Scene generation method and device, computer device, and storage medium

Abstract

The application relates to a scene generation method and device and a storage medium. The method comprises the following steps: acquiring multiple density maps, and determining at least one target pixel point set in each density map; generating a pretreatment model corresponding to each target pixel point set according to a gray value set corresponding to each target pixel point set; dividing the multiple pretreatment models according to the position coordinates of the target pixel point sets in the corresponding density maps to obtain at least one pretreatment model set; acquiring floor information of each density map respectively, and determining the floors corresponding to the pretreatment models according to the floor information; superimposing the pretreatment models in each pretreatment model set according to the floors to obtain multiple target models, and generating a target scene by comprehensively combining the multiple target models. The method can improve the efficiency of target scene generation.

Description

Technical Field

[0001] This application relates to the field of autonomous driving simulation technology, and in particular to a scene generation method, apparatus, computer equipment, and storage medium. Background Technology

[0002] In the field of autonomous driving simulation technology, constructing virtual scenarios for road testing can ensure the correctness of autonomous driving algorithms and help identify problems in the testing process. Therefore, constructing a relatively realistic virtual scenario is a crucial step in conducting road tests.

[0003] Currently, existing construction methods typically employ pre-collected map and point cloud data, and use sensors generated through physical simulation to perform environmental perception, thereby obtaining the raw scene required for road testing. However, when the perceived environment is a city-level scene, perceiving each data point individually significantly reduces the efficiency of scene generation. Therefore, how to efficiently generate city-level scenes without relying on pre-collected data is the problem this application aims to solve. Summary of the Invention

[0004] Therefore, it is necessary to provide a scene generation method, apparatus, computer equipment, storage medium, and computer program that can improve the construction efficiency of the target scene in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a scene generation method. The method includes:

[0006] Multiple density maps are acquired, and at least one set of target pixels is determined in each density map; the set of target pixels includes pixels in the density maps whose grayscale values ​​have been adjusted.

[0007] Based on the grayscale value set corresponding to each of the target pixel sets, a preprocessing model corresponding to each of the target pixel sets is generated;

[0008] Based on the position coordinates of each target pixel set in the corresponding density map, the multiple preprocessing models are divided to obtain at least one preprocessing model set;

[0009] The floor information of each density map is obtained, and the floor corresponding to each preprocessed model is determined according to the floor information.

[0010] According to the floor, the preprocessed models in each of the preprocessed model sets are superimposed to obtain multiple target models, and the multiple target models are combined to generate a target scene.

[0011] In one embodiment, before acquiring multiple density maps, the method further includes: generating a simplified semantic map based on preset semantic information, and performing grayscale conversion processing on the simplified semantic map to obtain an initial density map; determining selected candidate pixels from the initial density map; acquiring multiple floor information, and adjusting the grayscale values ​​of the candidate pixels according to the floor information to obtain a density map corresponding to each floor information.

[0012] In one embodiment, the grayscale value set includes the grayscale value corresponding to each pixel in the target pixel set; generating a preprocessing model corresponding to each target pixel set based on the grayscale value set corresponding to each target pixel set includes: for each target pixel set in the plurality of target pixel sets, obtaining the initial height of the density map in which the current target pixel set is located; the initial height is obtained by weighted averaging of the grayscale values ​​of the pixels in the density map; determining the candidate model height corresponding to each pixel based on the correspondence between grayscale values ​​and model heights, and the grayscale value corresponding to each pixel; determining the plurality of target model heights corresponding to the current target pixel set based on the difference between the initial height and each candidate model height; determining the set range size of the current target pixel set, and generating the preprocessing model corresponding to the current target pixel set based on the plurality of target model heights and the set range size.

[0013] In one embodiment, the step of superimposing the preprocessed models in each of the preprocessed model sets according to the floor to obtain multiple target models includes: for each of the multiple preprocessed model sets, superimposing the multiple preprocessed models in the current preprocessed model set according to the floor to obtain candidate preprocessed models; determining the edge pixels in the target pixel set corresponding to each preprocessed model; updating the contour of the candidate preprocessed models based on the multiple edge pixels to obtain a target preprocessed model; and searching for a target model corresponding to the target preprocessed model from a preset database.

[0014] In one embodiment, updating the contour of the candidate preprocessing model based on a plurality of edge pixels to obtain a target preprocessing model includes: determining a current edge pixel among a plurality of edge pixels, and a set of current target pixels where the current edge pixel is located; determining a set of adjacent target pixels that have the same position coordinates as the current target pixel set and are located in adjacent density maps; selecting a plurality of adjacent edge pixels adjacent to the current edge pixel from the set of adjacent target pixels; and updating the contour of the candidate preprocessing model based on the current edge pixel and the plurality of adjacent edge pixels to obtain the target preprocessing model.

[0015] In one embodiment, updating the contour of the candidate preprocessing model based on the current edge pixel and the plurality of adjacent edge pixels to obtain a target preprocessing model includes: selecting a model contour region including the current edge pixel and the plurality of adjacent edge pixels from the candidate preprocessing models; determining the pixel gradient between the current edge pixel and each of the adjacent edge pixels; inputting each pixel gradient into a preset contour transformation model to obtain the model contour corresponding to the current edge pixel; and updating the model contour region using the model contour to obtain the target preprocessing model.

[0016] In one embodiment, generating a target scene by integrating multiple target models includes: acquiring a simplified semantic map and a target sensor; performing model transformation on the target models to obtain a transformed model; determining the underlying density map in the multiple density maps, and the target position coordinates of the target pixel set in the underlying density map; placing the transformed model at the target position indicated by the target position coordinates to obtain target scene data to be perceived; and performing environmental perception on the target scene data through the target sensor to obtain the target scene.

