chart of a method for generating a simulated farmland terrain according to an embodiment of the present application.

[0064] like Figure 1 As shown, the method comprises the following steps:

[0065] Step S101, obtaining current farmland data, the current farmland data including current terrain length, current terrain width and current terrain flatness;

[0066] Step S102, using a neural network model technology to process the current terrain flatness to obtain an optimal height and an optimal number of clouds;

[0067] like Figure 2 As shown, step S102, i.e., using the neural network model technology to process the current terrain flatness to obtain the optimal height and the optimal number of clouds, includes the following steps:

[0068] Step S201, using the current terrain flatness as an input of an optimal neural network model, so that the optimal neural network model processes the current terrain flatness;

[0069] Step S202, obtaining the output of the above optimal neural network model;

[0070] Step S203, determining the output of the above-mentioned optimal neural network model as the above-mentioned optimal height and the above-mentioned optimal number of point clouds.

[0071] Specifically, a neural network learning model is used to directly obtain a terrain model that meets the test requirements by inputting parameters in advance, so that it is convenient to obtain the optimal height and the optimal number of clouds based on the current terrain flatness in the current farmland data obtained.

[0072] The neural network model can be tested using random point clouds with real collected data.

[0073] like Figure 3 As shown, before step S102, i.e., using the neural network model technology to process the current terrain flatness to obtain the optimal height and the optimal number of clouds, the method further includes the following steps:

[0074] Step S301, obtaining a plurality of test random point clouds and a real farmland point cloud, wherein the real farmland point cloud is a point cloud obtained by scanning the real farmland with a data acquisition device;

[0075] The test random point cloud is a preset point cloud, and the real farmland point cloud is subsequently used to correct the test random point cloud to improve the accuracy of the neural network.

[0076] In one embodiment of the present application, the step S301 of obtaining a plurality of test random point clouds includes:

[0077] Acquire multiple groups of test data, each group of the test data includes an initial test farmland length, an initial test farmland width, an initial test height, and an initial test point cloud quantity;

[0078] The above-mentioned multiple groups of test data are processed respectively using the above-mentioned random point cloud generation technology to generate the above-mentioned multiple test random point clouds.

[0079] Specifically, by using random point cloud generation technology, multiple test random point clouds can be generated to facilitate subsequent point cloud correction.

[0080] In one embodiment of the present application, obtaining a real farmland point cloud in step S301 includes:

[0081] Acquire a real scan image acquired by the above data acquisition device;

[0082] The above-mentioned real scanned image is processed by using Colmap sparse point cloud building technology to generate the above-mentioned real farmland point cloud.

[0083] Specifically, the real scanned images can be collected by using lidar and cameras, and then the Colmap sparse point cloud building technology is used to process the above-mentioned real scanned images to generate the above-mentioned real farmland point cloud, so as to facilitate the subsequent use of the real farmland point cloud to correct the test random point cloud, so that the test farmland terrain is more in line with the actual farmland terrain.

[0084] Step S302, performing a Poisson surface reconstruction process on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively, to obtain multiple test farmland terrains and multiple corresponding test flatnesses;

[0085] Farmland terrain is different from asphalt and cement roads. It is composed of uneven and undulating surfaces. According to traditional 3D modeling methods, to construct an unstructured terrain similar to farmland, you need to use 3D modeling software to manually draw a rectangular surface with a length of X and a width of Y, and then divide this surface into multiple sub-surfaces, and then manually stretch and carve all the sub-surfaces. After repeated and lengthy operations, the initial 3D model can finally be formed.

[0086] However, the surface flatness of the terrain model cannot be constrained from the beginning of modeling, resulting in great randomness in the flatness after modeling. Therefore, this application provides a method of taking the terrain length (X), width (Y) and flatness (P) as pre-inputs, and then generating a terrain model that meets the pre-conditions and can be imported into the simulation platform through an automated modeling method. Then, how to associate the pre-input with the output terrain model will be the focus of this application.

