WO2023051671A1 - 基于体素数据的数据处理方法、服务器、介质及计算机程序产品 - Google Patents

基于体素数据的数据处理方法、服务器、介质及计算机程序产品 Download PDF

Info

Publication number
WO2023051671A1
WO2023051671A1 PCT/CN2022/122494 CN2022122494W WO2023051671A1 WO 2023051671 A1 WO2023051671 A1 WO 2023051671A1 CN 2022122494 W CN2022122494 W CN 2022122494W WO 2023051671 A1 WO2023051671 A1 WO 2023051671A1
Authority
WO
WIPO (PCT)
Prior art keywords
voxel
scene
data
game
point
Prior art date
Application number
PCT/CN2022/122494
Other languages
English (en)
French (fr)
Inventor
熊鑫
廖焕华
李志凯
赵浩南
Original Assignee
上海莉莉丝互娱网络科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202111160592.XA external-priority patent/CN113877211A/zh
Priority claimed from CN202111163612.9A external-priority patent/CN113786617A/zh
Priority claimed from CN202111160564.8A external-priority patent/CN113877210A/zh
Application filed by 上海莉莉丝互娱网络科技有限公司 filed Critical 上海莉莉丝互娱网络科技有限公司
Priority to CN202280020323.XA priority Critical patent/CN116963811A/zh
Publication of WO2023051671A1 publication Critical patent/WO2023051671A1/zh

Links

Images

Classifications

    • AHUMAN NECESSITIES
    • A63SPORTS; GAMES; AMUSEMENTS
    • A63FCARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/50Controlling the output signals based on the game progress
    • A63F13/52Controlling the output signals based on the game progress involving aspects of the displayed game scene
    • AHUMAN NECESSITIES
    • A63SPORTS; GAMES; AMUSEMENTS
    • A63FCARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/60Generating or modifying game content before or while executing the game program, e.g. authoring tools specially adapted for game development or game-integrated level editor
    • A63F13/69Generating or modifying game content before or while executing the game program, e.g. authoring tools specially adapted for game development or game-integrated level editor by enabling or updating specific game elements, e.g. unlocking hidden features, items, levels or versions
    • AHUMAN NECESSITIES
    • A63SPORTS; GAMES; AMUSEMENTS
    • A63FCARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/70Game security or game management aspects
    • A63F13/75Enforcing rules, e.g. detecting foul play or generating lists of cheating players