[0017] In one embodiment, the method further includes: acquiring original scene data and an initial sensor, and determining the scene ground truth in the target scene; determining a loss function corresponding to the initial sensor based on the difference between the original scene data and the scene ground truth; and calibrating and updating the model parameters in the initial sensor using the loss function and based on a backpropagation algorithm until the training stopping condition is met, thereby obtaining a trained target sensor.

[0018] In one embodiment, after generating a target scene by integrating multiple target models, the method further includes: determining target scene data for the target scene and obtaining a grayscale mask; the grayscale mask is a mask obtained by adjusting the grayscale values ​​of some pixels in an initial grayscale mask; determining an additional set of pixels in the grayscale mask; the additional set of pixels includes at least one pixel in the grayscale mask whose grayscale value has been adjusted; determining the target grayscale value corresponding to the additional set of pixels, and determining the additional model height of the additional set of pixels based on the target grayscale value, the correspondence between grayscale value and model height; generating an additional model corresponding to the additional set of pixels based on the additional model height; and modifying and expanding the target scene data using the additional model to obtain a modified and expanded target scene.

[0019] Secondly, this application also provides a scene generation apparatus. The apparatus includes:

[0020] A density map acquisition module is used to acquire multiple density maps and determine at least one set of target pixels in each density map; the set of target pixels includes pixels in the density map whose grayscale values ​​have been adjusted.

[0021] The model generation module is used to generate a preprocessing model corresponding to each target pixel set based on the gray value set corresponding to each target pixel set; and to divide the multiple preprocessing models according to the position coordinates of each target pixel set in the corresponding density map to obtain at least one preprocessing model set.

[0022] The model overlay module is used to acquire the floor information of each density map, and determine the floor corresponding to each preprocessed model according to the floor information; according to the floor, the preprocessed models in each preprocessed model set are overlaid to obtain multiple target models, and the multiple target models are combined to generate a target scene.

[0023] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0024] Multiple density maps are acquired, and at least one set of target pixels is determined in each density map; the set of target pixels includes pixels in the density maps whose grayscale values ​​have been adjusted.

[0025] Based on the grayscale value set corresponding to each of the target pixel sets, a preprocessing model corresponding to each of the target pixel sets is generated;

[0026] Based on the position coordinates of each target pixel set in the corresponding density map, the multiple preprocessing models are divided to obtain at least one preprocessing model set;

[0027] The floor information of each density map is obtained, and the floor corresponding to each preprocessed model is determined according to the floor information.

[0028] According to the floor, the preprocessed models in each of the preprocessed model sets are superimposed to obtain multiple target models, and the multiple target models are combined to generate a target scene.

[0029] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, performs the following steps:

[0030] Multiple density maps are acquired, and at least one set of target pixels is determined in each density map; the set of target pixels includes pixels in the density maps whose grayscale values ​​have been adjusted.

[0031] Based on the grayscale value set corresponding to each of the target pixel sets, a preprocessing model corresponding to each of the target pixel sets is generated;

[0032] Based on the position coordinates of each target pixel set in the corresponding density map, the multiple preprocessing models are divided to obtain at least one preprocessing model set;

[0033] The floor information of each density map is obtained, and the floor corresponding to each preprocessed model is determined according to the floor information.

[0034] According to the floor, the preprocessed models in each of the preprocessed model sets are superimposed to obtain multiple target models, and the multiple target models are combined to generate a target scene.

[0035] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0036] Multiple density maps are acquired, and at least one set of target pixels is determined in each density map; the set of target pixels includes pixels in the density maps whose grayscale values ​​have been adjusted.

[0037] Based on the grayscale value set corresponding to each of the target pixel sets, a preprocessing model corresponding to each of the target pixel sets is generated;

[0038] Based on the position coordinates of each target pixel set in the corresponding density map, the multiple preprocessing models are divided to obtain at least one preprocessing model set;

[0039] The floor information of each density map is obtained, and the floor corresponding to each preprocessed model is determined according to the floor information.

[0040] According to the floor, the preprocessed models in each of the preprocessed model sets are superimposed to obtain multiple target models, and the multiple target models are combined to generate a target scene.

[0041] The aforementioned scene generation method, apparatus, computer equipment, storage medium, and computer program acquire multiple density maps and determine at least one set of target pixels in each density map. Then, based on the grayscale value set corresponding to each set of target pixels, a preprocessing model corresponding to each set of target pixels is generated. The multiple preprocessing models are divided using the position coordinates of each set of target pixels in the corresponding density maps to obtain at least one set of preprocessing models. The floor information of each density map is then acquired, and the corresponding floor for each preprocessing model is determined based on this information. Thus, the preprocessing models in each set of preprocessing models can be superimposed according to the floor to obtain multiple target models, until the target scene is generated by combining the multiple target models. Since this application directly generates the preprocessing model corresponding to each set of target pixels from multiple density maps, compared to traditional methods that rely on pre-collected data, this application can superimpose preprocessing models based on the corresponding floor, thereby obtaining multiple target models in large quantities, greatly improving the efficiency of target scene generation. Attached Figure Description

[0042] Figure 1 This is an application environment diagram of the scene generation method in one embodiment;

[0043] Figure 2 This is a flowchart illustrating a scene generation method in one embodiment;

[0044] Figure 3 This is a rendering of the initial density map in one embodiment;

[0045] Figure 4 This is a schematic diagram of the process for generating a preprocessed model in one embodiment;

[0046] Figure 5 This is a flowchart illustrating the process of determining the target model in one embodiment;

[0047] Figure 6 This is a structural block diagram of a scene generation device in one embodiment;

[0048] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The scene generation method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. Terminal 102 can send multiple density maps and floor information of each density map to server 104. Server 104 is used to determine at least one set of target pixels in each density map, and generate a preprocessing model corresponding to each set of target pixels based on the grayscale value set corresponding to each set of target pixels. Server 104 is also used to divide the multiple preprocessing models according to the position coordinates of each set of target pixels in the corresponding density map to obtain at least one set of preprocessing models, and determine the floor corresponding to each preprocessing model according to the floor information. Server 104 is also used to overlay the preprocessing models in each set of preprocessing models according to the floor to obtain multiple target models, and generate a target scene by combining the multiple target models, and send the target scene to terminal 102 so that terminal 102 can display the target scene. The terminal 102 may be, but is not limited to, various personal computers, laptops, smartphones, smart in-vehicle devices, tablets, IoT devices, and portable wearable devices. In one embodiment, the terminal 102 is an autonomous driving computing platform configured on an autonomous vehicle. The server 104 may be implemented using a standalone server or a server cluster consisting of multiple servers.