[0087] First, we need to automatically generate 3D surface terrain, and ensure that the structure of the terrain generated each time is different, so we use random point cloud generation and point cloud surface reconstruction methods. If the length, width, and height of a cuboid, as well as the number of point clouds to be generated, are known, we can use the random method to randomly generate point clouds inside the cuboid. The length (X) and width (Y) of the cuboid are known pre-input parameters, so the point cloud density can only be controlled by the height (Z) and the number of point clouds (N). After generating the point cloud, use Poisson surface reconstruction technology to connect the independent point clouds into a similar surface composed of mesh grids, and then smooth the surface to form a surface model.

[0088] Then, the surface model can be generated from the known length (X) and width (Y) and the height (Z) and number of point clouds (N) that require secondary input, but there is no constraint on flatness (P), so the problem becomes that a surface model that satisfies P needs to be generated through Z and N, that is, it is necessary to find the correspondence from Z and N to P. Therefore, deep learning is introduced to train a network model that can express this correspondence, with the input of the network being P and the output being Z and N.

[0089] Step S303, according to the size of the above-mentioned test flatness, a training data set is selected from all the above-mentioned test farmland terrains, and the above-mentioned training data set includes multiple groups of training data, and each group of the above-mentioned training data includes a test height, a number of test point clouds and the corresponding above-mentioned test flatness;

[0090] like Figure 4 As shown, step S303, i.e., selecting a training data set from all the above-mentioned test farmland terrains according to the size of the above-mentioned test flatness, includes the following steps:

[0091] Step S401, when the test flatness is within a preset range, determine to select the training data set from the test farmland terrain corresponding to the test flatness;

[0092] Step S402: When the test flatness is not within the preset range, determine not to select the training data set from the test farmland terrain corresponding to the test flatness.

[0093] Specifically, the preset range is [0,3]. When the test flatness is within [0,3], it is determined that the training data set is selected from the test farmland terrain corresponding to the test flatness, which means that the farmland terrain corresponding to the test flatness meets the requirements. When the test flatness is not within [0,3], it is determined not to select the training data set from the test farmland terrain corresponding to the test flatness, which means that the farmland terrain corresponding to the test flatness does not meet the requirements.

[0094] Step S304, using the above training data set to iteratively train the initial neural network model, and determining the current neural network model as the optimal neural network model at least when the number of iterations reaches a preset maximum number of iterations.

[0095] Specifically, the BP neural network training model is used to clean the data based on the existing data set: first, the data in the data set is cleaned and standardized to ensure data quality; second, a neural network structure with an appropriate number of layers is built, where the number of neurons in the input layer is set to 1, the number of neurons in the output layer is set to 2, and the hidden layer is set according to the complexity. In this application, it can be set to 2 layers, and the number of neurons in each hidden layer is 10. At the same time, the activation function selects ReLU to prevent the gradient from disappearing during training. The optimizer selects the SGD algorithm and combines regularization technology to prevent overfitting. The MSE mean square error function is used as the objective function to determine the update direction of the parameters and determine the hyperparameters in the network such as the learning rate and the maximum number of iterations for training to obtain a neural network model.

[0096] The preset maximum number of iterations can be 100.

[0097] Step S103, using a random point cloud generation technology to generate a current random point cloud based on the optimal height, the optimal point cloud quantity, the current terrain length, and the current terrain width;

[0098] Step S104, performing Poisson surface reconstruction processing and texture rendering processing on the above-mentioned current random point cloud in turn to obtain the final farmland simulation terrain, and importing the above-mentioned final farmland simulation terrain into the simulation platform to simulate the preset scene.

[0099] In the above steps, the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds are respectively subjected to Poisson surface reconstruction processing to obtain multiple test farmland terrains and multiple corresponding test flatnesses, so that the construction of the farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is adopted to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the above-mentioned current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform for simulation of the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem that the existing scheme has low construction efficiency and high time cost in the process of constructing the farmland simulation terrain.

[0100] The preset scene may be a scene simulating the different effects of different weather conditions on farmland terrain.