Definitions

  • the present invention relates to the technical field of game data processing, in particular to a data processing method based on voxel data, a server, a medium and a computer program product.
  • a game scene is a collection of all scene elements in a virtual space in a video game, including map landforms, buildings, game characters, equipment props, etc.
  • the interface of the game scene seen by the game user is often displayed in the form of a pixel scene, that is, the content of the game scene is displayed on the display screen according to the pixel data format.
  • AI functions that is, to add one or more simulated players in a game, and the behavior of simulated players is calculated by the server.
  • behavior trees are commonly used to calculate AI behavior.
  • This calculation method uses CPU, which requires traversal of tree data structures and detection and navigation tasks that consume CPU performance. For high real-time competitive games Said that the calculation efficiency is low and cannot meet business needs.
  • the object of the present invention is to provide a data processing method based on voxel data, a server, a medium and a computer program product with higher computing efficiency.
  • the invention discloses a data processing method based on voxel data.
  • the original data of a game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the method includes:
  • the voxel modules of all the scene elements are spliced to obtain the voxel scene.
  • the scene elements include at least one of terrain, vegetation, buildings and outdoor decorations;
  • Said deriving the original data of several kinds of said scene elements respectively includes:
  • the setting of the desired unit voxel side length, combined with the unit voxel side length, converting the raw data of several kinds of scene elements into voxel data includes:
  • the surface height data of the terrain is converted into voxel data by sampling point by point according to the side length of a unit voxel.
  • the method also includes:
  • the voxel area within the target space range of the target object is obtained by clipping in the voxel scene, and the voxel area is used as an input of a neural network in the form of a three-dimensional tensor to obtain the spatial characteristics of the target object.
  • the clipping obtains the voxel area within the target space range of the target object, and using the voxel area in the form of a three-dimensional tensor as the input of the neural network to obtain the spatial characteristics of the target object includes:
  • the clipping obtains the voxel area within the target space range of the target object, and using the voxel area in the form of a three-dimensional tensor as the input of the neural network to obtain the spatial characteristics of the target object includes:
  • the input of the neural network is used to obtain the spatial features far ahead of the target object.
  • the clipping obtains the voxel area within the target space range of the target object, and using the voxel area in the form of a three-dimensional tensor as the input of the neural network to obtain the spatial characteristics of the target object includes:
  • the voxel data in the voxel scene For the specified target object and field of view, use the voxel data in the voxel scene to obtain a depth map with a resolution of M*N through ray detection, and use the depth map as the input of the neural network to obtain the The spatial characteristics of the target object.
  • obtaining a depth map with a resolution of M*N through ray detection includes:
  • the end of the viewing cone is a curved surface, and in the voxel scene, the curved surface includes M*N points;
  • the detection results of the M*N paths constitute the depth map.
  • performing ray detection on the M*N paths along the direction from the start point to the end point until a solid point on the path is detected includes:
  • the path includes a start point and an end point
  • calculation is performed on the M*N paths at the same time, so as to index to the corresponding point of each point in the voxel scene.
  • the method also includes:
  • the game processing process of the game object is carried out in the voxel scene, the voxel scene includes the game object and the field of view scene of the game object, and each customer The end corresponds to one or more of the game objects, and the determination of whether the behavior of the game objects is fraudulent includes:
  • the server side calculates the field of view scene that the game object should acquire at the current time, and sends the field of view scene to the client in real time;
  • the game account corresponding to the game object and the fraudulent behavior are recorded.
  • the area range of the field of view scene is the area range of the viewing cone of the game object
  • the judging whether the decision-making behavior satisfies the occurrence condition in the field of view based on the decision-making behavior that has occurred on the game object includes:
  • the calculation of the field of view scene that the game object should acquire at the current time includes:
  • the updated coordinates of all contour points are written into the voxel data, and the voxel data of the original position of all contour points before the dynamic game object is refreshed is erased to complete the dynamics of all game objects in the field of view. Refresh, so as to obtain the field of view scene after calculation refresh.
  • the judging whether the decision-making behavior satisfies the occurrence condition in the field of vision scene based on the decision-making behavior that has occurred in the game object, and if the occurrence condition is not satisfied, then it is considered that there is a fraudulent behavior includes:
  • the number of the navigation path is one or several, and the number of the travel mode is one or several;
  • the calculating and obtaining the navigation path and travel mode between the game object and the destination includes:
  • the z coordinate of is the height, forming several foundation columns, and there are L element points on each of the foundation columns;
  • Collect the continuous hollow point segments on each of the foundation columns if the height of the continuous hollow point segments is greater than or equal to the first preset height, define the continuous hollow point segments as a voxel layer;
  • the judgment step of radiation detection includes:
  • the judging step of the ray detection also includes: performing ray detection on the m paths in the GPU:
  • the m paths include m starting points and m end points; at the same time, calculating and obtaining n points passed by each path in the voxel scene;
  • the path is blocked; otherwise, the path is unblocked, so as to obtain the detection results of m paths.
  • it also includes:
  • a detection area includes x detection targets and y nodes, and dynamic planning is performed on the detection area, including:
  • the detection results of different paths are retrieved at different times for the detection area to perform the dynamic programming.
  • it also includes:
  • Dynamic planning for multiple detection areas including:
  • the detection result of the detection area is retrieved from the detection result table.
  • the present invention also discloses a data processing server based on voxel data.
  • the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the server includes:
  • An export module for exporting raw data of several kinds of scene elements respectively
  • the conversion module sets the desired side length of the unit voxel, combines the side length of the unit voxel, respectively converts the original data of several kinds of the scene elements into voxel data, and the voxel data in the volume In the voxel scene, it is represented as a voxel module;
  • the splicing module splices the voxel modules of all the scene elements according to the relative positions of the voxel modules of all the scene elements in the pixel scene to obtain the voxel scene.
  • the invention also discloses a computer-readable storage medium for storing data processing instructions based on voxel data.
  • the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the instruction Implements the following steps when executed:
  • the voxel modules of all the scene elements are spliced to obtain the voxel scene.
  • the invention also discloses a computer program product comprising computer-executable instructions executed by a processor to implement the following steps:
  • the voxel modules of all the scene elements are spliced to obtain the voxel scene.
  • voxel data For the calculation of spatial data, pixel data is converted into voxel data, which is highly compatible with GPU computing. Compared with the existing CPU computing mode, the performance of GPU-based data computing solutions is 2 to 3 orders of magnitude higher; Since all calculations are based on the voxel scene, on the server platform integrated with GPU, a large number of parallel calculations in the voxel scene are very fast, so that the server can bear a large amount of calculation load; using voxel data as the input of the neural network, It can make the AI behavior obtained by calculation more refined, intelligent, fast-responsive, and close to human behavior; use voxel data for high-performance ray detection to perceive the environment, and have high computing efficiency.
  • Fig. 1 is the flowchart of the data processing method based on voxel data provided by the present invention
  • FIG. 2 is a pixel scene of the prior art
  • Fig. 3 is a voxel scene converted from the pixel scene in Fig. 2 provided by the present invention.
  • first, second, third, etc. may be used in the present disclosure to describe various information, the information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of the present disclosure, first information may also be called second information, and similarly, second information may also be called first information. Depending on the context, the word “if” as used herein may be interpreted as “at” or “when” or “in response to a determination.”
  • connection should be understood in a broad sense, for example, it can be mechanical connection or electrical connection, or two
  • connection should be understood in a broad sense, for example, it can be mechanical connection or electrical connection, or two
  • connection should be understood in a broad sense, for example, it can be mechanical connection or electrical connection, or two
  • the internal communication of each element may be directly connected or indirectly connected through an intermediary.
  • intermediary Those skilled in the art can understand the specific meanings of the above terms according to specific situations.
  • voxel is the abbreviation of volume element, and the volume containing the voxel can be expressed by stereo rendering or extracting the polygonal isosurface of the given threshold contour.
  • a voxel is the smallest unit of digital data in three-dimensional space division, and the unit voxel mentioned in the present invention can be understood as a single voxel.
  • Voxels are used in areas such as 3D imaging, scientific data, and medical imaging.
  • pixels pixels are used in the image data of two-dimensional computer images.
  • the CPU-based computing mode is usually adopted, that is, various logical task operations and data preprocessing are performed on the CPU, while the GPU is more suitable for concurrent data operations, such as image rendering.
  • the combination effect of the existing spatial data structure and the GPU is not ideal, so the existing spatial data structure takes pixel data as the mainstream, which cannot reflect the high concurrent computing performance of the GPU.
  • the present invention converts the pixel data used for 3D imaging into volume Pixel data, which is highly compatible with the high-concurrency computing characteristics of the GPU, and performs various operations.
  • the performance of the GPU-based data computing solution is 2 to 3 orders of magnitude higher. Therefore, the voxel data is used for high-performance ray detection to sense the environment, and the calculation efficiency is high.
  • the present invention provides a specific embodiment of a data processing method based on voxel data, wherein the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and
  • the methods include:
  • the game scene is a game world created by UE4 game engine, referring to accompanying drawing 2, in this game world, scene elements include terrain, vegetation, buildings and outdoor decorations. Terrain such as slopes, hills, rivers, etc.; vegetation such as trees, flowers, shrubs, etc.; buildings such as houses, warehouses, etc.; outdoor decorations such as oil tanks, platforms, etc. All scene elements in Figure 2 are displayed in the form of pixels, and each element is composed of pixel blocks, that is, the game scene is described and displayed by pixel data, and the pixel scene is a way of expressing the game scene.
  • Terrain such as slopes, hills, rivers, etc.
  • vegetation such as trees, flowers, shrubs, etc.
  • buildings such as houses, warehouses, etc.
  • outdoor decorations such as oil tanks, platforms, etc.
  • All scene elements in Figure 2 are displayed in the form of pixels, and each element is composed of pixel blocks, that is, the game scene is described and displayed by pixel data, and the pixel scene is a way of expressing the game scene
  • the data types used by different scene elements are different when they are constructed. According to the types and characteristics of the scene elements, the original data of different scene elements are exported from the UE4 game engine in different ways. specific:
  • Outdoor decorations and buildings belong to the Actor type containing StaticMesh in the UE4 game engine, which can directly export OBJ files and synchronize the coordinate information of Actors;
  • OBJ files are 3D model files;
  • Vegetation is not an independent Actor type in UE4, so it is necessary to record coordinates and shapes and export it as a CSV information file;
  • the CSV information file is a comma-separated value file;
  • Terrain only uses the surface height information in the voxel world, and the present invention can export the image data of the terrain by adopting the method of orthogonal shooting with the depth camera.
  • a voxel module includes one or several unit voxels, combined with the side length of the unit voxel, respectively converts the data of several scene elements into voxel data, specifically:
  • OPEN3D voxel module
  • Terrain occupies only one layer in the voxel world, see area D in Figure 3, according to the image data stored with height information, it is sampled point by point according to the side length of the unit voxel to convert it into voxel data (voxel module ).
  • the voxel scene is represented as a large number of three-dimensional coordinate points in the program.
  • the principle of splicing the voxel module is actually to integrate the coordinate point information representing the voxel module into the same data according to the relative position of the voxel module in the game map.
  • the UE4 game engine itself contains the position information and rotation information of each module, so it is necessary to pay special attention to the inconsistent rules of the Euler angle rotation transformation of the 3D model in different systems when splicing.
  • the voxel scene shown in Figure 3 is obtained.
  • three-dimensional tensors are used to fully express the spatial information of the three-dimensional world.
  • the sampling accuracy determines the spatial resolution
  • the voxel data is used to record the game scene.
  • the voxel scene is a representation of the game scene.
  • it is flat vector data, which is very suitable for GPU computing, and the computing of the voxel scene through the GPU greatly improves the computing speed compared with the way of using the CPU in the prior art.
  • the target object can be a game object or a scene element, where the game object can be a character controlled by a player or a character controlled by an AI.
  • the voxel area within the target space range of the target object is obtained by clipping in the voxel scene, and the voxel area is used as an input of the neural network in the form of a three-dimensional tensor to obtain the spatial characteristics of the target object.
  • the target spatial range includes a first spatial characteristic range, a second spatial characteristic range...the Nth spatial characteristic range, wherein the target spatial range may be one of the spatial characteristic ranges, or a combination of multiple spatial characteristic ranges.
  • the present invention provides a preferred embodiment, including three spatial feature ranges, the first spatial feature range is the surrounding space range, that is, the range of adjacent spaces around the target object; the second spatial feature range is the near front space range, that is, the target The space range within a certain distance in front of the object; the third space feature range is the far front space range, that is, the space range beyond a certain distance in front of the target object; both the near front space range and the far front space range belong to the front space range.
  • the target spatial range does not include the spatial range of the target object itself.
  • the specific application examples of the above-mentioned surrounding space range, front space range, near front space range, and far front space range are as follows:
  • An example of the application of the surrounding space range with the target object as the center, crop the voxel cube within the surrounding space range of the target object, and use the voxel cube in the form of a three-dimensional tensor as the input of the neural network to obtain the target object surrounding spatial characteristics.
  • the clipped voxel cube is a combination of several unit voxels.
  • the voxel cube is used as the input of the neural network in the form of three-dimensional tensor.
  • the model of the neural network can be selected according to the specific application requirements, such as the convolutional neural network; the output result obtained through the calculation of the neural network is a set of vector data, the vector data That is, the spatial features around the target object.
  • the target object can be understood as a character in the game scene.
  • the character needs to know the equipment that can be picked up around the character, that is, it needs to capture the environmental information around the character, and extract and identify whether there is equipment information from the environmental information.
  • the environmental information is usually limited to a preset range. For example, when there is equipment information within two meters around the character, the character is notified that there is equipment that can be picked up around the character. In this case, the size of the cube corresponds to the value "two meters".
  • the front space range In front of the target object, cut and obtain a first voxel cuboid, and use the first voxel cuboid in the form of a three-dimensional tensor as the input of the neural network to obtain the space range in front of the target object. internal spatial characteristics.
  • the clipped first voxel cuboid is a combination of several unit voxels, and the first voxel cuboid may be a wide cuboid or a slender cuboid.
  • the target object can also be understood as a character in the game scene.
  • the character needs to recognize the behavior of the enemy in front of the field of vision, so as to make its own behavior selection according to the behavior, such as action dodge in fighting games; another example is missile dodge in scene games. That is, it is necessary to capture the environmental information and enemy behavior information in front of the character's field of vision, and extract from the environmental information and enemy behavior information within the environmental range to identify whether there are actions that can be cooperated with or avoided.
  • the environmental information and enemy behavior information are usually limited to a preset range. For example, when there is missile information within 50 meters in front of the character's field of vision, the character is informed that it needs to dodge or take other defensive actions. In this case, the length dimension of the cuboid corresponds to the value of "fifty meters".
  • a second voxel cuboid is obtained by cropping.
  • the length of the second voxel cuboid is much greater than the length of the first voxel cuboid.
  • the second voxel The cuboid is a slender cuboid, and the length of the first voxel cuboid is an order of magnitude multiple; the second voxel cuboid is used as the input of the neural network in the form of a three-dimensional tensor to obtain the target object in the far front space (ie The spatial characteristics of the distance ahead).
  • the clipped second voxel cuboid is a combination of several unit voxels.
  • the second voxel cuboid can be used independently or in combination with the first voxel cuboid.
  • the target object is also a character in the game.
  • the character's gun needs to use the scope to obtain a distant view, so the environment range that needs to be extracted at this time should be farther to correspond to the distant aiming object.
  • the length of the slender cuboid corresponds to the distant The distance between the aiming object and the character.
  • the length of the slender cuboid corresponds to the farthest distance that the scope can see, so as to ensure that the cropped area can meet the requirement of the sight distance of the scope.
  • the present invention applies voxel data to GPU ray detection, and generates depth , and then input the depth map as the feature extraction of the neural network, so that the performance of voxel data can be better utilized.
  • the voxel data in the voxel scene uses the voxel data in the voxel scene to obtain a depth map with a resolution of M*N through ray detection, and use the depth map as the input of the neural network to obtain the target object.
  • spatial features For the specified target object and field of view, use the voxel data in the voxel scene to obtain a depth map with a resolution of M*N through ray detection, and use the depth map as the input of the neural network to obtain the target object.
  • the generation of the depth map specifically includes:
  • the curved surface includes M*N points;
  • the viewing cone is an abstract conical spatial range in which the line of sight radiates from the eyes of a "person" who is one of the characters in the game scene.
  • the viewing cone is usually a plane (a curved surface with a curvature of 0).
  • the calculation is usually based on the spherical coordinate system, in which the viewing frustum is a curved surface.
  • the specific steps of radiation detection are:
  • the path between the starting point (x 1 , y 1 , z 1 ) and the ending point (x 2 , y 2 , z 2 ) is on the same straight line.
  • the so-called ray detection means starting from the starting point (x 1 , y 1 , z 1 ) to A ray is emitted from the end point (x 2 , y 2 , z 2 ), and a ray path is traveled between the two points.
  • the character in the course of the game, it is necessary to determine whether the character can hit a distant target with a gun.
  • the starting point of the detection is the character, and the end point is the distant target. If the detection result is no obstacle, it means that under the current aiming path, the normal shooting will be able to hit the distant target; if the detection result is there is an obstacle, it means that under the current aiming path, normal shooting Can't hit the distant target either.
  • the ray detection process of obtaining the depth map needs to detect where there is a blockage, that is, to obtain the depth and stop the detection, it can also be understood as a pixel map.
  • different spatial features can also be extracted and input into the neural network for spatial object segmentation, spatial object classification and behavior reinforcement learning.
  • the method further includes: determining whether the behavior of the game object is fraudulent.
  • the game processing process of the game object is carried out in the voxel scene.
  • the voxel scene includes the game object and the field of view scene of the game object.
  • the game object moves and executes decision-making behavior in the field of vision scene.
  • the field of view scene of a game object is a collection of scene elements within the field of view of the game object.
  • the data of the voxel scene is processed by the server and the client, and the server and the client are connected wirelessly or by wire.
  • the client can be a mobile phone, a tablet computer, a computer and other devices, and the server is a server or a server group. Usually each client corresponds to operate a game object, but for some game types, a client can also operate multiple game objects.
  • a voxel scene is a three-dimensional space composed of unit voxels. In this three-dimensional space, various spatial elements (such as terrain, buildings, game characters, and equipment props) are constructed by the data format of unit voxels
  • the method for judging the behavior of the game object of the present invention mainly includes performing preventive behaviors and halfway blocking behaviors on game fraudulent behaviors, specifically:
  • the server side calculates the field of view scene that the game object should acquire at the current time, and sends the field of view scene to the client in real time;
  • the idea of preventing fraudulent behavior is to put all the relevant game data (ie player data) of the game object on the server side, calculate the game data of each game object on the server side, and then send the game data to the client , the game data received by the client is only the data that the game object should receive, and there is no data that the game object should not receive.
  • a user obtains the perspective function through a plug-in program on the client,
  • Other game objects behind obstacles can be seen in the field of view of the game object; and if all the above data are placed on the server side, the calculation is performed on the server side, and the data actually required by each game object is sent to the corresponding client
  • the client does not distribute unnecessary data, it can prevent users from maliciously obtaining and using data, thereby avoiding the occurrence of game fraud from the source, that is, even if users use plug-ins on the client, they cannot see them in the field of vision.
  • the game object behind the obstacle because the data of the game object behind the obstacle will not be sent to the client.
  • the game data sent by the server to the client mentioned in the present invention is the dynamic data required to construct the overall game scene, but not all the data of the overall game scene. It can be understood as: including other game objects Data that can change dynamically, such as game props that can be picked up; does not include existing scene data that does not change, such as terrain and vegetation.
  • the size of the dynamic data is far smaller than the total data size of the overall game scene, which can reduce the communication data burden between the server and the client, and support real-time data communication during the running of the game.
  • game fraud also includes the use of normal data to conduct abnormal decision-making operations during the game. For example, under normal operating conditions, there is an occluder between this game object and other game objects, making it difficult to achieve a successful shooting action, or the distance between this game object and other game objects is too large, making it difficult to achieve a successful shooting action; and fraudulent behavior occurs During the game, this game object has finished shooting other game objects.
  • game object mentioned in the present invention can be understood as the game object operated by the currently monitored client, while other game objects can be understood as the game object operated by other clients, the AI game of the game server object, or a non-player object, etc.
  • the game objects can be game characters, or game elements other than game characters, such as vehicles, airdrops, equipment, etc.
  • the present invention also proposes an anti-fraud method that prevents halfway, that is, during the game process, the decision-making behavior of the game object is monitored in real time, and it is judged whether the current game environment supports the decision-making behavior. It has game fraud.
  • the anti-fraud method of midway blocking is also carried out on the server side, but it does not exclude that the anti-fraud method of midway blocking can also be carried out on the client side when the client side also obtains relevant permissions and is configured with relevant modules.
  • the field of view of this game object can be used as a part of the game scene to participate in the judgment.
  • the field of view is the game scene that the user can observe when operating the game object on the client.
  • game objects perform game actions (such as shooting), such game actions or decision-making behaviors are eligible to occur.
  • game actions such as shooting
  • other game objects behind the obstacle cannot be observed in the field of view of the game object, so that the game operation on other game objects is not supported, that is, the occurrence condition is not satisfied.
  • Fraudulent behavior is determined to exist if the game operation on other game objects that cannot be observed is detected.
  • the fraudulent behavior After the fraudulent behavior is determined, immediately record the game account corresponding to the game object and the fraudulent behavior. Furthermore, the game can be terminated immediately, or other measures can be taken according to the seriousness of the fraudulent behavior.
  • the area range of the field of view scene is the area range of the viewing cone of the game object, and the area within the area range of the viewing cone is considered to be the area that the player (that is, the client) should see on the client display interface.
  • the game object may be a game character in the game, and the game character has an anthropomorphic field of vision, and the field of view of the game character can be observed from a first perspective or a third perspective on the client.
  • this game object should not be able to see the other game objects.
  • the game object outside the house should not be able to see other game objects in the house.
  • the decision-making behavior of the game object includes shooting behavior, hitting behavior, healing behavior and so on.
  • the present invention judges whether the path between the game object and other game objects is unimpeded through ray detection, if it is unblocked, and the distance between the game object and other game objects meets the shooting requirements, it is considered that the shooting condition is met, and it can be Complete the shooting action; otherwise, if the path between this game object and other game objects is blocked or the distance between this game object and other game objects does not meet the shooting requirements, the shooting (hitting) behavior should not occur.
  • the decision-making behavior is determined only after the decision-making behavior is detected. Therefore, it can be determined that the player’s decision-making behavior is fraudulent after it is determined that the current game environment does not satisfy the decision-making behavior.
  • the hitting behavior it is usually judged whether the condition is satisfied by the distance between the game object and other game objects. If the distance between the game object and other game objects exceeds the distance supported by the hitting behavior, it is considered There is fraud.
  • the ray detection process based on the voxel scene includes:
  • each coordinate point in the voxel scene has an attribute value. If the attribute value is 1, it means that there is an object or a part of the object on the coordinate point, such as game characters, buildings and other objects will occupy a large number of coordinate positions , the attributes of the coordinate points of these occupied positions are all 1, that is, solid points; correspondingly, if the attribute value is 0, it means that there is no object on the coordinate point, which is an open scene, that is, a hollow point.
  • the index can be understood as a search. When detecting whether n points are solid points, you first need to "reach” the n points, and the index is the step of "arriving". Can be understood as a proprietary step in computer processing.
  • 1 In the world of voxel data, 1 represents a solid point, and 0 represents a hollow point.
  • the embodiment of an object in the world of voxel data is a series of 1s at certain positions. To write an object means to write 1s to certain positions in the voxel data world; to erase an object means to write 0s to certain positions in the voxel data world.
  • the present invention applies voxel data to GPU ray detection, which can realize high-speed multi-path concurrent operation. It can be understood that, in GPU, the time for using voxel data to detect multiple rays is the same as the time to detect one ray. Therefore, the time for multi-ray detection can be greatly shortened, thereby greatly reducing the calculation load on the server side, making it possible to centralize the calculation of each player's game data on the server side.
  • the voxel scene can also be other application scenarios.
  • the scene elements will not only include terrain, vegetation, buildings and outdoor decorations, or be different from terrain, vegetation, buildings and outdoor decorations.
  • the data export of other scene elements and the conversion of voxel data may adopt methods different from those in this embodiment, which are not limited here.
  • voxel data can only be used for calculations in GPU. In CPU, voxel data can still be used for calculations. However, parallel calculations cannot be performed during ray detection, and only multiple paths can be used. The detection tasks are executed sequentially.
  • the so-called ray detection method means that a ray is sent from the starting point to the end point, and there is a line between the starting point and the end point , the points on the line segment are the n points that need to pass through.
  • array A the number of array elements is n, and each array element contains voxel scene coordinate information, that is, three-dimensional coordinate (x, y, z) information);
  • each point in the array A is calculated at the same time, that is, n threads are concurrent, and each thread performs the following process:
  • Index points in voxel data according to their coordinates in the voxel scene, and detect whether they are solid points. If a solid point is detected, write 1 to the result, which is the first detection result, indicating that the thread is blocked; if no solid point is detected, write 0 to the result, which is the second detection result, indicating that the thread corresponds
  • the path can be passed without obstruction.
  • each pair of start and end points in the array B is calculated at the same time, that is, m parent threads are concurrent, and each thread executes the following process:
  • n points into array A' (the number of array elements is n, and each array element contains voxel scene coordinate information, that is, three-dimensional coordinate (x, y, z) information);
  • each point in the array A is calculated at the same time, that is, n sub-threads are concurrent, and the following process is executed in each sub-thread (at this time, n sub-threads are concurrently executed for each parent thread, that is, a total of n *m threads): Index the point in the voxel data according to the coordinates in the voxel scene, and check whether it is a solid point; if a solid point is detected, write 1 to the result, indicating that the path corresponding to the sub-thread is blocked Blocking; if no solid point is detected, write 0 to the result, indicating that the path corresponding to the sub-thread has no obstacles to pass through;
  • the invention is measured on the processor RTX3090 and AMD threadriper 3990X: when 1 million or more rays are detected at the same time, the detection speed of the GPU is about 550 times that of a single CPU core, and the operation speed is significantly improved.
  • voxel scene dynamic programming is usually required to calculate the navigation task, and the output result is a path.
  • a dynamic programming task several nodes to be explored are included. It should be noted that this node is a node in the sense of the task process, and does not represent a specific reference in the data. In some embodiments, a node may be a specific voxel scene coordinate point.
  • the exploration of nodes is sequential. It is unknown whether ray detection is required when exploring nodes, and which targets are to be ray detected, and a time sequence task is formed.
  • the present invention can carry out parallelization time dispersion task owing to use GPU:
  • the detection target may be a voxel scene coordinate point, such as a target object in the game world or a coordinate point on the exterior of a scene building.
  • the storage space occupied by the ray detection results of the specified detection area is very small, one result occupies one byte at most, and 1 million detection results can be stored in 1MB, so the execution efficiency of the dynamic programming algorithm task of the present invention is relatively high. The time to read it.
  • the detection subjects include several detection targets, that is, simultaneously perform ray detection on x*y paths of the detection areas corresponding to different detection targets, and
  • the detection results are saved, and a detection result table divided by detection areas is formed; when a detection area is dynamically planned, the detection results of the detection area can be retrieved from the detection result table.
  • ray detection is performed on x*y paths in each area at the same time, and the detection time is only a single ray detection time t.
  • the detection result of area A is a
  • the detection result of area B is b
  • the detection result of area C is c
  • the detection result of area D is d
  • the detection result table of A-a, B-b, C-c, D-d
  • the ray detection time is only a single ray detection time t.
  • follow-up in the actual process of dynamic programming can be read in time, and the calculation speed is very fast.
  • the game object often has a displacement during the game, that is, the position of the game object is different at different times.
  • the game object is called a dynamic game object here, and it is necessary to refresh the dynamic game in the field of vision scene according to a preset cycle.
  • the field of view scenes that game objects should obtain at the current time are calculated on the server side, including:
  • the coordinates of the contour points are the coordinates of the contours that make up the object, and the writing voxel data of the object is to write 1 in the positions of several contour points.
  • the center point coordinate of an airdrop in the game is (10.10.30), and one of the outline point coordinates of the airdrop is (0.20.50), then the offset coordinate of the outline point relative to the center point is (-10.10 .20).
  • the center point coordinate of the dynamic game object after refreshing is (10.10.25), then the coordinate of the contour point after refreshing is (0.20.45). Write 1 at the position of the refreshed contour point (0.20.45), and write 0 at the original position (0.20.50).
  • Several outline points that make up the airdrop are refreshed at the same time, which realizes the refresh of the airdrop position.
  • Dynamic game objects can be all game objects that may move, such as game characters, vehicles, and airdrops.
  • fraudulent behaviors can also be judged by judging whether the route traveled by the game object and the way of passing are compliant. specific:
  • the object of this game is on flat ground, and the destination to go is a house on a cliff. If you take a straight line (ie, the navigation path), you need to use the equipment to spray (fly) (ie, the way of passage) to reach it; if you take a detour If you follow a curve (i.e., a navigation path), you can reach it by walking (i.e., a way of access). That is to say: the straight path only supports jet (flying) traffic mode, and the curved path supports walking, jumping, jet (flying) and other traffic modes. If it is detected that the game object reaches the house by walking in a straight line and does not have jet (flying) equipment, it can be determined that the game object has fraudulent behavior.
  • the object of this game is by the river, and the destination to go to is the other side of the river.
  • the game sets the depth of the river and only supports taking water vehicles or jetting (flying) to the other side of the river. Pass without using any water vehicle or equipment. If it is detected that the game character passes through the river without using any water vehicle or equipment, it will be considered fraudulent.
  • the game object For another example, if it is set that the game object will consume an energy value of s after moving a certain distance, but it is actually detected that the energy consumed by the game object after moving a certain distance is less than s, then the game object is considered to be fraudulent .
  • the calculation and acquisition of the navigation path and way of travel between the game object and the destination is based on the layer data of the voxel scene and the connection data between the layer data and the layer data, specifically including:
  • the invention also discloses a data processing system based on voxel data.
  • the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the system includes:
  • the export module is used to export the original data of several scene elements respectively;
  • the conversion module sets the desired side length of the unit voxel, combines the side length of the unit voxel, and converts the original data of several scene elements into voxel data, and the voxel data is expressed as a voxel module in the voxel scene ;
  • the splicing module splices the voxel modules of all scene elements according to the relative positions of the voxel modules of all scene elements in the pixel scene to obtain a voxel scene.
  • the system includes a hardware structure and a computer-readable storage medium.
  • the above functional modules may be integrated on the hardware structure or on the computer-readable storage medium, which is not limited here.
  • the connection relationship of the above-mentioned functional modules may be a tangible connection or an intangible cross-regional connection, which is not limited here.
  • the system and the corresponding method embodiments belong to the same idea, and its specific implementation process is detailed in the corresponding method embodiments, which will not be repeated here.
  • the present invention also discloses a data processing server based on voxel data.
  • the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the server includes:
  • the export module is used to export the original data of several scene elements respectively;
  • the conversion module sets the desired side length of the unit voxel, combines the side length of the unit voxel, and converts the original data of several scene elements into voxel data, and the voxel data is expressed as a voxel module in the voxel scene ;
  • the stitching module according to the relative positions of the voxel modules of all scene elements in the pixel scene, stitches the voxel modules of all scene elements to obtain the voxel scene.
  • server and the corresponding method embodiments belong to the same idea, and its specific implementation process is detailed in the corresponding method embodiments, which will not be repeated here.
  • the invention also discloses a computer-readable storage medium for storing data processing instructions based on voxel data.
  • the original data of the game scene constitutes a pixel scene, and the pixel scene includes several scene elements of different data types, and the instruction is executed Implement the following steps:
  • the voxel modules of all scene elements are spliced to obtain a voxel scene.
  • the computer-readable storage medium can be integrated in hardware, and when the hardware is running, the computer-readable storage medium can be supported to read and run.
  • the invention also discloses a computer program product, including computer-executable instructions, the instructions are executed by a processor to implement the following steps:
  • the voxel modules of all scene elements are spliced to obtain a voxel scene.