[0051] In one embodiment, such as Figure 2 As shown, a scene generation method is provided. Taking the application of this method to a computer device as an example, the computer device can provide... Figure 1 The terminal or server in the process includes the following steps:

[0052] Step 202: Obtain multiple density maps and determine at least one set of target pixels in each density map; the set of target pixels includes pixels in the density map whose grayscale values ​​have been adjusted.

[0053] The density map is obtained by adjusting the grayscale values ​​of some pixels in the initial density map. A density map is a grayscale image with multiple pixels, and its resolution is related to the geographic coordinate range corresponding to the model's height. For example, a density map with a resolution of 1000 dpi × 1000 dpi can represent a geographic coordinate range of 100m × 100m; therefore, the grayscale value of each pixel corresponding to 1 dpi corresponds to a geographic coordinate range of 0.1m. The target pixel set is a set of points where the actual building is placed.

[0054] Specifically, the number of target pixel sets in each density map can be different. For any current target pixel set in a density map, multiple pixels with adjustment markers and consecutive positions are selected from the pixels in the density map using computer equipment to obtain the current target pixel set. Alternatively, the center pixel with adjustment markers can be selected from the pixels, and then pixels within a preset set size located at the center pixel can be selected to obtain the current target pixel set.

[0055] In one embodiment, before acquiring multiple density maps, the method further includes: generating a simplified semantic map based on preset semantic information, and performing grayscale value conversion processing on the simplified semantic map to obtain an initial density map; determining selected candidate pixels from the initial density map; acquiring multiple floor information, and adjusting the grayscale values ​​of the candidate pixels according to the floor information to obtain a density map corresponding to each floor information.

[0056] In one embodiment, determining the selected candidate pixels from the initial density map includes: in response to a selection operation for pixels in the initial density map, determining the candidate pixels selected by the selection operation.

[0057] The simplified semantic map includes road labeling information, such as road areas, lane line types, and building areas. Preset semantic information is obtained through user-defined building placement rules or generated via an editor system. Floor information is used to determine the density map of the target model at different floor heights; for example, density... Figure 1 The corresponding floor information is the building density at a height of 0-10m. Figure 2 This corresponds to buildings at a height of 10m-20m.

[0058] Specifically, the computer device performs grayscale conversion on the simplified semantic map based on the semantic information in the simplified semantic map to obtain an initial density map. For example, building areas are converted into pixel areas with larger grayscale values, making these areas appear close to white, while road areas are converted into pixel areas with smaller grayscale values, making these areas appear close to black, and so on. Figure 3 As shown, Figure 3 This is a rendering of an initial density map based on a specific terrain region. The computer device can display a density map editing interface, where the initial density map is shown. In response to a user's selection of pixels in the initial density map within the editing interface, the computer device obtains candidate pixels. The computer device acquires multiple floor information and a preset floor ground truth value corresponding to each floor information, such as density. Figure 1 The truth value of floor 1 includes the set range size, location coordinates including (X1, Y1) and (X2, Y2), and density. Figure 2 The true value of the floor 2 includes the location coordinates (X1, Y1), etc. In response to the user's operation of adjusting the grayscale value of the candidate pixel in the editing interface based on the true value of the floor, until it is adjusted to match the true value of the floor information, the initial density map and the pixels after grayscale value adjustment are combined to obtain the density map corresponding to each floor information, and displayed in the editing interface.

[0059] In one embodiment, the computer device determines the top-level density map to be generated corresponding to each location coordinate, and gradually increases the grayscale values ​​of pixels farther away from the candidate pixels, i.e., the whiter the color. The resulting top-level density map is shown below. Here, the location coordinates are the coordinates of the center point of the target pixel set. Since the height of the target model to be generated at each location coordinate is not exactly the same, the top-level density maps corresponding to the location coordinates of each target model will not be exactly the same after the target model is subsequently generated. For example, if there is an 8m building 1 and a 12m building 2, the top-level density map of building 1 is [density map not specified]. Figure 1 The density map of the top floor of building 2 is as follows: Figure 2 .

[0060] In one embodiment, a simplified semantic map can be obtained by data collection from an autonomous vehicle equipped with LiDAR. When the collected raw data is insufficient in quantity or density, a computer device can use an orthogonal camera (i.e., without perspective) to query the simplified semantic map from a top-down view. By querying the height value of each coordinate point on the original map, the simplified semantic map can be programmed to edit, generating at least one density map. Here, "editing" broadly refers to any computer program with a visual interface and interactive methods that can load and edit resources, including various 3D model editing software, game engine software, etc.

[0061] In this embodiment, since the density map is generated only on the basis of the simplified semantic map, the generation difficulty of the density map is reduced. Furthermore, the gray value of the pixel can be flexibly adjusted according to the different floor information to ensure that each density map is the same, making the generation of the density map more flexible and reducing the computational cost of determining the height of the target model by the gray value of the pixel.

[0062] Step 204: Generate a preprocessing model for each set of target pixels based on the set of gray values ​​corresponding to each set of target pixels.

[0063] The preprocessing model provides basic information about the target model to be generated, such as its outline shape and various texture maps of the surface (color, gloss, reflectivity, normal direction, etc.). The preprocessing model usually exists in a specific file in a certain description format.