[0101] In one embodiment of the present application, the above-mentioned current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain the final farmland simulation terrain, including:

[0102] After texture rendering is performed on the result obtained by the Poisson surface reconstruction process, the result obtained by the Poisson surface reconstruction process is optimized by using Open3D to smooth the mesh curvature of the result obtained by the Poisson surface reconstruction process, and bridge the cross section of the result obtained by the Poisson surface reconstruction process to obtain an intermediate farmland simulation surface;

[0103] The above intermediate farmland simulation surface is texture rendered to obtain the above final farmland simulation terrain.

[0104] After using Open3D to smooth the mesh curvature and bridge the sections, you can use 3ds Max for texture rendering to get the final farmland simulation terrain.

[0105] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the method for generating simulated farmland terrain of the present application will be described in detail below in conjunction with specific embodiments.

[0106] This embodiment relates to a specific method for generating farmland simulation terrain. Figure 5 As shown, the following steps are included:

[0107] Acquire multiple groups of test data, each group of test data includes an initial test farmland length, an initial test farmland width, an initial test height, and an initial test point cloud quantity; use a random point cloud generation technology to process the multiple groups of test data respectively to generate multiple test random point clouds;

[0108] Acquire a real scanned image collected by a data acquisition device; use Colmap sparse point cloud building technology to process the real scanned image to generate a real farmland point cloud, where the real farmland point cloud is a point cloud obtained by scanning a real farmland using a data acquisition device;

[0109] The real farmland point cloud and all the test random point clouds are reconstructed by Poisson surface respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses;

[0110] When the test flatness is within a preset range, it is determined to select a training data set from the test farmland terrain corresponding to the test flatness; when the test flatness is not within the preset range, it is determined not to select the training data set from the test farmland terrain corresponding to the test flatness, the training data set includes multiple groups of training data, each group of training data includes a test height, a number of test point clouds and a corresponding test flatness;

[0111] Iteratively training the initial neural network model using the training data set, and determining that the current neural network model is the optimal neural network model at least when the number of iterations reaches a preset maximum number of iterations;

[0112] Obtain current farmland data, including current terrain length, current terrain width and current terrain flatness; and use the optimal neural network model to process the current terrain flatness to obtain the optimal height and optimal number of clouds;

[0113] Using random point cloud generation technology, the current random point cloud is generated based on the optimal height, the optimal number of point clouds, the current terrain length and the current terrain width;

[0114] After texture rendering is performed on the result obtained by Poisson surface reconstruction of the current random point cloud, Open3D is used to optimize the result obtained by Poisson surface reconstruction to smooth the mesh curvature of the result obtained by Poisson surface reconstruction, and bridge the cross section of the result obtained by Poisson surface reconstruction to obtain the intermediate farmland simulation surface;

[0115] The intermediate farmland simulation surface is texture rendered to obtain the final farmland simulation terrain; the final farmland simulation terrain is imported into the simulation platform to simulate the preset scene.

[0116] By performing Poisson surface reconstruction processing on the real farmland point cloud and all the test random point clouds respectively, multiple test farmland terrains and multiple corresponding test flatnesses are obtained, so that the construction of the farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is used to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the above-mentioned current random point cloud is successively subjected to Poisson surface reconstruction processing and texture rendering processing to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform for simulation of the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem of low construction efficiency and high time cost in the process of constructing the farmland simulation terrain in the existing scheme.

[0117] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0118] The embodiment of the present application also provides a device for generating farmland simulation terrain. It should be noted that the device for generating farmland simulation terrain of the embodiment of the present application can be used to execute the method for generating farmland simulation terrain provided by the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions thereof will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0119] The following is an introduction to the device for generating simulated farmland terrain provided in an embodiment of the present application.

[0120] Figure 61 is a structural block diagram of a device for generating farmland simulation terrain according to an embodiment of the present application. Figure 6 As shown, the device comprises:

[0121] A first acquisition unit 61 is used to acquire current farmland data, where the current farmland data includes current terrain length, current terrain width and current terrain flatness;

[0122] The first processing unit 62 is used to process the current terrain flatness using a neural network model technology to obtain an optimal height and an optimal number of clouds;

[0123] The second processing unit 63 is used to generate a current random point cloud based on the optimal height, the optimal point cloud quantity, the current terrain length and the current terrain width by using a random point cloud generation technology;

[0124] The third processing unit 64 is used to perform Poisson surface reconstruction processing and texture rendering processing on the above-mentioned current random point cloud in sequence to obtain the final farmland simulation terrain, and import the above-mentioned final farmland simulation terrain into the simulation platform to simulate the preset scene.