Abstract

本发明提供了一种基于体素数据的数据处理方法、服务器、介质及计算机程序产品,所述方法包括:分别导出若干种所述场景元素的原始数据;设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。本发明对于空间数据的计算,将像素数据转化为体素数据,从而与GPU运算高度契合,相对于既有的CPU运算模式,以GPU为基础的数据运算方案性能要高2到3个数量级。

Description

基于体素数据的数据处理方法、服务器、介质及计算机程序产品 技术领域
本发明涉及游戏数据处理技术领域,尤其涉及一种基于体素数据的数据处理方法、服务器、介质及计算机程序产品。
背景技术
游戏场景是电子游戏中一虚拟空间内所有场景元素的集合,包括地图地貌、建筑物、游戏角色、装备道具等内容。游戏用户所看到的游戏场景的界面往往以像素场景的方式展现,即按照像素的数据格式将游戏场景的内容展示在显示屏上。
在多人竞技游戏中,往往需要加入AI功能,也就是在一局游戏中加入一个或多个模拟玩家,模拟玩家的行为由服务器来运算。
现有的技术手段中,常用行为树来计算AI行为,这样的计算方式使用CPU进行,需要进行树状数据结构的遍历以及非常消耗CPU性能的检测和导航任务,对于高实时性的竞技游戏来说,计算效率较低,无法满足业务需求。
发明内容
为了克服上述技术缺陷,本发明的目的在于提供一种运算效率更高的基于体素数据的数据处理方法、服务器、介质及计算机程序产品。
本发明公开了一种基于体素数据的数据处理方法,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且所述方法包括:
分别导出若干种所述场景元素的原始数据;
设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
优选的,所述场景元素包括地形、植被、建筑和室外摆件中的至少一种;
所述分别导出若干种所述场景元素的原始数据包括:
导出所述室外摆件与所述建筑的3D模型文件格式数据和坐标信息;
导出所述植被的逗号分隔值文件格式数据;
通过深度相机正交拍摄以导出地形的图片,该图片中包含地形的表面高度数据。
优选的,所述设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据包括:
通过开源库中的“读写三角网格”功能和“从三角网格创建”功能将所述室外摆件与所述建筑的所述3D模型文件格式数据转换为体素数据;
获取所述植被的碰撞体大小,结合所述单元体素的边长,计算获取所述植被在所述体素场景中需要占用的体素个数及体素形状;
根据存储有所述地形的所述表面高度数据的所述图片,按照单元体素的边长逐点采样以将地形的所述表面高度数据转换为体素数据。
优选的,所述方法还包括:
在所述体素场景内裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征。
优选的,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
以所述目标对象为中心,裁剪获取所述目标对象的周围空间范围内的体素正方体,将所述体素正方体以三维张量的形式作为神经网络的输入,以获取所述目标对象周围的空间特征。
优选的,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
在所述目标对象的前方空间范围内,裁剪获取一第一体素长方体,将所述第一体素长方体以三维张量的形式作为神经网络的输入,以获取所述目标对象前方的空间特征,
在所述目标对象的前方空间范围内,裁剪获取一第二体素长方体,其中所述第二体素长方体的长度远大于所述第一体素长方体的长度,将所述第二体素长方体以三维张量的形式作为神经网络的输入,以获取所述目标对象前方远处的空间特征。
优选的,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
对指定的目标对象和视野,使用所述体素场景中的体素数据,通过射线检测获取一分辨率为M*N的深度图,将所述深度图作为神经网络的输入,以获取所述目标对象的空间特征。
优选的,所述对指定的目标对象和视野,使用所述体素场景中的体素数据,通过射线 检测获取一分辨率为M*N的深度图包括:
在所述视野的方向生成一视锥,所述视锥的尽头为曲面,在所述体素场景中,所述曲面包括M*N个点;
以所述目标对象为起点、所述M*N个点为终点,形成M*N条路径,对M*N条路径沿自起点至终点的方向进行射线检测,直到检测到路径上的实心点;
所述M*N条路径的检测结果构成所述深度图。
优选的,所述对M*N条路径沿自起点至终点的方向进行射线检测,直到检测到路径上的实心点包括:
所述路径包括起点和终点;
计算获取所述路径在所述体素场景中经过的n个点;
在GPU中,对所述M*N条路径同时进行计算,以索引到每个点在所述体素场景中对应的点。
优选的,所述方法还包括:
判定游戏对象的行为是否存在欺诈,其中,所述游戏对象的游戏处理过程在所述体素场景中进行,所述体素场景包括所述游戏对象和所述游戏对象的视野场景,每个客户端对应一个或多个所述游戏对象,并且所述判定游戏对象的行为是否存在欺诈包括:
服务器端计算所述游戏对象在当前时间应该获取的所述视野场景,将所述视野场景实时下发至所述客户端;
基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件,若不满足发生条件,则认为存在欺诈行为;
记录所述游戏对象所对应的游戏账号和所述欺诈行为。
优选的,所述视野场景的区域范围为游戏对象的视锥的区域范围;
在所述视锥的区域范围内,若通过射线检测判断本游戏对象与其他游戏对象之间的路径存在阻挡,则不显示所述其他游戏对象。
优选的,所述基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件包括:
通过射线检测判断本游戏对象与其他游戏对象之间的路径是否畅通,若畅通,则所述决策行为满足发生条件,若不畅通,则所述决策行为不满足发生条件。
优选的,所述计算所述游戏对象在当前时间应该获取的所述视野场景包括:
在所述视野场景中,获取所有动态的游戏对象的类型及每一类型的动态的游戏对象的 轮廓点相对于该游戏对象的中心点的偏移坐标;
根据最近一预设周期结束时的动态的游戏对象的中心点坐标和所述轮廓点的偏移坐标计算获取刷新后的所述轮廓点的更新坐标;
在动态的游戏对象刷新后的所有轮廓点的更新坐标写入体素数据,并擦除动态的游戏对象刷新前的所有轮廓点的原始位置的体素数据,完成视野场景内所有游戏对象的动态刷新,从而得到计算刷新后的视野场景。
优选的,所述基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件,若不满足发生条件,则认为存在欺诈行为包括:
计算获取所述游戏对象与目的地之间的导航路径和通行方式;所述导航路径的数量为一条或若干条,所述通行方式的数量为一种或若干种;
判断所述导航路径是否包括所述游戏对象的实际行走路径,并判断所述实际行走路径上所支持的所述通行方式是否包含所述游戏对象实际所采用的通行方式,若否则认为存在欺诈行为。
优选的,所述计算获取所述游戏对象与目的地之间的导航路径和通行方式包括:
在所述体素场景的x,y,z三维空间中,取z=z′的一基础平面,所述基础平面上存在若干个基础点;以所述基础点为底点,以该基础点的z坐标为高度,形成若干根基础柱,每根所述基础柱上都存在L个元素点;
在GPU中并行地访问所有所述基础柱,并在每个并行的线程中遍历所述基础柱上的每个所述元素点,并根据体素数据索引到其对应的体素场景坐标点,判断该体素场景坐标点是空心点或实心点;
收集每根所述基础柱上的连续空心点段,若所述连续空心点段的高度大于等于第一预设高度,则定义该连续空心点段为一个体素层;
计算获取游戏对象当前所位于的所述体素层与相邻所述基础柱上的每个所述体素层之间的位置关系,从而获取游戏对象当前所位于的所述体素层与相邻所述基础柱上的每个所述体素层之间的通行方式。
优选的,射线检测的判断步骤包括:
计算获取本游戏对象(x 1,y 1,z 1)和其他游戏对象(x 2,y 2,z 2)形成的射线路径在所述体素场景中经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)、...、(a n,b n,c n);
在GPU中,同时对所述n个点进行计算:根据所述体素场景中的坐标索引到所述体素场景中n个点,并检查其是否为实心点;
若检测到所述n个点中有实心点,则所述本游戏对象与其他游戏对象之间的路径存在阻挡;若检测到所述n个点中没有实心点,则本游戏对象与其他游戏对象之间的路径畅通。
优选的,所述射线检测的判断步骤还包括:在GPU中对m条路径进行射线检测:
所述m条路径包括m个起点和m个终点;同时计算获取每条所述路径在体素场景中经过的n个点;
同时对m条路径中的所述n个点进行计算:根据体素场景中的坐标索引每个点在所述体素场景中对应的点,并检测其是否为实心点;
若检测到所述n个点中有实心点,则所述路径存在阻挡;否则所述路径畅通,以此获得m条路径的检测结果。
优选的,还包括:
一检测区域包括x个检测目标和y个节点,对所述检测区域进行动态规划,包括:
同时对所述x个检测目标和y个节点所形成的x*y个路径进行射线检测,并保存检测结果;
对所述检测区域在不同时间调取不同路径的所述检测结果以进行所述动态规划。
优选的,还包括:
对多个检测区域进行动态规划,包括:
同时对不同所述检测区域的x*y个路径进行射线检测,并保存检测结果,形成以所述检测区域划分的检测结果表;
对一检测区域进行所述动态规划时、则从所述检测结果表中调取所述检测区域的所述检测结果。
本发明还公开了一种基于体素数据的数据处理服务器,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且所述服务器包括:
导出模块,分别导出若干种所述场景元素的原始数据;
转换模块,设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
拼接模块,根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
本发明还公开了一种存储基于体素数据的数据处理指令的计算机可读存储介质,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且 所述指令在被执行时实施以下步骤:
分别导出若干种所述场景元素的原始数据;
设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
本发明还公开了一种计算机程序产品,包括计算机可执行指令,所述指令被处理器执行以实施以下步骤:
分别导出若干种所述场景元素的原始数据;
设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
采用了上述技术方案后,与现有技术相比,具有以下有益效果:
1.对于空间数据的计算,将像素数据转化为体素数据,从而与GPU运算高度契合,相对于既有的CPU运算模式,以GPU为基础的数据运算方案性能要高2到3个数量级;所有计算由于是基于体素场景的,在集成有GPU的服务器平台上,在体素场景中的大量并行计算都十分快速,使得服务器可以负担大量计算负载;使用体素数据作为神经网络的输入,可以使得计算获得的AI行为更加精细,智能化,反应快,贴近人的行为方式;使用体素数据进行高性能射线检测,用以感知环境,运算效率高。
2.将校验计算由客户端挪到服务器端执行,杜绝了客户端欺诈行为发生的可能;具体对每个游戏对象进行视野裁剪,即仅向该游戏对象下发该游戏对象应该看见的视野数据,而客户端仅计算所述视野场景内的空间数据;还通过射线检测等技术实时判断游戏对象的决策是否合理,两者结合可以做到预先防止和中途阻止游戏欺诈行为;
3.通过体素数据进行射线检测以生成深度图,将深度图作为神经网络的特征提取,可以进行后续一系列的空间物体分割、空间物体分类和行为强化学习等任务。
附图说明
图1为本发明提供的基于体素数据的数据处理方法的流程图;
图2为现有技术的像素场景;
图3为本发明提供的将图2的像素场景转换成的体素场景。
具体实施方式
以下结合附图与具体实施例进一步阐述本发明的优点。
这里将详细地对示例性实施例进行说明,其示例表示在附图中。下面的描述涉及附图时,除非另有表示,不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开相一致的所有实施方式。相反,它们仅是与如所附权利要求书中所详述的、本公开的一些方面相一致的装置和方法的例子。
在本公开使用的术语是仅仅出于描述特定实施例的目的,而非旨在限制本公开。在本公开和所附权利要求书中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式,除非上下文清楚地表示其他含义。还应当理解,本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解,尽管在本公开可能采用术语第一、第二、第三等来描述各种信息,但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如,在不脱离本公开范围的情况下,第一信息也可以被称为第二信息,类似地,第二信息也可以被称为第一信息。取决于语境,如在此所使用的词语“如果”可以被解释成为“在……时”或“当……时”或“响应于确定”。
在本发明的描述中,需要理解的是,术语“纵向”、“横向”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”、“内”、“外”等指示的方位或位置关系为基于附图所示的方位或位置关系,仅是为了便于描述本发明和简化描述,而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作,因此不能理解为对本发明的限制。
在本发明的描述中,除非另有规定和限定,需要说明的是,术语“安装”、“相连”、“连接”应做广义理解,例如,可以是机械连接或电连接,也可以是两个元件内部的连通,可以是直接相连,也可以通过中间媒介间接相连,对于本领域的普通技术人员而言,可以根据具体情况理解上述术语的具体含义。
在后续的描述中,使用用于表示元件的诸如“模块”、“部件”或“单元”的后缀仅为了有利于本发明的说明,其本身并没有特定的意义。因此,“模块”与“部件”可以混合地使用。
关于体素场景,体素是体积元素的简称,包含体素的立体可以通过立体渲染或者提取给定阈值轮廓的多边形等值面表现出来。体素是数字数据于三维空间分割上的最小单位,本发明所提及的单元体素可以理解为单个体素。体素用于三维成像、科学数据与医学影像 等领域。概念上类似二维空间的最小单位:像素,像素用在二维计算机图像的影像数据上。有些真正的三维显示器运用体素来描述它们的分辨率,例如:可以显示512×512×512体素的显示器。
在三维成像技术领域,通常采用以CPU为主的运算模式,即在CPU上进行各类逻辑任务运算及数据预处理,而GPU上更适合并发类的数据运算,如图像渲染。现有的空间数据结构与GPU的结合效果并不理想,故现有的空间数据结构以像素数据为主流,无法体现GPU的高并发运算性能,本发明将用于三维成像的像素数据转化为体素数据,从而与GPU高并发运算特性高度契合,进行各类运算,相对于既有的CPU运算模式,以GPU为基础的数据运算方案性能要高2到3个数量级。从而使用体素数据进行高性能射线检测,用以感知环境,运算效率高。