[0064] Specifically, the computer device performs grayscale value analysis on the pixels in each target pixel set to obtain the grayscale value set corresponding to the target pixel set, and obtains the correspondence between grayscale values ​​and model height. Then, based on the grayscale value of each pixel and the correspondence between grayscale values ​​and model height, the computer device determines the model height set corresponding to each target pixel set. The computer device determines the set size of each target pixel set, that is, the pixel area it occupies in the density map, and then constructs a preprocessed model corresponding to each target pixel set based on the model height set and the set size.

[0065] In one embodiment, for each pixel in the density map, its grayscale value determines the associated height information within the geographic coordinate range represented by the pixel. Users can pre-set the relationship between the upper and lower bounds of the height value and the grayscale value, that is, pre-determine prior information such as the upper and lower bound constraints of the grayscale value. For example, a grayscale value of 255 represents a model height of 0m, a grayscale value of 0 represents a model height of 255m, and for every decrease of 1 in the grayscale value, the height of the pixel increases by 1m.

[0066] Step 206: Based on the position coordinates of each target pixel set in the corresponding density map, divide the multiple preprocessing models to obtain at least one preprocessing model set.

[0067] Specifically, the computer device determines the position coordinates of the center point of each target pixel set in the corresponding density map, and uses these position coordinates as a basis to divide the preprocessing model. For example, density... Figure 1 The center point of the target pixel set 1 in the image has coordinates (X1, Y1) and density. Figure 1 The center point of the target pixel set 2 in the image has coordinates (X2, Y2) and density. Figure 2 The center point of the target pixel set 3 is located at coordinates (X1, Y1). Therefore, the density... Figure 1 Preprocessing Model 1 and density at (X1, Y1) Figure 2 Preprocessed model 3 at (X1, Y1) is grouped together to obtain a preprocessed model set a; the density is... Figure 1 The preprocessed model 2 at (X2, Y2) is used as a group to obtain another preprocessed model set b.

[0068] Step 208: Obtain the floor information for each density map and determine the corresponding floor for each preprocessing model based on the floor information.

[0069] Specifically, for each of the multiple density maps, at least one set of target pixels in the current density map is associated with the floor information of the current density map. For example, for density... Figure 1 If the floor level information is 0-10m, then the density... Figure 1 The floor information of both target pixel set 1 and target pixel set 2 in the image is associated with 0-10m, and the density is... Figure 2 If the floor level information is 10-20m, then the density... Figure 2 The floor information of target pixel set 3 is associated with 10-20m, from which the corresponding floor of each preprocessed model can be obtained. For example, the floor of preprocessed model 1 of target pixel set 1 is the first floor, the floor of preprocessed model 2 of target pixel set 2 is the first floor, and the floor of preprocessed model 3 of target pixel set 3 is the second floor.

[0070] Step 210: According to the floor, the preprocessed models in each preprocessed model set are superimposed to obtain multiple target models, and the target scene is generated by combining the multiple target models.

[0071] When a density map corresponding to each floor information is obtained, the computer equipment can also annotate the density map according to the obtained floor information to determine the stacking order of the preprocessed model body. For example, buildings usually have a structural shape that is wider at the bottom and narrower at the top. By annotating the stacking order, buildings that are narrower at the bottom and wider at the top and do not conform to physical rules can be avoided by accident.

[0072] Specifically, for each preprocessed model set, the computer device determines multiple preprocessed models within the current set and overlays them according to their floor level to form a target model. Referring to the example above, preprocessed model set a includes preprocessed model 1 and preprocessed model 3. Preprocessed model 1 is on the first floor, and preprocessed model 3 is on the second floor. Therefore, preprocessed model 3 (second floor) is overlaid onto preprocessed model 1 (first floor) to obtain the target model with position coordinates (X1, Y1). If the current preprocessed model set contains only one preprocessed model, that preprocessed model is directly used as the target model. For example, since preprocessed model set b only contains preprocessed model 2, preprocessed model 2 is directly used as the target model with position coordinates (X2, Y2). It is easy to understand that for buildings of different heights, different numbers of density maps are needed to generate the target model.

[0073] In one embodiment, when computing resources are limited, a target scene can be generated based on existing raw data using relatively little computing power. Here, computing power represents the number of data operations a computer device can perform per second. If computing resources are sufficient, more computing power can be used to build the target scene from scratch, thus enabling flexible allocation of computing power.

[0074] The aforementioned scene generation method, apparatus, computer equipment, storage medium, and computer program acquire multiple density maps and determine at least one set of target pixels in each density map. Then, based on the grayscale value set corresponding to each set of target pixels, a preprocessing model corresponding to each set of target pixels is generated. The multiple preprocessing models are divided using the position coordinates of each set of target pixels in the corresponding density maps to obtain at least one set of preprocessing models. The floor information of each density map is then acquired, and the corresponding floor for each preprocessing model is determined based on this information. Thus, the preprocessing models in each set of preprocessing models can be superimposed according to the floor to obtain multiple target models, until the target scene is generated by combining the multiple target models. Since this application directly generates the preprocessing model corresponding to each set of target pixels from multiple density maps, compared to traditional methods that rely on pre-collected data, this application can superimpose preprocessing models based on the corresponding floor, thereby obtaining multiple target models in large quantities, greatly improving the efficiency of target scene generation.

[0075] In one embodiment, such as Figure 4 As shown, Figure 4 This is a flowchart illustrating the process of generating the preprocessed model. Based on the grayscale value set corresponding to each set of target pixels, a preprocessed model is generated for each set of target pixels, including the following steps:

[0076] Step 402: For each set of target pixels in the multiple sets of target pixels, obtain the initial height of the density map in which the current set of target pixels is located.