[0125] In the above-mentioned device, by performing Poisson surface reconstruction processing on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively, multiple test farmland terrains and multiple corresponding test flatnesses are obtained, so that the construction of the farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is adopted to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the above-mentioned current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform for simulation of the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem that the existing scheme has low construction efficiency and high time cost in the process of constructing the farmland simulation terrain.

[0126] In one embodiment of the present application, the first processing unit includes a first processing module, a first acquisition module and a second processing module.

[0127] The first processing module is used to use the current terrain flatness as an input of the optimal neural network model, so that the optimal neural network model processes the current terrain flatness;

[0128] The first acquisition module is used to obtain the output of the above optimal neural network model;

[0129] The second processing module is used to determine the output of the above-mentioned optimal neural network model as the above-mentioned optimal height and the above-mentioned optimal number of point clouds.

[0130] In one embodiment of the present application, the apparatus further includes a second acquisition unit, a fourth processing unit, a fifth processing unit, and a sixth processing unit. Before the current terrain flatness is processed by using a neural network model technology to obtain the optimal height and the optimal number of clouds, the method further includes:

[0131] The second acquisition unit is used to acquire a plurality of test random point clouds and a real farmland point cloud, wherein the real farmland point cloud is a point cloud obtained by scanning the real farmland with a data acquisition device;

[0132] The fourth processing unit is used to perform a Poisson surface reconstruction process on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively, to obtain multiple test farmland terrains and multiple corresponding test flatnesses;

[0133] The fifth processing unit is used to select a training data set from all the above-mentioned test farmland terrains according to the size of the above-mentioned test flatness, and the above-mentioned training data set includes multiple groups of training data, and each group of the above-mentioned training data includes a test height, a number of test point clouds and the corresponding above-mentioned test flatness;

[0134] The sixth processing unit is used to iteratively train the initial neural network model using the above training data set, and determine that the current neural network model is the optimal neural network model at least when the number of iterations reaches a preset maximum number of iterations.

[0135] In one embodiment of the present application, the second acquisition unit includes a second acquisition module and a first generation module;

[0136] The second acquisition module is used to acquire multiple groups of test data, each group of the above test data includes an initial test farmland length, an initial test farmland width, an initial test height and an initial test point cloud quantity;

[0137] The first generating module is used to process the above-mentioned multiple groups of test data respectively by adopting the above-mentioned random point cloud generation technology to generate the above-mentioned multiple test random point clouds.

[0138] In one embodiment of the present application, to obtain a real farmland point cloud, the second acquisition unit includes a third acquisition module and a second generation module.

[0139] The third acquisition module is used to acquire the real scan image acquired by the above data acquisition device;

[0140] The second generation module is used to process the above-mentioned real scanned image using Colmap sparse point cloud building technology to generate the above-mentioned real farmland point cloud.

[0141] In one embodiment of the present application, the fifth processing unit includes a first determining module and a second determining module.

[0142] The first determination module is used for determining, when the test flatness is within a preset range, to select the training data set from the test farmland terrain corresponding to the test flatness;

[0143] The second determination module is used for determining not to select the training data set from the test farmland terrain corresponding to the test flatness when the test flatness is not within the preset range.

[0144] In one embodiment of the present application, the third processing unit includes a third processing module and a fourth processing module.

[0145] The third processing module is used to optimize the result obtained by the Poisson surface reconstruction process by using Open3D after texture rendering is performed on the result obtained by the Poisson surface reconstruction process, so as to smooth the mesh curvature of the result obtained by the Poisson surface reconstruction process, and bridge the cross section of the result obtained by the Poisson surface reconstruction process to obtain an intermediate farmland simulation surface;

[0146] The fourth processing module is used to perform texture rendering processing on the intermediate farmland simulation surface to obtain the final farmland simulation terrain.