具体的,参见附图1,本发明提供一种基于体素数据的数据处理方法的具体实施例,其中,游戏场景的原始数据构成像素场景,像素场景包括若干种不同数据类型的场景元素,并且所述方法包括:
S100、分别导出若干种场景元素的原始数据;
S200、设定期望的单元体素的边长,结合单元体素的边长,分别将若干种场景元素的原始数据转换为体素数据,体素数据在体素场景中表现为体素模块;
S300、根据所有场景元素的体素模块在像素场景中的相对位置,拼接所有场景元素的体素模块,获得体素场景。
在本发明的优选实施例中,游戏场景为UE4游戏引擎建立的游戏世界,参见附图2,在该游戏世界中,场景元素包括地形、植被、建筑和室外摆件。地形如土坡、山丘,河流等;植被如树木、花草、灌木等;建筑如房屋、仓库等;室外摆件如油罐、平台等。附图2中的所有场景元素都是以像素的形式展现的,每个元素由一个个像素块组成,即游戏场景以像素数据来描述并展现,像素场景是游戏场景的一种表现方式。
而不同的场景元素在构建时所用到的数据类型是不同的。根据场景元素的类型及其特征,采用不同的方式从UE4游戏引擎中分别导出不同场景元素的原始数据。具体的:
室外摆件与建筑在UE4游戏引擎中属于包含StaticMesh的Actor类型,可直接导出OBJ文件,同步导出Actor的坐标信息;OBJ文件即为3D模型文件;
植被在UE4中不属于独立的Actor类型,故需要采取记录坐标和形状的方式,导出为CSV信息文件;CSV信息文件即为逗号分隔值文件;
地形在体素世界中只利用到了表面高度信息,本发明采取深度相机正交拍摄的方式则 可以导出地形的图片数据。
以上几种原始数据的导出方式均可参考UE4游戏引擎的说明文档实现,是本领域技术人员所掌握的技术手段。
由于场景元素的类型及其特征不同,故需要采取不同的转换方式以将不同场景元素的各类原始数据转换为在游戏场景中表现为体素模块的体素数据,首先需要设定我们所期望的单元体素的边长,一个体素模块包括一个或若干个单元体素,结合单元体素的边长,分别将若干种场景元素的数据转换为体素数据,具体的:
对于OBJ(3D模型)文件,可以借助OPEN3D这一开源库中的“读写三角网格”(read_triangle_mesh)功能和“从三角网格创建”(create_from_triangle_mesh)功能直接将OBJ文件转换为体素数据(体素模块);OPEN3D并不为本发明的限制的开源库,其他可以实现上述两功能的开源库也可以被用于进行数据转换;
对于植被,需要根据植被的碰撞体的大小,结合体素的边长,直接计算出它在体素世界中占用的体素模块的个数及形状;
地形在体素世界中只占一层,参见图3中的D区域,根据储存有高度信息的图片数据,按照单元体素的边长逐点采样以将其转换为体素数据(体素模块)。
至此已获取到建立体素场景的所有数据,最后需要拼接具有所有场景元素的体素模块,获得所有场景元素的体素场景。
体素场景在程序中表示为大量的三维坐标点,拼接体素模块的原理实际上是根据体素模块在游戏地图中的相对位置,把表示体素模块的坐标点信息,整合到同一个数据结构中,而UE4游戏引擎本身包含各个模块的位置信息和旋转信息,故在拼接时需要特别注意3D模型的欧拉角的旋转变换在不同体系中规则不一致的问题。
经过拼接获得图3所示的体素场景,在该体素场景中,使用三维张量完备的表达三维世界的空间信息,其采样精度决定了空间分辨率,实现了用体素数据记录游戏场景,即体素场景是游戏场景的一种表现方式。基于该体素场景的数据格式特点,为扁平化的向量数据,非常适合GPU运算,通过GPU对体素场景进行运算,比现有技术中使用CPU的方式极大提升了运算速度。
较佳的,基于像素场景转换为体素场景的基础,用于裁剪目标对象的目标空间范围,以此目标空间范围作为神经网络的输入,通过神经网络来获取空间特征,该空间特征包括空间内所包含的信息,例如目标空间范围内的体素点是否为实心点等。目标对象可以是游戏对象,也可以是场景元素,其中游戏对象可以是玩家控制的角色,也可以是AI控制的 角色。
具体的,在体素场景内裁剪获取目标对象的目标空间范围内的体素区域,将体素区域以三维张量的形式作为神经网络的输入,以获取目标对象的空间特征。
目标空间范围包括第一空间特征范围、第二空间特征范围...第N个空间特征范围,其中,目标空间范围可以是其中一种空间特征范围,也可以是多种空间特征范围的组合。本发明给出一优选实施例,包括三个空间特征范围,第一空间特征范围为周围空间范围,即目标对象的四周相邻空间的范围;第二空间特征范围为近前方空间范围,即目标对象前方某个距离内的空间范围;第三空间特征范围为远前方空间范围,即目标对象前方某个距离之外的空间范围;近前方空间范围和远前方空间范围都属于前方空间范围。综上所述,明显的,目标空间范围不包括目标对象本身的空间范围。上述周围空间范围、前方空间范围、近前方空间范围、远前方空间范围具体的应用示例如下:
一、有关周围空间范围的应用示例,以目标对象为中心,裁剪获取目标对象的周围空间范围内的体素正方体,将体素正方体以三维张量的形式作为神经网络的输入,以获取目标对象周围的空间特征。所裁剪获得的体素正方体为若干个单元体素的组合。体素正方体以三维张量的形式作为神经网络的输入,该神经网络的模型可以根据具体应用需求选取,如卷积神经网络;经过神经网络计算得到的输出结果是一组向量数据,该向量数据即目标对象周围的空间特征。
在本实例中,目标对象可理解为游戏场景内的角色。角色在游戏过程中需要获知周围的可拾取的装备,即需要捕捉到角色周围的环境信息,从环境信息中进行提取识别是否存在装备信息。该环境信息通常被限制在一个预设范围,例如,当角色周围两米内有装备信息,则告知角色周围有可拾取的装备。此种情况下,该正方体的尺寸即对应“两米”这一数值。
之所以说“对应”而不是“相等”,是由于游戏场景的画面数据的尺寸不一定与角色实际感知的尺寸完全一致,可能存在放大或缩小的对应关系。
二、有关前方空间范围的应用示例,在目标对象的前方,裁剪获取一第一体素长方体,将第一体素长方体以三维张量的形式作为神经网络的输入,以获取目标对象前方空间范围内的空间特征。所裁剪获得的第一体素长方体为若干个单元体素的组合,该第一体素长方体可以是较宽形状的长方体,也可以是细长形状的长方体。
在本实例中,目标对象同样可理解为游戏场景内的角色。角色需要识别视野前方的敌方行为动作,从而根据该行为动作进行自身的行为选择,例如搏击游戏中的动作闪避;又 例如场景游戏中的飞弹躲避等。即需要捕捉到角色视野前方的环境信息和敌方行为信息,从环境信息和环境范围内的敌方行为信息中进行提取识别是否存在可以与之相合作或躲避的动作。该环境信息和敌方行为信息通常被限制在一个预设范围,例如,当角色视野前方五十米内有飞弹信息,则告知角色需要进行相关躲避或采取其他防御行为。此种情况下,该长方体的长度尺寸即对应“五十米”这一数值。
三、有关远前方空间范围的应用示例,在目标对象的前方,裁剪获取一第二体素长方体,第二体素长方体的长度远大于第一体素长方体的长度,例如,该第二体素长方体为一细长长方体,与第一体素长方体的长度呈数量级的倍数关系;将第二体素长方体以三维张量的形式作为神经网络的输入,以获取目标对象远前方空间范围内(即前方远处)的空间特征。所裁剪获得的第二体素长方体为若干个单元体素的组合。第二体素长方体可独立应用,也可与第一体素长方体组合应用。
在本实例中,目标对象同样为游戏内的角色。角色的持枪需要借助瞄准镜来获取远方的视野,则此时所需要提取的环境范围要更远以对应远处的瞄准物,此种情况下,该细长长方体的长度尺寸即对应远处的瞄准物与角色之间的距离。
在设置时,可以设置为该细长长方体的长度尺寸对应瞄准镜可以看见的最远的距离,即可保证所裁剪的区域能满足瞄准镜的视野距离要求。
除了将体素数据直接应用于神经网络的输入,由于体素数据格式规范、整齐,易于计算、特别是并行计算,故本发明将体素数据应用于GPU射线检测,并通过射线检测来生成深度图,再将深度图作为神经网络的特征提取来输入,以此可以更好的发挥体素数据的性能。
具体的,对指定的目标对象和视野,使用体素场景中的体素数据,通过射线检测获取一分辨率为M*N的深度图,将深度图作为神经网络的输入,以获取目标对象的空间特征。
较佳的,深度图的生成具体包括:
1)在视野的方向生成一视锥,视锥的尽头为曲面,在体素场景中,曲面包括M*N个点;
2)以目标对象为起点、M*N个点为终点,形成M*N条路径,对该M*N条路径沿自起点至终点的方向进行射线检测,直到检测到该路径上的实心点,检测结果表现为该路径的像素;
3)M*N条路径的检测结果构成深度图。
需要说明的,视锥为视线从作为游戏场景的角色之一的“人”的眼睛放射出去的呈一 抽象的圆锥状的空间范围,视锥通常为平面(曲率为0的曲面),在体素场景中,为了便于计算,通常基于球坐标系进行计算,在该球坐标系中,视锥即为曲面。
在沿自起点至终点的方向(也可理解为由近及远)的射线检测过程中,若在某个位置第一次检测到实心点,则说明该路径在这个位置被阻挡,其视野深度也就到这里为止,由此形成深度图。
较佳的,射线检测的具体步骤为:
1)设定起点(x 1,y 1,z 1)和终点(x 2,y 2,z 2);
2)计算获取路径在体素场景中经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)...(a n,b n,c n);
3)在GPU中,对M*N条路径同时进行计算,以索引到每个点在体素场景中对应的点,并判断其是否为实心点。
起点(x 1,y 1,z 1)和终点(x 2,y 2,z 2)之间的路径在同一条直线上,所谓射线检测即从起点(x 1,y 1,z 1)向终点(x 2,y 2,z 2)发出一条射线,两点之间行程一条射线路径。
在射线检测过程中,若检测到实心点,即体素坐标点的属性值为1,则说明该路径有阻挡、不通,若没有实心点,即体素坐标点的属性值为0,则可以说明该路径没有阻挡可以通行。通行可以理解为角色的空间意义上通行,还可以理解为信号的发送传递的畅通性。
例如,在游戏过程中,需要判断角色持枪是否可以击中远处一目标物,则检测起点为角色,终点为远处的目标物。若检测结果为无阻挡物,则表明在当前的瞄准路径下,正常开枪将可以击中该远处目标物;若检测结果为有阻挡物,则表明在当前的瞄准路径下,正常开枪也无法击中该远处目标物。
下面提供一对指定的角色和视野,使用体素数据得到分辨率为320*180的深度图的实例:
1)根据视野方向生成视锥,视锥的尽头的曲面在体素数据中对应了320*180个点;
2)以指定角色为起点,上述320*180个点为终点,定义320*180条检测射线;
3)将320*180对起点和终点,放入数组A(数组元素个数为320*180,每个数组元素包含起点和终点);
4)在GPU中,对数组A中的每对起点和终点同时进行计算,即320*180个线程并发,在每个线程都执行如下过程:
a、使用常规射线检测方法算出其在体素世界中需要经过的n个点(a1,b1,c1)、(a2,b2,c2)、...、(an,bn,cn);
b、由近及远的根据体素场景中的坐标索引到体素数据中相对应的点,并检查是否为实心点;
c、遇到的第m个点为实心点,则将m写入结果,全部为空心点则将n写入结果;
5)收集所有结果,获得深度图。
由于获得深度图的射线检测过程中需要检测在哪里不通,即得出深度,并停止检测,也可以理解为一像素图。
通过确定不同的游戏目标和视野,还可以提取不同的空间特征输入神经网络以进行空间物体分割、空间物体分类和行为强化学习。
另外,所述方法还包括:判定游戏对象的行为是否存在欺诈。
游戏对象的游戏处理过程在体素场景中进行,体素场景包括游戏对象和游戏对象的视野场景,游戏对象在视野场景中移动、执行决策行为。游戏对象的视野场景即该游戏对象视野范围内的场景元素的集合。该体素场景的数据由服务器端和客户端进行处理,服务器端和客户端无线或有线连接,客户端可以是手机、平板电脑、计算机等设备,服务器端为服务器或服务器群组。通常每个客户端对应操作一个游戏对象,但对于某些游戏类型,一个客户端也可以对应操作多个游戏对象。体素场景为由单元体素组成的三维空间,在该三维空间内,各个空间元素(如地形、建筑物、游戏角色、装备道具)通过单元体素的数据格式构建。
本发明的游戏对象的行为判定方法主要包括对游戏欺诈行为执行预防行为和中途阻止行为,具体的:
1)服务器端计算游戏对象在当前时间应该获取的视野场景,将视野场景实时下发至客户端;
2)基于游戏对象已发生的决策行为,判断决策行为是否满足在视野场景下的发生条件,若不满足发生条件,则认为存在欺诈行为;
3)记录游戏对象所对应的游戏账号和欺诈行为。
本实施例中有关预防欺诈行为的思路,将游戏对象的相关游戏数据(即玩家数据)全部放在服务器端,在服务器端计算每个游戏对象的游戏数据,然后将游戏数据下发至客户端,客户端所接收到的游戏数据即仅为该游戏对象应该接收的数据,而不存在该游戏对象不应该接收数据。
可以理解为(并不限定):一局游戏进行时,需要本游戏对象的角色数据、其他游戏对象的角色数据、视野场景数据、地图数据、配置数据等,若以上所有游戏数据都下发至 客户端,在每个客户端分别进行计算来实现游戏的进程,则会存在用户将上述数据进行恶意获取而破坏游戏规则的风险,如某个用户在客户端上通过外挂程序获得了透视功能,其游戏对象的视野内可以看到障碍物后面的其他游戏对象;而若将上述数据全部放在服务器端,在服务器端进行计算,将每个游戏对象实际所需要的数据下发至对应的客户端,不下发非必要数据,则可以避免用户对数据的恶意获取和利用,由此从源头避免了游戏欺诈行为的发生,即用户即便在客户端上使用外挂程序,也无法在视野内看到障碍物后面的游戏对象,因为障碍物后面的游戏对象的数据不会被下发至该客户端。
需要说明的是,本发明所提及的服务器向客户端下发的游戏数据为构建整体游戏场景所需的动态数据,而并不为整体游戏场景的全部数据,可以理解为:包括其他游戏对象、可拾取的游戏道具等能够发生动态变化的数据;不包括地形、植被等既有的不发生变化的场景数据。该动态数据的大小远远小于整体游戏场景的全部数据大小,这样可以减轻服务器与客户端之间的通信数据负担,支持游戏运行过程中的实时数据通信。
除了数据的获取,游戏欺诈行为还包括在游戏过程中利用正常数据进行不正常的决策操作的行为。例如正常操作情况下,本游戏对象和其他游戏对象之间存在遮挡物使得成功射击动作难以实现,或本游戏对象和其他游戏对象之间的距离过大使得成功射击动作难以实现;而发生欺诈行为的游戏过程中,本游戏对象却完成了对其他游戏对象的射击。
需要说明的是,本发明所提及的本游戏对象可以理解为当前被监测的客户端所操作的游戏对象,而其他游戏对象可以理解为其它客户端所操作的游戏对象、游戏服务器的AI游戏对象、或者非玩家对象等。而游戏对象可以是游戏角色,也可以是游戏角色以外的游戏元素,例如车辆,空投,装备等。
对于上述游戏欺诈行为,本发明还提出了中途阻止的反欺诈方法,即在游戏过程中,实时监测地游戏对象的决策行为,并判断当前游戏环境是否支持该决策行为,若不支持,则认为其存在游戏欺诈行为。通常,中途阻止的反欺诈方法也在服务器端进行,但并不排除在客户端也获取相关权限并配置有相关模块的情况下,中途阻止的反欺诈方法也可以在客户端进行。举例来说,可以将本游戏对象的视野场景作为游戏场景的一部分参与判断,视野场景即用户在客户端上操作该游戏对象能够观察到的游戏场景,若游戏对象针对游戏场景中可以看到的其他游戏对象执行游戏操作(比如射击),这样的游戏操作即决策行为是符合发生条件的。但是视野场景中若存在一障碍物,本游戏对象的视野场景中并不能观察到障碍物后面的其他游戏对象,这样就不支持对其他游戏对象的游戏操作,也就是不满足发生条件,如果发生了对无法观察到的其他游戏对象的游戏操作,就判定存在欺诈行为。