[0077] The grayscale value set includes the grayscale value corresponding to each pixel in the target pixel set; the initial height of the density map is the ground elevation information of the current geographic area, which is obtained by weighted averaging of the grayscale values ​​of the pixels in the density map; it is easy to understand that the initial height corresponding to each density map can be different.

[0078] Step 404: Based on the correspondence between grayscale values ​​and model height, and the grayscale value corresponding to each pixel, determine the candidate model height corresponding to each pixel.

[0079] Specifically, when the computer device determines the initial height of the density map, it determines the candidate model height corresponding to each pixel in the target pixel set according to the preset correspondence between gray values ​​and model height. For example, if the gray value of pixel 1 is 245 and the gray value of pixel 2 is 244, then the candidate model height of pixel 1 is 10m and the candidate model height of pixel 2 is 11m.

[0080] Step 406: Determine the heights of multiple target models corresponding to the current set of target pixels based on the difference between the initial height and the height of each candidate model.

[0081] Since the target model to be generated will be built on the current geographic area, the computer device determines the height of multiple target models corresponding to the current target pixel set based on the difference between the initial height and the candidate model height. For example, if the initial height is 2m, the target model height of pixel 1 in the current target pixel set is 8m, and the target model height of pixel 2 is 9m.

[0082] Step 408: Determine the set range size of the current target pixel set, and generate the preprocessed model corresponding to the current target pixel set based on the height of multiple target models and the set range size.

[0083] The set range size represents the range occupied by the target pixel set in the density map, which is the actual bottom area of ​​the target model to be generated.

[0084] Specifically, the computer device determines whether the center point of the current target pixel set has a special identifier. If the center point has a special identifier, it indicates that the current target pixel set corresponds to the top layer density map, and the preprocessed model in this case may be a model with a special shape. Referring to the example above, if the pixels in the current target pixel set are close to the center point and their grayscale values ​​are relatively small, it can be determined that the heights of multiple target models form a dome shape. When the set size is a circle with a radius of 5m, the preprocessed model corresponding to the current target pixel set is a cylindrical building with a dome shape. If the center point does not have a special identifier, it is easy to understand that the preprocessed model may be a regular shape associated with the floor information of the density map. For example, if multiple target models are all 10m high, the generated preprocessed model is a regular cylindrical building with a height of 10m and a radius of 5m.

[0085] In this embodiment, by taking the initial height corresponding to the density map into the target model height, the obtained target model height can be ensured to be more accurate. Furthermore, by determining the preprocessed model generated by the top-level density map with special identification, the subsequently generated target model can better match the actual constructed virtual scene.

[0086] In one embodiment, such as Figure 5 As shown, Figure 5 To determine the flowchart of the target model, the preprocessed models in each preprocessed model set are superimposed according to the floor, resulting in multiple target models, including the following steps:

[0087] Step 502: For each of the multiple preprocessing model sets, according to the floor, the multiple preprocessing models in the current preprocessing model set are superimposed to obtain the candidate preprocessing model.

[0088] The candidate preprocessing model is typically a simple model with a conventional shape, which does not reflect the detailed contour information of the target model to be generated. The detailed contour information can be a contour protrusion or a contour depression. For example, referring to step 408, if multiple sub-cylinders are superimposed, the resulting candidate preprocessing model will be a cylinder that does not include detailed contour information.

[0089] Step 504: Determine the edge pixels in the target pixel set corresponding to each preprocessing model.

[0090] Specifically, when the set of target pixels is circular, the edge pixels are the pixels at the inner edge of the circle; when the set of target pixels is rectangular, the edge pixels are the pixels at the inner edge of the rectangle.

[0091] Step 506: Update the contour of the candidate preprocessing model based on multiple edge pixels to obtain the target preprocessing model.

[0092] In one embodiment, the candidate preprocessing model is updated based on multiple edge pixels to obtain the target preprocessing model, including: determining the current edge pixel among multiple edge pixels and the current target pixel set where the current edge pixel is located; determining the set of adjacent target pixels that have the same position coordinates as the current target pixel set and are located in adjacent density maps; and selecting multiple adjacent edge pixels that are adjacent to the current edge pixel from the set of adjacent target pixels.

[0093] For example, the current target pixel set where the current edge pixel m is located is target pixel set 1. Since target pixel set 1 has a density... Figure 1 Given that the position coordinates are (X1, Y1), the set of adjacent target pixels is the density. Figure 2 The target pixel set 3 is a set of pixels with position coordinates (X1, Y1). Thus, we can select the adjacent edge pixels k1 and k2 that are adjacent to the current edge pixel n from the target pixel set 3.

[0094] Further, based on the current edge pixel and its multiple adjacent edge pixels, the candidate preprocessing model is updated to obtain the target preprocessing model. This includes: selecting a model contour region from the candidate preprocessing models that includes the current edge pixel and multiple adjacent edge pixels; determining the pixel gradient between the current edge pixel and each adjacent edge pixel; inputting each pixel gradient into a preset contour transformation model to obtain the model contour corresponding to the current edge pixel; and updating the model contour region using the model contour to obtain the target preprocessing model.

[0095] The contour transformation model can be trained using deep learning. By learning the shape ratio of buildings at different heights in the real world, it can simulate the shape difference of the target model at different heights, thus generating an irregular model contour. The pixel gradient represents the difference between pixel values. The larger the pixel gradient, the greater the difference between any two pixel values. The pixel gradient can also be the density gradient of a density map.