[0147] The device for generating farmland simulation terrain includes a processor and a memory. The first acquisition unit, the first processing unit, the second processing unit, and the third processing unit are all stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions. The modules are all located in the same processor; or, the modules are located in different processors in any combination.

[0148] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the kernel parameters are adjusted to solve the problem of low construction efficiency and high time cost in the construction of farmland simulation terrain in the existing solution.

[0149] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0150] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for generating the farmland simulation terrain.

[0151] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the program executes the method for generating the farmland simulation terrain when running.

[0152] An embodiment of the present invention provides a device, the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and the processor implements at least the following steps when executing the program: obtaining current farmland data, the current farmland data including current terrain length, current terrain width, and current terrain flatness; using a neural network model technology to process the current terrain flatness to obtain an optimal height and an optimal number of point clouds; using a random point cloud generation technology to generate a current random point cloud based on the optimal height, the optimal number of point clouds, the current terrain length, and the current terrain width; sequentially performing Poisson surface reconstruction and texture rendering on the current random point cloud to obtain a final farmland simulation terrain, and importing the final farmland simulation terrain into a simulation platform to simulate a preset scene. The device in this article may be a server, a PC, a PAD, a mobile phone, etc.

[0153] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes at least the following method steps: obtaining current farmland data, the current farmland data including current terrain length, current terrain width and current terrain flatness; using a neural network model technology to process the current terrain flatness to obtain an optimal height and an optimal number of point clouds; using a random point cloud generation technology to generate a current random point cloud based on the optimal height, the optimal number of point clouds, the current terrain length and the current terrain width; performing Poisson surface reconstruction processing and texture rendering processing on the current random point cloud in turn to obtain a final farmland simulation terrain, and importing the final farmland simulation terrain into a simulation platform to simulate a preset scene.

[0154] The present application also provides a system for generating farmland simulation terrain, the system comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing any of the above-mentioned methods for generating farmland simulation terrain. By performing Poisson surface reconstruction processing on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively, multiple test farmland terrains and multiple corresponding test flatnesses are obtained, so that the construction of the farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, random point cloud generation technology is used to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the above-mentioned current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform for simulation of the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem of low construction efficiency and high time cost in the process of constructing the farmland simulation terrain in the existing scheme.

[0155] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0156] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0157] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0158] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0159] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0160] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0161] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0162] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0163] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0164] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0165] 1) The method for generating farmland simulation terrain of the present application performs Poisson surface reconstruction processing on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses, so that the construction of farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is adopted to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the Poisson surface reconstruction processing and texture rendering processing are performed on the above-mentioned current random point cloud in turn to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform to simulate the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem that the existing scheme has low construction efficiency and high time cost in the process of constructing farmland simulation terrain.

[0166] 2) The device for generating farmland simulation terrain of the present application performs Poisson surface reconstruction processing on the above-mentioned real farmland point cloud and all the above-mentioned test random point clouds respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses, so that the construction of farmland terrain has the advantages of high ease of use, no need for modeling experience, low cost of use, high efficiency and reduced time cost. Then, the random point cloud generation technology is adopted to generate the current random point cloud based on the above-mentioned optimal height, the above-mentioned optimal number of point clouds, the above-mentioned current terrain length and the above-mentioned current terrain width, and the Poisson surface reconstruction processing and texture rendering processing are performed on the above-mentioned current random point cloud in turn to obtain the final farmland simulation terrain, and the above-mentioned final farmland simulation terrain is imported into the simulation platform to simulate the preset scene, so that the present application can have the advantages of low time cost and high efficiency compared with the existing scheme, thereby solving the problem that the existing scheme has low construction efficiency and high time cost in the process of constructing farmland simulation terrain.