在判定为欺诈行为后,立即记录该游戏对象对应的游戏账号和该欺诈行为。进一步的,还可以立即终止该局游戏,或者根据欺诈行为的恶劣程度采取其他措施。
较佳的,视野场景的区域范围为游戏对象的视锥的区域范围,视锥的区域范围以内的区域认为是玩家(即客户)在客户端显示界面上应该看见的区域。这里,游戏对象可以是游戏中的游戏角色,该游戏角色具备拟人的视野,在客户端上能够以第一视角或第三视角观察到游戏角色的视野范围。
进一步的,在视锥的区域内,若通过射线检测到本游戏对象与其他游戏对象之间的路径存在阻挡,则不显示该其他游戏对象。
例如,本游戏对象视野范围内的草丛有其他游戏对象,则本游戏对象应当看不见该其他游戏对象。
又例如,本游戏对象在房屋外,而其他游戏对象在房屋内,则房屋外的本游戏对象应当看不见房屋内的其他游戏对象。
较佳的,游戏对象的决策行为包括射击行为、击打行为、治疗行为等。
关于射击行为,本发明通过射线检测判断本游戏对象与其他游戏对象之间的路径是否畅通,若通畅,且本游戏对象与其他游戏对象之间的距离满足射击要求,则认为满足射击条件,可以完成射击动作;反之,若本游戏对象与其他游戏对象之间的路径有阻挡或本游戏对象与其他游戏对象之间的距离不满足射击要求,则射击(击中)行为不应当发生。
通常,决策行为的判定是当该决策行为发生后被检测到才进行的判定,故,至需要在判定当前游戏环境不满足该决策行为之后,即可判定该玩家的该决策行为存在欺诈。
关于击打行为,通常也是通过本游戏对象与其他游戏对象之间的距离来判断其是否满足条件,若本游戏对象与其他游戏对象之间的距离超过该击打行为所支持的距离,则认为存在欺诈。
较佳的,基于体素场景的射线检测过程包括:
1)计算获取本游戏对象(x 1,y 1,z 1)和其他游戏对象(x 2,y 2,z 2)形成的路径在体素场景中经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)、...、(a n,b n,c n);即从起点(x 1,y 1,z 1)开始向终点(x 2,y 2,z 2)发出射线,起点与终点之间存在一线段,该线段上的点即需要经过的n个点。
2)在GPU中,同时对n个点进行计算:根据体素场景中的三维坐标(x、y、z)索引到每个点在体素场景中对应的点,并检查其是否为实心点;具体地,体素场景中每个坐标点都具备属性值,该属性值若为1,则表示该坐标点上存在物体或物体的一部分,如游戏 角色、建筑物等物体会占据大量坐标位置,这些被占据的位置的坐标点的属性均为1,也就是实心点;相应的,若该属性值为0,则表示该坐标点上没有物体,即为空旷场景,也就是空心点。
3)若检测到n个点中有实心点,则本游戏对象与其他游戏对象之间的路径存在阻挡;
4)若检测到n个点中没有实心点,则本游戏对象与其他游戏对象之间的路径畅通。
索引可以理解为查找,在检测判断n个点是否为实心点时,首先需要“到达”这n个点,索引即是“到达”这一步骤。可以理解为计算机处理过程中的专有步骤。
在体素数据世界中,1体现实心点,0体现空心点,一个物体在体素数据世界中的体现就是在若干特定位置的一系列的1。写入一个物体,即在体素数据世界中的若干特定位置写入1;擦除一个物体,即在体素数据世界中的若干特定位置写入0。
另外,本发明将体素数据应用于GPU射线检测,可以实现高速的多路径并发运算,可以理解为,在GPU中,使用体素数据进行多条射线检测的时间与检测一条射线的时间相同,故可以极大缩短多条射线检测的时间,由此大大减轻了服务器端的计算负载,使得每个玩家的游戏数据的计算统一放在服务器端进行称为可能。
在其他实施例中,体素场景还可为其他应用场景,在该其他应用场景中,场景元素将不仅仅包括地形、植被、建筑和室外摆件,或区别于地形、植被、建筑和室外摆件的其他场景元素。而对应的,对该其他场景元素的数据导出及体素数据的转换将可能采用区别于本实施例的方法,此处并不限制。
需要说明的是,并不意味着体素数据只可用于在GPU中进行计算,在CPU中,依然可以使用体素数据进行计算,只不过,在射线检测时无法并行计算,只能多个路径检测任务依次执行。
下面提供一GPU并行实现单个路径射线检测的实例:
1)使用射线检测方法算出起点(x 1,y 1,z 1)和终点(x 2,y 2,z 2)在体素场景中需要经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)、...、(a n,b n,c n);所谓射线检测方法,即从起点开始向终点发出射线,起点与终点之间存在一线段,该线段上的点即需要经过的n个点。
2)将n个点放入数组A(数组元素个数为n,每个数组元素包含体素场景坐标信息,即三维坐标(x、y、z)信息);
3)在GPU中,对数组A中的每个点同时进行计算,即n个线程并发,在每个线程都执行如下过程:
根据体素场景中的坐标索引体素数据中的点,并检测是否为实心点。若检测到实心点,则往结果中写1,即第一检测结果,表明该线程被阻挡;若没有检测到实心点,则往结果中写0,即第二检测结果,表明该线程所对应的路径无阻挡物可以通过。
下面提供一GPU并行实现多个射线检测的实例:
1)对于m个射线检测任务,有m对起点和终点,将他们放入数组B(数组元素个数为m,每个数组元素包含起点和终点);
2)在GPU中,对数组B中的每对起点和终点同时进行计算,即m个父线程并发,在每个线程都执行如下过程:
a.对起点(x 1,y 1,z 1)和终点(x 2,y 2,z 2)进行射线检测,
b.使用射线检测方法算出其在体素场景中需要经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)、...、(a n,b n,c n);
c.将n个点放入数组A’(数组元素个数为n,每个数组元素包含体素场景坐标信息,即三维坐标(x、y、z)信息);
d.在GPU中,对数组A中的每个点同时进行计算,即n个子线程并发,在每个子线程都执行如下过程(此时是针对每个父线程再并发n个子线程,即总共n*m个线程):根据体素场景中的坐标索引体素数据中的点,并检测是否为实心点;若检测到实心点,则往结果中写1,表明该子线程所对应的路径被阻挡;若没有检测到实心点,则往结果中写0,表明该子线程所对应的路径无阻挡物可以通过;
3)收集m条射线的检测结果。
本发明在处理器RTX3090和AMD threadriper 3990X上测得:在同时进行100万条以及以上射线检测时,GPU的检测速度为单个CPU核心的约550倍,运算速度获得明显的提升。
在体素场景中通常需要进行动态规划,用于计算导航任务,输出结果为路径。在一个动态规划任务中,包括若干个待探索的节点,需要说明的是,该节点为任务过程意义上的节点,不代表数据中的具体指代。在某些实施例中,节点可以是某个具体的体素场景坐标点。对动态规划来说,节点的探索是有先后的,探索节点时是否需要射线检测、以及对哪些目标进行射线检测都是未知的,则形成了一个时间序列任务。本发明由于使用GPU可以进行并行化时间分散任务:
1)预估动态规划将会用到的检测区域;
2)对检测区域中的所有节点、以及每个节点需要检测射线的所有检测目标都进行射 线检测,假设有x个检测目标,y个节点,则有x*y个路径,也就是共有x*y个并行检测任务;其中,检测目标可以是一个体素场景坐标点,如游戏世界中某个目标对象或场景建筑外表上的一个坐标点。
3)保存检测结果;
4)当动态规划算法执行时,只需要查询已经运算好的检测结果即可。
由于存储指定检测区域的射线检测结果所占据存储空间很小,一个结果最多占用一个字节,1MB即可存储100万条检测结果,故本发明的动态规划算法任务的执行效率较高,只需要读取的时间即可。
较佳的,还存在需要对多个检测主体同时进行动态规划,在该检测主体中包括若干个检测目标,即同时对不同检测目标所对应的检测区域的x*y个路径进行射线检测,并保存检测结果,形成以检测区域划分的检测结果表;对一检测区域进行动态规划时、则从检测结果表中调取所述检测区域的检测结果即可。
例如,需要对区域A、区域B、区域C和区域D进行动态规划,则同时对每个区域的x*y个路径进行射线检测,检测时间仅为单个射线检测时间t。
将检测结果保存,区域A的检测结果为a,区域B的检测结果为b,区域C的检测结果为c,区域D的检测结果为d,形成A-a、B-b、C-c、D-d的检测结果表,当对区域A进行动态规划时,则调取结果a即可。
不论是存在多个检测目标,还是多个检测主体,射线检测时间都仅为单个射线检测时间t。后续在动态规划的实际过程中进行及时读取即可,运算速度十分快速。
较佳的,游戏对象在游戏过程中常常有位移,即不同时间该游戏对象的位置不同,该游戏对象在此被称为动态的游戏对象,需要按照一预设周期刷新视野场景中动态的游戏对象的体素数据。在服务器端计算游戏对象在当前时间应该获取的视野场景包括:
1)在视野场景中,获取所有动态的游戏对象的类型及每一类型的动态的游戏对象的轮廓点相对于该游戏对象的中心点的偏移坐标;
2)根据最近一预设周期结束时的动态的游戏对象的中心点坐标和轮廓点的偏移坐标计算获取刷新后的轮廓点的更新坐标;
3)在动态的游戏对象刷新后的所有轮廓点的更新坐标写入体素数据,并擦除动态的游戏对象刷新前的所有轮廓点的原始位置的体素数据,完成视野场景内所有游戏对象的动态刷新,从而得到计算刷新后的视野场景。
轮廓点坐标即组成对象的轮廓的坐标,对象的写入体素数据即在若干个轮廓点的位置 写入1。
例如,一个空投在游戏中的中心点坐标为(10.10.30),该空投的其中一个轮廓点坐标为(0.20.50),则该轮廓点相对于中心点的偏移坐标则为(-10.10.20)。刷新后动态的游戏对象的中心点坐标为(10.10.25),则该轮廓点刷新后的坐标为(0.20.45)。在刷新后的轮廓点的位置(0.20.45)写入1,并在原来位置(0.20.50)写入0。组成空投的若干个轮廓点同时刷新,则实现了空投位置的刷新。
通过使用体素数据,即使是全局擦写,在GPU上并行实现,也可以在10微秒(0.01毫秒)级别高速完成体素场景的实时刷新。在本发明其他实施方式中,也可按预设周期对整个游戏内的所有动态游戏对象进行刷新,并按照不同游戏对象的视野场景分别下发相应的数据给不同客户端。
动态的游戏对象可以是游戏角色、车辆、空投等一切可能发生移动的游戏对象。
较佳的,除了判断射击行为、击打行为和治疗行为等决策行为,还可以通过判断游戏对象所行走的路线、以及通行方式是否为合规来判定欺诈行为。具体的:
1)计算获取游戏对象与目的地之间的导航路径和通行方式;导航路径的数量为一条或若干条,通行方式的数量为一种或若干种;
2)判断导航路径是否包括游戏对象的实际行走路径,并判断该实际行走路径上所支持的通行方式是否包含游戏对象实际所采用的通行方式,若否则认为存在该游戏对象欺诈行为。
例如,本游戏对象在平地上,所要去的目的地为悬崖上的房屋,若走直线(即导航路径),则需要通过装备进行喷射(飞行)(即通行方式)才可以到达;若绕路走曲线(即导航路径),则可以通过行走(即通行方式)到达。即代表:直线路径仅支持喷射(飞行)的通行方式,曲线路径支持行走、跳跃、喷射(飞行)等通行方式。若检测到该游戏对象通过行走直线达到该房屋,且不具备喷射(飞行)的装备,则可以判定该游戏对象存在欺诈行为。
又例如,本游戏对象在在河边,所要去的目的地为河对岸,游戏设定该河的较深,仅支持乘坐水上交通工具、或喷射(飞行)至河对岸,而不支持角色在不使用任何水上交通工具、不使用任何装备的情况下通行,若检测到该游戏角色在不使用任何水上交通工具、不使用任何装备的情况下通过和该河流,则认为存在欺诈行为。
再例如,设定游戏对象在移动一定距离后将会消耗s值的能量值,而实际检测到本游戏对象在移动了一定距离后所消耗的能量值小于s,则认为该游戏对象存在欺诈行为。
较佳的,计算获取游戏对象与目的地之间的导航路径和通行方式是基于体素场景的层数据和层数据与层数据之间的连接数据,具体包括:
1)在体素场景的x,y,z三维空间中,取z=z′的一基础平面,基础平面上存在若干个基础点(x,y,z′);以基础点为底点,以该基础点的z坐标为高度,形成若干根基础柱,每根基础柱上都存在L个元素点(x,y,z);
2)在GPU中并行地访问所有基础柱,并在每个并行的线程中遍历基础柱上的每个元素点,并根据体素数据索引到其对应的体素场景坐标点,判断该体素场景坐标点是空心点或实心点;
3)收集每根基础柱上的连续空心点段,若连续空心点段的高度大于等于第一预设高度,则定义该连续空心点段为一个体素层;
4)计算获取游戏对象当前所位于的体素层与相邻基础柱上的每个体素层之间的位置关系,从而获取游戏对象当前所位于的体素层与相邻基础柱上的每个体素层之间的通行方式。
本发明还公开了一种基于体素数据的数据处理系统,游戏场景的原始数据构成像素场景,像素场景包括若干种不同数据类型的场景元素,并且所述系统包括:
导出模块,分别导出若干种场景元素的原始数据;
转换模块,设定期望的单元体素的边长,结合单元体素的边长,分别将若干种场景元素的原始数据转换为体素数据,体素数据在体素场景中表现为体素模块;
拼接模块,根据所有场景元素的体素模块在像素场景中的相对位置,拼接所有场景元素的体素模块,获得体素场景。
该系统包括硬件结构和计算机可读存储介质,上述功能模块可以集成在硬件结构上,也可集成在计算机可读存储介质上,此处并不限制。且上述功能模块的连接关系可以是有形连接,也可以是跨区域的无形连接,此处并不限制。另外,该系统与相应方法实施例属于同一构思,其具体实现过程详见对应方法实施例,这里不再赘述。
本发明还公开了一种基于体素数据的数据处理服务器,游戏场景的原始数据构成像素场景,像素场景包括若干种不同数据类型的场景元素,并且所述服务器包括:
导出模块,分别导出若干种场景元素的原始数据;
转换模块,设定期望的单元体素的边长,结合单元体素的边长,分别将若干种场景元素的原始数据转换为体素数据,体素数据在体素场景中表现为体素模块;
拼接模块,根据所有场景元素的体素模块在像素场景中的相对位置,拼接所有场景元 素的体素模块,获得体素场景。
另外,该服务器与相应方法实施例属于同一构思,其具体实现过程详见对应方法实施例,这里不再赘述。
本发明还公开了一种存储基于体素数据的数据处理指令的计算机可读存储介质,游戏场景的原始数据构成像素场景,像素场景包括若干种不同数据类型的场景元素,并且指令在被执行时实施以下步骤:
分别导出若干种场景元素的原始数据;
设定期望的单元体素的边长,结合单元体素的边长,分别将若干种场景元素的原始数据转换为体素数据,体素数据在体素场景中表现为体素模块;
根据所有场景元素的体素模块在像素场景中的相对位置,拼接所有场景元素的体素模块,获得体素场景。
该计算机可读存储介质可以集成在硬件上,当硬件运行时,该计算机可读存储介质可以被支持读取运行。
另外,该计算机可读存储介质与相应方法实施例属于同一构思,其具体实现过程详见对应方法实施例,这里不再赘述。
本发明还公开了一种计算机程序产品,包括计算机可执行指令,指令被处理器执行以实施以下步骤:
分别导出若干种场景元素的原始数据;
设定期望的单元体素的边长,结合单元体素的边长,分别将若干种场景元素的原始数据转换为体素数据,体素数据在体素场景中表现为体素模块;
根据所有场景元素的体素模块在像素场景中的相对位置,拼接所有场景元素的体素模块,获得体素场景。
另外,该计算机程序产品与相应方法实施例属于同一构思,其具体实现过程详见对应方法实施例,这里不再赘述。
应当注意的是,本发明的实施例有较佳的实施性,且并非对本发明作任何形式的限制,任何熟悉该领域的技术人员可能利用上述揭示的技术内容变更或修饰为等同的有效实施例,但凡未脱离本发明技术方案的内容,依据本发明的技术实质对以上实施例所作的任何修改或等同变化及修饰,均仍属于本发明技术方案的范围内。