[0096] Specifically, the computer device selects the model contour region from the candidate preprocessing models. This model contour region is equivalent to the side region at the overlap of two sub-cylinders, associated with the current edge pixel and multiple adjacent edge pixels. This side region does not include detailed contour information. Based on the difference between the pixel value of the current edge pixel and the corresponding pixel values ​​of each adjacent edge pixel, the computer device determines the pixel gradient between the current edge pixel and each adjacent edge pixel. For example, the current edge pixel n and adjacent edge pixel k1 can be determined as pixel gradient 1, and adjacent edge pixel k2 as pixel gradient 2. The computer device inputs pixel gradient 1 and pixel gradient 2 into a preset contour transformation model, resulting in a model contour with a contour bulge of 0.5m between the current edge pixel n and adjacent edge pixel k1, and a model contour with a contour bulge of 1m between the current edge pixel n and adjacent edge pixel k2. The model contour region is then updated using the model contour to obtain a target preprocessing model with detailed contour information.

[0097] Step 508: Search for the target model corresponding to the target preprocessing model from the preset database.

[0098] In this embodiment, by determining the edge pixels in the target pixel set, the model contour region is updated by the model contour corresponding to the current edge pixel. In this way, the target preprocessing model can be quickly determined, so that when building a city-level scene, a target model with detailed contour information can still be generated, thus ensuring the efficiency of building the target scene.

[0099] In one embodiment, generating a target scene by integrating multiple target models includes: acquiring a simplified semantic map and a target sensor; performing model transformation on the target models to obtain a transformed model; determining the underlying density map in multiple density maps and the target position coordinates of the target pixel set in the underlying density map; placing the transformed model at the target position indicated by the target position coordinates to obtain target scene data to be perceived; and performing environmental perception on the target scene data through the target sensor to obtain the target scene.

[0100] Among them, the target sensor is used to perceive the environment of the data to be generated, so as to obtain the simulation scene required for autonomous driving; the bottom density map is determined by floor information, for example, the floor information is 0-10m height as the bottom density map.

[0101] Specifically, since the target model is generated based on multiple density maps, scale changes may occur during the editing process of generating at least one density map based on a simplified semantic map. Therefore, the computer device needs to perform model transformation on the target model, such as volume transformation, color transformation, or texture transformation, to make the surface of the target model more realistic, resulting in a transformed model. The computer device determines the target position coordinates of the center point of the target pixel set in the underlying density map, and determines the target position in the simplified semantic map based on the target position coordinates. Then, the transformed model is placed at the target position to obtain the target scene data to be perceived. The computer device uses a target sensor to perform environmental perception on the target scene data, obtaining a target scene that can be used for road testing.

[0102] In this embodiment, by determining the target position coordinates of the center point of the target pixel set in the underlying density map, the target model is transformed into a transformation model that can be placed in a simplified semantic map, thus realizing the fusion of the target model and the simplified semantic map and ensuring the accuracy of the target scene generation.

[0103] In one embodiment, the method further includes: acquiring original scene data and an initial sensor, and determining the scene ground truth in the target scene; determining the loss function corresponding to the initial sensor based on the difference between the original scene data and the scene ground truth; calibrating and updating the model parameters in the initial sensor using the loss function and based on the backpropagation algorithm until the training stopping condition is met, thereby obtaining a trained target sensor.

[0104] The ground truth of the scene includes the semantic information of all models in the target scene, such as coordinates, volume, and normal vector of a point on the surface; the loss function can be a hinge loss function, exponential loss function, etc.; the backpropagation algorithm is an algorithm that adjusts the weight parameters of the network model based on the error between the true value and the estimated value. For example, the backpropagation algorithm can be a gradient descent algorithm. In the gradient descent algorithm, the sum of squares of the error can be used as the objective function. By calculating the partial derivative of the objective function with respect to the weights of each neuron layer by layer, the gradient of the objective function with respect to the weight parameters is formed. Thus, the model parameters are modified according to the gradient descent direction. The training of the network model is completed in the process of modifying the model parameters.

[0105] Specifically, when the true value of the scene in the target scene is obtained, the computer device determines the loss function corresponding to the initial sensor based on the difference between the original scene data and the true value of the scene. Then, the gradient of the loss value of the loss function is determined through the backpropagation algorithm, and the model parameters in the initial sensor are updated along the gradient direction of the loss value. That is, the loss value is minimized as much as possible by using the model parameter decay method, so that the loss function converges and the trained target sensor is obtained.

[0106] In this embodiment, the initial sensor is updated and trained using the scene ground truth provided by the target scene, thereby improving the overall perception performance of the target sensor, calibrating the initial sensor, and ensuring that a perception scene that is more in line with the autonomous driving environment is obtained.

[0107] In one embodiment, after generating a target scene by integrating multiple target models, the method further includes: determining target scene data for the target scene and obtaining an additional density map; the additional density map is a density map obtained after converting the target scene data to grayscale values; determining an additional set of pixels in the additional density map; determining the additional grayscale values ​​corresponding to the additional set of pixels, and determining the additional model height of the additional set of pixels based on the additional grayscale values ​​and the correspondence between grayscale values ​​and model height; generating an additional model corresponding to the additional set of pixels based on the additional model height; and modifying and expanding the target scene data using the additional model to obtain the modified and expanded target scene.

[0108] For example, when the simplified semantic map represents a school's campus A, the above embodiment can quickly construct a large batch of all target models to be generated in campus A based on the simplified semantic map, thus obtaining the target scene campus A. However, for situations where it is necessary to modify or add some models in the target scene campus A, the method in this embodiment can be used to modify and expand the target scene campus A by adding models, resulting in the modified and expanded campus A.

[0109] The computer device can pre-extract model data corresponding to the target model from the target scene data, or extract background data from the target scene, and then perform grayscale conversion processing on the model data or background data to obtain an initial additional density map. The computer device then adjusts the pixels in the initial additional density map to obtain an additional density map corresponding to the additional model to be generated. The additional pixel set includes at least one pixel in the additional density map that has undergone grayscale adjustment. The additional pixel set represents a set of points in the additional density map where buildings are actually placed, referring to the target pixel set in the above embodiment.