[0167] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for generating farmland simulation terrain, characterized in that: include: Acquire current farmland data, wherein the current farmland data includes current terrain length, current terrain width and current terrain flatness; The current terrain flatness is processed by using a neural network model technology to obtain an optimal height and an optimal number of clouds; Using random point cloud generation technology, based on the optimal height, the optimal point cloud quantity, the current terrain length and the current terrain width, a current random point cloud is generated; The current random point cloud is sequentially subjected to Poisson surface reconstruction processing and texture rendering processing to obtain a final farmland simulation terrain, and the final farmland simulation terrain is imported into a simulation platform to simulate a preset scene.

2. The method according to claim 1, characterized in that The current terrain flatness is processed using a neural network model technology to obtain the optimal height and optimal number of clouds, including: Using the current terrain flatness as an input of an optimal neural network model so that the optimal neural network model processes the current terrain flatness; Obtaining an output of the optimal neural network model; The output of the optimal neural network model is determined as the optimal height and the optimal number of point clouds.

3. The method according to claim 1, characterized in that Before using the neural network model technology to process the current terrain flatness to obtain the optimal height and the optimal number of clouds, the method further includes: Acquire multiple test random point clouds and real farmland point clouds, wherein the real farmland point clouds are point clouds obtained by scanning the real farmland with a data acquisition device; Perform a Poisson surface reconstruction process on the real farmland point cloud and all the test random point clouds respectively to obtain multiple test farmland terrains and multiple corresponding test flatnesses; According to the size of the test flatness, a training data set is screened out from all the test farmland terrains, the training data set includes multiple groups of training data, each group of the training data includes a test height, a number of test point clouds and the corresponding test flatness; The initial neural network model is iteratively trained using the training data set, and the current neural network model is determined to be the optimal neural network model at least when the number of iterations reaches a preset maximum number of iterations.

4. The method according to claim 3, characterized in that Get multiple test random point clouds, including: Acquire multiple groups of test data, each group of test data includes an initial test farmland length, an initial test farmland width, an initial test height, and an initial test point cloud quantity; The random point cloud generation technology is used to process the multiple groups of test data respectively to generate the multiple test random point clouds.

5. The method according to claim 3, characterized in that: Get real farmland point cloud, including: Acquire a real scan image acquired by the data acquisition device; The real scanned image is processed by using Colmap sparse point cloud building technology to generate the real farmland point cloud.

6. The method according to claim 3, characterized in that: According to the size of the test flatness, a training data set is selected from all the test farmland terrains, including: When the test flatness is within a preset range, determining to select the training data set from the test farmland terrain corresponding to the test flatness; When the test flatness is not within the preset range, it is determined not to filter the training data set from the test farmland terrain corresponding to the test flatness.

7. The method according to claim 3, characterized in that The current random point cloud is sequentially subjected to Poisson surface reconstruction and texture rendering to obtain the final farmland simulation terrain, including: After texture rendering is performed on the result obtained by the Poisson surface reconstruction process, the result obtained by the Poisson surface reconstruction process is optimized by using Open3D to smooth the mesh curvature of the result obtained by the Poisson surface reconstruction process, and bridge the cross section of the result obtained by the Poisson surface reconstruction process to obtain an intermediate farmland simulation surface; The intermediate farmland simulation surface is texture rendered to obtain the final farmland simulation terrain.

8. A device for generating simulated farmland terrain, characterized in that: include: A first acquisition unit, used to acquire current farmland data, wherein the current farmland data includes current terrain length, current terrain width and current terrain flatness; A first processing unit is used to process the current terrain flatness using a neural network model technology to obtain an optimal height and an optimal number of clouds; A second processing unit is used to generate a current random point cloud based on the optimal height, the optimal point cloud quantity, the current terrain length and the current terrain width by using a random point cloud generation technology; The third processing unit is used to perform Poisson surface reconstruction and texture rendering on the current random point cloud in sequence to obtain a final farmland simulation terrain, and import the final farmland simulation terrain into a simulation platform to simulate a preset scene.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for generating farmland simulation terrain according to any one of claims 1 to 7.

10. A system for generating farmland simulation terrain, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the generation method of the farmland simulation terrain described in any one of claims 1 to 7.

Images