Claims (22)

  1. 一种基于体素数据的数据处理方法,其特征在于,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且所述方法包括:
    分别导出若干种所述场景元素的原始数据;
    设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
    根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
  2. 根据权利要求1所述的方法,其特征在于,所述场景元素包括地形、植被、建筑和室外摆件中的至少一种;
    所述分别导出若干种所述场景元素的原始数据包括:
    导出所述室外摆件与所述建筑的3D模型文件格式数据和坐标信息;
    导出所述植被的逗号分隔值文件格式数据;
    通过深度相机正交拍摄以导出地形的图片,该图片中包含地形的表面高度数据。
  3. 根据权利要求2所述的方法,其特征在于,所述设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据包括:
    通过开源库中的“读写三角网格”功能和“从三角网格创建”功能将所述室外摆件与所述建筑的所述3D模型文件格式数据转换为体素数据;
    获取所述植被的碰撞体大小,结合所述单元体素的边长,计算获取所述植被在所述体素场景中需要占用的体素个数及体素形状;
    根据存储有所述地形的所述表面高度数据的所述图片,按照单元体素的边长逐点采样以将地形的所述表面高度数据转换为体素数据。
  4. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    在所述体素场景内裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征。
  5. 根据权利要求4所述的方法,其特征在于,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
    以所述目标对象为中心,裁剪获取所述目标对象的周围空间范围内的体素正方体,将所述体素正方体以三维张量的形式作为神经网络的输入,以获取所述目标对象周围的空间特征。
  6. 根据权利要求4所述的方法,其特征在于,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
    在所述目标对象的前方空间范围内,裁剪获取一第一体素长方体,将所述第一体素长方体以三维张量的形式作为神经网络的输入,以获取所述目标对象前方的空间特征,
    在所述目标对象的前方空间范围内,裁剪获取一第二体素长方体,其中所述第二体素长方体的长度远大于所述第一体素长方体的长度,将所述第二体素长方体以三维张量的形式作为神经网络的输入,以获取所述目标对象前方远处的空间特征。
  7. 根据权利要求4所述的方法,其特征在于,所述裁剪获取目标对象的目标空间范围内的体素区域,将所述体素区域以三维张量的形式作为神经网络的输入,以获取所述目标对象的空间特征包括:
    对指定的目标对象和视野,使用所述体素场景中的体素数据,通过射线检测获取一分辨率为M*N的深度图,将所述深度图作为神经网络的输入,以获取所述目标对象的空间特征。
  8. 根据权利要求7所述的方法,其特征在于,所述对指定的目标对象和视野,使用所述体素场景中的体素数据,通过射线检测获取一分辨率为M*N的深度图包括:
    在所述视野的方向生成一视锥,所述视锥的尽头为曲面,在所述体素场景中,所述曲面包括M*N个点;
    以所述目标对象为起点、所述M*N个点为终点,形成M*N条路径,对M*N条路径沿自起点至终点的方向进行射线检测,直到检测到路径上的实心点;
    所述M*N条路径的检测结果构成所述深度图。
  9. 根据权利要求8所述的方法,其特征在于,所述对M*N条路径沿自起点至终点的方向进行射线检测,直到检测到路径上的实心点包括:
    所述路径包括起点和终点;
    计算获取所述路径在所述体素场景中经过的n个点;
    在GPU中,对所述M*N条路径同时进行计算,以索引到每个点在所述体素场景中对应的点。
  10. 根据权利要求1所述的方法,其特征在于,所述方法还包括:
    判定游戏对象的行为是否存在欺诈,其中,所述游戏对象的游戏处理过程在所述体素场景中进行,所述体素场景包括所述游戏对象和所述游戏对象的视野场景,每个客户端对应一个或多个所述游戏对象,并且所述判定游戏对象的行为是否存在欺诈包括:
    服务器端计算所述游戏对象在当前时间应该获取的所述视野场景,将所述视野场景实时下发至所述客户端;
    基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件,若不满足发生条件,则认为存在欺诈行为;
    记录所述游戏对象所对应的游戏账号和所述欺诈行为。
  11. 根据权利要求10所述的方法,其特征在于,所述视野场景的区域范围为游戏对象的视锥的区域范围;
    在所述视锥的区域范围内,若通过射线检测判断本游戏对象与其他游戏对象之间的路径存在阻挡,则不显示所述其他游戏对象。
  12. 根据权利要求10所述的方法,其特征在于,所述基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件包括:
    通过射线检测判断本游戏对象与其他游戏对象之间的路径是否畅通,若畅通,则所述决策行为满足发生条件,若不畅通,则所述决策行为不满足发生条件。
  13. 根据权利要求10所述的方法,其特征在于,所述计算所述游戏对象在当前时间应该获取的所述视野场景包括:
    在所述视野场景中,获取所有动态的游戏对象的类型及每一类型的动态的游戏对象的轮廓点相对于该游戏对象的中心点的偏移坐标;
    根据最近一预设周期结束时的动态的游戏对象的中心点坐标和所述轮廓点的偏移坐标计算获取刷新后的所述轮廓点的更新坐标;
    在动态的游戏对象刷新后的所有轮廓点的更新坐标写入体素数据,并擦除动态的游戏对象刷新前的所有轮廓点的原始位置的体素数据,完成视野场景内所有游戏对象的动态刷新,从而得到计算刷新后的视野场景。
  14. 根据权利要求10所述的方法,其特征在于,所述基于所述游戏对象已发生的决策行为,判断所述决策行为是否满足在所述视野场景下的发生条件,若不满足发生条件,则认为存在欺诈行为包括:
    计算获取所述游戏对象与目的地之间的导航路径和通行方式;所述导航路径的数量为一条或若干条,所述通行方式的数量为一种或若干种;
    判断所述导航路径是否包括所述游戏对象的实际行走路径,并判断所述实际行走路径上所支持的所述通行方式是否包含所述游戏对象实际所采用的通行方式,若否则认为存在欺诈行为。
  15. 根据权利要求14所述的方法,其特征在于,所述计算获取所述游戏对象与目的地之间的导航路径和通行方式包括:
    在所述体素场景的x,y,z三维空间中,取z=z′的一基础平面,所述基础平面上存在若干个基础点;以所述基础点为底点,以该基础点的z坐标为高度,形成若干根基础柱,每根所述基础柱上都存在L个元素点;
    在GPU中并行地访问所有所述基础柱,并在每个并行的线程中遍历所述基础柱上的每个所述元素点,并根据体素数据索引到其对应的体素场景坐标点,判断该体素场景坐标点是空心点或实心点;
    收集每根所述基础柱上的连续空心点段,若所述连续空心点段的高度大于等于第一预设高度,则定义该连续空心点段为一个体素层;
    计算获取游戏对象当前所位于的所述体素层与相邻所述基础柱上的每个所述体素层之间的位置关系,从而获取游戏对象当前所位于的所述体素层与相邻所述基础柱上的每个所述体素层之间的通行方式。
  16. 根据权利要求1所述的方法,其特征在于,射线检测的判断步骤包括:
    计算获取本游戏对象(x 1,y 1,z 1)和其他游戏对象(x 2,y 2,z 2)形成的射线路径在所述体素场景中经过的n个点(a 1,b 1,c 1)、(a 2,b 2,c 2)、...、(a n,b n,c n);
    在GPU中,同时对所述n个点进行计算:根据所述体素场景中的坐标索引到所述体素场景中n个点,并检查其是否为实心点;
    若检测到所述n个点中有实心点,则所述本游戏对象与其他游戏对象之间的路径存在阻挡;
    若检测到所述n个点中没有实心点,则本游戏对象与其他游戏对象之间的路径畅通。
  17. 根据权利要求16所述的方法,其特征在于,所述射线检测的判断步骤还包括:在GPU中对m条路径进行射线检测:
    所述m条路径包括m个起点和m个终点;同时计算获取每条所述路径在体素场景中经过的n个点;
    同时对m条路径中的所述n个点进行计算:根据体素场景中的坐标索引每个点在所述体素场景中对应的点,并检测其是否为实心点;
    若检测到所述n个点中有实心点,则所述路径存在阻挡;否则所述路径畅通,以此获得m条路径的检测结果。
  18. 根据权利要求17所述的方法,其特征在于,还包括:
    一检测区域包括x个检测目标和y个节点,对所述检测区域进行动态规划,包括:
    同时对所述x个检测目标和y个节点所形成的x*y个路径进行射线检测,并保存检测结果;
    对所述检测区域在不同时间调取不同路径的所述检测结果以进行所述动态规划。
  19. 根据权利要求18所述的方法,其特征在于,还包括:
    对多个检测区域进行动态规划,包括:
    同时对不同所述检测区域的x*y个路径进行射线检测,并保存检测结果,形成以所述检测 区域划分的检测结果表;
    对一检测区域进行所述动态规划时、则从所述检测结果表中调取所述检测区域的所述检测结果。
  20. 一种基于体素数据的数据处理服务器,其特征在于,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且所述服务器包括:
    导出模块,分别导出若干种所述场景元素的原始数据;
    转换模块,设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
    拼接模块,根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
  21. 一种存储基于体素数据的数据处理指令的计算机可读存储介质,其特征在于,游戏场景的原始数据构成像素场景,所述像素场景包括若干种不同数据类型的场景元素,并且所述指令在被执行时实施以下步骤:
    分别导出若干种所述场景元素的原始数据;
    设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
    根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
  22. 一种计算机程序产品,包括计算机可执行指令,其特征在于,所述指令被处理器执行以实施以下步骤:
    分别导出若干种所述场景元素的原始数据;
    设定期望的单元体素的边长,结合所述单元体素的边长,分别将若干种所述场景元素的原始数据转换为体素数据,所述体素数据在所述体素场景中表现为体素模块;
    根据所有所述场景元素的体素模块在所述像素场景中的相对位置,拼接所有所述场景元素的体素模块,获得所述体素场景。
PCT/CN2022/122494 2021-09-30 2022-09-29 基于体素数据的数据处理方法、服务器、介质及计算机程序产品 WO2023051671A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202280020323.XA CN116963811A (zh) 2021-09-30 2022-09-29 基于体素数据的数据处理方法、服务器、介质及计算机程序产品