[0110] Specifically, the specific implementation process of the computer device determining the height of the additional model for the additional pixel set can be referred to the specific implementation process of step 406 above, which will not be repeated here. When the target scene data includes a target model corresponding to the additional pixel set, the additional model height is the height corresponding to the center point of the additional pixel set. At this time, the computer device extracts the model data corresponding to the target model from the target scene data, and then constructs the additional model corresponding to the additional pixel set through the additional model height and the model data. For example, for target model 1 with position coordinates (X1, Y1) in the target scene, if the model height of target model 1 generated in the above embodiment is 12m, an additional model of 15m can be generated through the additional model height. By first deleting the original target model in the target scene, and then expanding the additional model to the position of the deleted target model, the modified and expanded target scene is obtained.

[0111] When the target scene data does not include a target model corresponding to the additional pixel set, the additional model height includes the height of each pixel in the additional pixel set. In this case, the computer determines a preset horizontal dimension associated with the additional pixel set. Based on the preset horizontal dimension and the additional model height, the additional model corresponding to the additional pixel set can be generated. The preset horizontal dimension indicates the area of ​​the pixel region occupied in the additional density map, centered on the two-dimensional coordinates of the center point of the additional pixel set. For example, if a target model exists at position coordinates (X5, Y5) in the target scene, the additional model is directly expanded to (X5, Y5) to obtain the modified and expanded target scene.

[0112] This embodiment solves the problems of needing to add data to the target scene and expand parts of the model, greatly improving the flexibility of modifying and expanding the target scene. At the same time, since it is based on the target scene data, the accuracy of modifying and expanding the target scene is ensured.

[0113] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0114] Based on the same inventive concept, this application also provides a scene generation apparatus for implementing the scene generation method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more scene generation apparatus embodiments provided below can be found in the limitations of the scene generation method described above, and will not be repeated here.

[0115] In one embodiment, such as Figure 6 As shown, a scene generation device 600 is provided, including: a density map acquisition module 602, a model generation module 604, and a model overlay module 606, wherein:

[0116] The density map acquisition module 602 is used to acquire multiple density maps and determine at least one set of target pixels in each density map; the set of target pixels includes pixels in the density map whose grayscale values ​​have been adjusted.

[0117] The model generation module 604 is used to generate a preprocessing model corresponding to each target pixel set based on the gray value set corresponding to each target pixel set; and to divide multiple preprocessing models according to the position coordinates of the target pixel set in the corresponding density map to obtain at least one preprocessing model set.

[0118] The model overlay module 606 is used to acquire the floor information of each density map and determine the corresponding floor of each preprocessed model based on the floor information; the preprocessed models in each preprocessed model set are overlaid according to the floor to obtain multiple target models, and the target scene is generated by combining the multiple target models.

[0119] In one embodiment, the scene generation device 600 further includes a grayscale adjustment module 608, which is used to generate a simplified semantic map based on preset semantic information, and perform grayscale conversion processing on the simplified semantic map to obtain an initial density map; determine the selected candidate pixels from the initial density map; acquire multiple floor information, and adjust the grayscale values ​​of the candidate pixels according to the floor information to obtain a density map corresponding to each floor information.

[0120] In one embodiment, the model generation module 604 is further configured to: obtain the initial height of the density map in which the current target pixel set is located for each of the multiple target pixel sets; the initial height is obtained by weighted averaging of the gray values ​​of the pixels in the density map; determine the candidate model height corresponding to each pixel based on the correspondence between gray values ​​and model height, and the gray value corresponding to each pixel; determine the multiple target model heights corresponding to the current target pixel set based on the difference between the initial height and each candidate model height; determine the set range size of the current target pixel set; and generate a preprocessed model corresponding to the current target pixel set based on the multiple target model heights and the set range size.

[0121] In one embodiment, the model overlay module 606 further includes a contour update module 6061, which is used to overlay multiple preprocessed models in the current preprocessed model set according to floor level for each of the multiple preprocessed model sets to obtain a candidate preprocessed model; determine the edge pixels in the target pixel set corresponding to each preprocessed model; update the contour of the candidate preprocessed model according to the multiple edge pixels to obtain a target preprocessed model; and search for the target model corresponding to the target preprocessed model from a preset database.

[0122] In one embodiment, the contour update module 6061 is further configured to: determine the current edge pixel among multiple edge pixels and the current target pixel set in which the current edge pixel is located; determine the set of adjacent target pixels that have the same position coordinates as the current target pixel set and are located in adjacent density maps; select multiple adjacent edge pixels that are adjacent to the current edge pixel from the set of adjacent target pixels; and update the contour of the candidate preprocessing model based on the current edge pixel and the multiple adjacent edge pixels to obtain the target preprocessing model.

[0123] In one embodiment, the contour update module 6061 is further configured to: select a model contour region that includes the current edge pixel and multiple adjacent edge pixels from the candidate preprocessing models; determine the pixel gradient between the current edge pixel and each adjacent edge pixel; input each pixel gradient into a preset contour transformation model to obtain the model contour corresponding to the current edge pixel; and update the model contour region through the model contour to obtain the target preprocessing model.

[0124] In one embodiment, the model overlay module 606 further includes a model placement module 6062, which is used to acquire a simplified semantic map and a target sensor; perform model conversion on the target model to obtain a converted model; determine the underlying density map in multiple density maps and the target position coordinates of the target pixel set in the underlying density map; place the converted model at the target position indicated by the target position coordinates to obtain the target scene data to be perceived; and perform environmental perception on the target scene data through the target sensor to obtain the target scene.

[0125] In one embodiment, the scene generation device 600 further includes a calibration module 610, which is used to acquire original scene data and an initial sensor, and determine the scene ground truth in the target scene; determine the loss function corresponding to the initial sensor based on the difference between the original scene data and the scene ground truth; and calibrate and update the model parameters in the initial sensor using the loss function and based on the backpropagation algorithm until the training stopping condition is met, thereby obtaining a trained target sensor.