Applications Claiming Priority (6)

Application Number Priority Date Filing Date Title
CN202111163612.9 2021-09-30
CN202111160564.8 2021-09-30
CN202111160592.XA CN113877211A (zh) 2021-09-30 2021-09-30 基于体素数据的游戏对象的行为判定方法、系统及计算机可读存储介质
CN202111163612.9A CN113786617A (zh) 2021-09-30 2021-09-30 应用体素数据进行射线检测的方法、系统、服务器及计算机可读存储介质
CN202111160592.X 2021-09-30
CN202111160564.8A CN113877210A (zh) 2021-09-30 2021-09-30 游戏场景的转换方法、系统、服务器及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2023051671A1 true WO2023051671A1 (zh) 2023-04-06

Family

ID=85781334

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/122494 WO2023051671A1 (zh) 2021-09-30 2022-09-29 基于体素数据的数据处理方法、服务器、介质及计算机程序产品

Country Status (2)

Country Link
CN (1) CN116963811A (zh)
WO (1) WO2023051671A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116772886A (zh) * 2023-08-17 2023-09-19 腾讯科技(深圳)有限公司 虚拟场景中虚拟角色的导航方法、装置、设备及存储介质
CN116883612A (zh) * 2023-09-08 2023-10-13 东华理工大学南昌校区 一种三维场景模型生成方法及系统
CN117745728A (zh) * 2024-02-21 2024-03-22 法奥意威(苏州)机器人系统有限公司 点云平面检测方法、装置、电子设备和可读存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200226820A1 (en) * 2019-01-11 2020-07-16 Microsoft Technology Licensing, Llc Object permanence in surface reconstruction
CN111714895A (zh) * 2020-06-29 2020-09-29 天津亚克互动科技有限公司 游戏数据的处理方法及装置、存储介质、计算机设备
CN112740269A (zh) * 2020-05-13 2021-04-30 华为技术有限公司 一种目标检测方法及装置
CN112973127A (zh) * 2021-03-17 2021-06-18 北京畅游创想软件技术有限公司 一种游戏3d场景编辑方法及装置
CN113786617A (zh) * 2021-09-30 2021-12-14 上海莉莉丝计算机技术有限公司 应用体素数据进行射线检测的方法、系统、服务器及计算机可读存储介质
CN113877211A (zh) * 2021-09-30 2022-01-04 上海莉莉丝计算机技术有限公司 基于体素数据的游戏对象的行为判定方法、系统及计算机可读存储介质
CN113877210A (zh) * 2021-09-30 2022-01-04 上海莉莉丝计算机技术有限公司 游戏场景的转换方法、系统、服务器及计算机可读存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200226820A1 (en) * 2019-01-11 2020-07-16 Microsoft Technology Licensing, Llc Object permanence in surface reconstruction
CN112740269A (zh) * 2020-05-13 2021-04-30 华为技术有限公司 一种目标检测方法及装置
CN111714895A (zh) * 2020-06-29 2020-09-29 天津亚克互动科技有限公司 游戏数据的处理方法及装置、存储介质、计算机设备
CN112973127A (zh) * 2021-03-17 2021-06-18 北京畅游创想软件技术有限公司 一种游戏3d场景编辑方法及装置
CN113786617A (zh) * 2021-09-30 2021-12-14 上海莉莉丝计算机技术有限公司 应用体素数据进行射线检测的方法、系统、服务器及计算机可读存储介质
CN113877211A (zh) * 2021-09-30 2022-01-04 上海莉莉丝计算机技术有限公司 基于体素数据的游戏对象的行为判定方法、系统及计算机可读存储介质
CN113877210A (zh) * 2021-09-30 2022-01-04 上海莉莉丝计算机技术有限公司 游戏场景的转换方法、系统、服务器及计算机可读存储介质

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116772886A (zh) * 2023-08-17 2023-09-19 腾讯科技(深圳)有限公司 虚拟场景中虚拟角色的导航方法、装置、设备及存储介质
CN116772886B (zh) * 2023-08-17 2023-10-20 腾讯科技(深圳)有限公司 虚拟场景中虚拟角色的导航方法、装置、设备及存储介质
CN116883612A (zh) * 2023-09-08 2023-10-13 东华理工大学南昌校区 一种三维场景模型生成方法及系统
CN116883612B (zh) * 2023-09-08 2023-11-21 东华理工大学南昌校区 一种三维场景模型生成方法及系统
CN117745728A (zh) * 2024-02-21 2024-03-22 法奥意威(苏州)机器人系统有限公司 点云平面检测方法、装置、电子设备和可读存储介质

Also Published As

Publication number Publication date
CN116963811A (zh) 2023-10-27

Similar Documents

Publication Publication Date Title
WO2023051671A1 (zh) 基于体素数据的数据处理方法、服务器、介质及计算机程序产品
US11321924B2 (en) Caching and updating of dense 3D reconstruction data
US11238644B2 (en) Image processing method and apparatus, storage medium, and computer device
US10349033B2 (en) Three-dimensional map generating and displaying apparatus and method
CA2818695C (en) Guided navigation through geo-located panoramas
US20230201718A1 (en) Method, device, and storage medium for displaying azimuth in virtual scene
JP7361502B2 (ja) 立体視効果を調節するためのシステムおよび方法
JP7432595B2 (ja) 連携仮想インターフェース
US11853080B2 (en) Spray operation method and device for unmanned aerial vehicle
US20230147759A1 (en) Viewpoint dependent brick selection for fast volumetric reconstruction
WO2023142609A1 (zh) 虚拟场景中的对象处理方法、装置、设备、存储介质及程序产品
CN113877210A (zh) 游戏场景的转换方法、系统、服务器及计算机可读存储介质
WO2023061474A1 (zh) 基于体素数据的数据处理方法、服务器、介质及计算机程序产品
CN113877211A (zh) 基于体素数据的游戏对象的行为判定方法、系统及计算机可读存储介质
CN112891940B (zh) 图像数据处理方法及装置、存储介质、计算机设备
CN113786617A (zh) 应用体素数据进行射线检测的方法、系统、服务器及计算机可读存储介质
CN115631320B (zh) 预计算单元格显示方法、预计算单元格生成方法及装置
CN116109756A (zh) 光线追踪方法、装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22875054

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 202280020323.X

Country of ref document: CN