[0126] Each module in the aforementioned scene generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0127] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown, this computer device includes a processor, memory, input / output interfaces (I / O), a communication interface, a display unit, and input devices. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface, display unit, and input devices are also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a three-dimensional modeling method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0128] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0129] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0130] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0131] In one embodiment, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. The camera's processor reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the camera to perform the steps described in the method embodiments above.

[0132] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. 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), magnetic 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 take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0133] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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, they should be considered to be within the scope of this specification.

[0134] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A scene generation method characterized by, The method comprises: obtaining a plurality of density maps, and determining at least one target pixel point set in each of the density maps; the target pixel point set comprises pixel points in the density map that have undergone gray value adjustment; wherein the density map is determined based on a preset simple semantic map and floor information, and each of the density maps corresponds to each of the floor information; for each of the plurality of target pixel point sets, an initial height of the density map in which the current target pixel point set is located is obtained; the initial height is obtained by weighted average calculation on the gray values of the pixel points in the density map; determining a candidate model height corresponding to each pixel point according to the corresponding relationship between the gray value and the model height, and the respective corresponding gray value of each pixel point; determining a plurality of target model heights corresponding to the current target pixel point set according to the difference between the initial height and each of the candidate model heights; determining the size of the set range of the current target pixel point set, and generating a pre-processing model corresponding to the current target pixel point set according to the plurality of target model heights and the size of the set range; dividing a plurality of pre-processing models according to the position coordinates of each of the target pixel point sets in the corresponding density map to obtain at least one pre-processing model set; obtaining floor information of each of the density maps, and determining a floor corresponding to each of the pre-processing models according to each of the floor information; stacking the pre-processing models in each of the pre-processing model sets according to the floor to obtain a plurality of target models, and generating a target scene by integrating the plurality of target models.

2. The method of claim 1, wherein, Before the plurality of density maps are obtained, the method further comprises: generating a simple semantic map according to preset semantic information, and performing gray value conversion processing on the simple semantic map to obtain an initial density map; determining selected candidate pixel points from the initial density map; obtaining a plurality of floor information, and adjusting the gray values of the candidate pixel points according to the floor information to obtain a density map corresponding to each of the floor information.

3. The method of claim 1, wherein, The stacking of the pre-processing models in each of the pre-processing model sets according to the floor to obtain a plurality of target models comprises: for each of the plurality of pre-processing model sets, stacking a plurality of pre-processing models in the current pre-processing model set according to the floor to obtain a candidate pre-processing model; determining an edge pixel point in each of the target pixel point sets corresponding to each of the pre-processing models; updating the contour of the candidate pre-processing model according to a plurality of edge pixel points to obtain a target pre-processing model; finding a target model corresponding to the target pre-processing model from a preset database.

4. The method of claim 3, wherein, The updating of the contour of the candidate pre-processing model according to a plurality of edge pixel points to obtain a target pre-processing model comprises: determining a current edge pixel point in a plurality of edge pixel points, and a current target pixel point set in which the current edge pixel point is located; determine a neighboring target pixel point set which has the same position coordinate as the current target pixel point set and is located in a neighboring density map; select a plurality of neighboring edge pixel points adjacent to the current edge pixel point from the neighboring target pixel point set; update the candidate pre-processing model according to the current edge pixel point and the plurality of neighboring edge pixel points to obtain a target pre-processing model.

5. The method of claim 4, wherein, The method further comprises: obtain original scene data and an initial sensor, and determine a scene ground truth in the target scene; determine a loss function corresponding to the initial sensor according to a difference between the original scene data and the scene ground truth; update model parameters in the initial sensor based on a back propagation algorithm through the loss function until a training stop condition is reached to stop, thereby obtaining a trained target sensor. After the plurality of target models are integrated to generate a target scene, the method further comprises:

6. The method of claim 1, wherein, determine target scene data of the target scene, and obtain an additional density map; the additional density map is a density map obtained by performing gray value conversion processing on the target scene data; determine an additional pixel point set in the additional density map; the additional pixel point set includes at least one pixel point in the additional density map that has undergone gray value adjustment; determine an additional gray value corresponding to the additional pixel point set, and determine an additional model height of the additional pixel point set according to a corresponding relationship between the additional gray value, a gray value, and a model height; generate an additional model corresponding to the additional pixel point set according to the additional model height; perform modification and expansion processing on the target scene data through the additional model to obtain a modified and expanded target scene. The device comprises:

7. The method of claim 1, wherein, ​ ​ ​ ​ 8. The method of claim 1, wherein, ​ ​ ​ ​ ​ ​ 9. A scene generation apparatus, characterized by comprising: ​ The density map acquisition module is configured to acquire a plurality of density maps and determine at least one target pixel point set in each of the density maps; the target pixel point set includes pixel points in the density map that have undergone gray value adjustment; wherein the density map is determined based on a preset simple semantic map and floor information, and each of the density maps corresponds to each of the floor information; The model generation module is configured to, for each of a plurality of target pixel point sets, acquire an initial height of the density map in which the target pixel point set is located; the initial height is obtained by weighted average calculation on the gray values of the pixel points in the density map; determine a candidate model height corresponding to each pixel point according to the corresponding relationship between the gray value and the model height and the gray value corresponding to each pixel point; determine a plurality of target model heights corresponding to the target pixel point set according to the difference between the initial height and each of the candidate model heights; determine the range size of the target pixel point set, and generate a pre-processing model corresponding to the target pixel point set according to the plurality of target model heights and the range size of the target pixel point set; The model superposition module is configured to acquire floor information of each of the density maps respectively, and determine a floor corresponding to each of the pre-processing models according to each of the floor information; superimpose the pre-processing models in each of the pre-processing model sets according to the floor to obtain a plurality of target models, and generate a target scene by integrating the plurality of target models.

10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

11. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8.

12. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8.

Images