Method, apparatus, device, and storage medium for processing game images

Abstract

The present application provides a method, apparatus, device, and storage medium for processing game images. The method includes: obtaining the feature information of a second virtual character that is within the field of view of a first virtual character in a game scene image and is closest to the midline position of the field of view; obtaining the skill landing position of the first virtual character in the game scene image; training a neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position, where the feature information of the second virtual character and the skill landing position are used as label data for the game scene image, thereby solving the technical problem of how to introduce artificial intelligence players in asymmetric confrontation competitive games.

Description

Technical Field

[0001] The present application relates to the field of games, and in particular, to a method, device, equipment and storage medium for processing game images. Background Art

[0002] An asymmetrical battle arena (ABA) game refers to a game in which the two sides in the confrontation are unequal in terms of the number of people, resources, information, and rules during the game process. For example, the hide-and-seek game.

[0003] Technicians in the field of online games have proposed the idea of introducing artificial intelligence (AI) players in ABA games, so that novice players can become familiar with the game mechanisms in ABA games through the confrontation between AI players and novice players. However, technicians in this field have not further pointed out how to implement the technical solution for introducing AI players in ABA games, that is, they have not further pointed out how to implement the technical solution for the game behaviors of AI players.

[0004] Therefore, how to introduce AI players in ABA games has become a technical problem to be solved urgently. Summary of the Invention

[0005] In the prior art, there are few technologies for introducing AI players in online games in the ABA mode, and it is relatively difficult to implement. Instead, the technologies for introducing AI players in symmetrical battle arena games are relatively mature. Introducing AI players in symmetrical battle arena games is usually achieved by training AI players using expert rule systems such as finite state machines or behavior trees. If the technology for introducing AI players in symmetrical battle arena games is used to introduce AI players in online games in the ABA mode, due to the limitations of the expert rule system, the trained AI players cannot show diversification in ABA games, resulting in real players being very easy to identify the AI players, reducing the game experience of real players, and thus resulting in a lower degree of anthropomorphism of the trained AI players. Therefore, there is a problem in the current prior art of how to introduce AI players in ABA games and ensure a relatively high degree of anthropomorphism of the introduced AI players.

[0006] The embodiments of the present application provide a method, device, equipment and storage medium for processing game images, which are used to solve the problem of how to introduce AI players in ABA games and ensure a relatively high degree of anthropomorphism of the introduced AI players.

[0007] In a first aspect, an embodiment of the present application provides a method for processing game images. The method includes: obtaining feature information of a second virtual character that is within the field of view of a first virtual character in a game scene image and is closest to the midline position of the field of view; obtaining the skill landing position of the first virtual character in the game scene image; training a neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position, where the feature information of the second virtual character and the skill landing position are used as label data of the game scene image.

[0008] The technical solution provided by the embodiment of the present application may include the following beneficial effects: By obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained through supervised learning can have a high degree of anthropomorphism, thereby achieving the purpose of introducing an AI player in the ABA game and ensuring a relatively high degree of anthropomorphism of the AI player.

[0009] In a possible implementation manner, obtaining the feature information of the second virtual character that is within the field of view of the first virtual character in the game scene image and is closest to the midline position of the field of view includes: determining the target virtual character in the game scene image; obtaining multiple target points on the target virtual character; performing ray detection on the first virtual character and the multiple target points to determine whether there is a collision body between the first virtual character and the target point; if there is no collision body between the first virtual character and at least one target point, determining that the target virtual character is within the field of view; obtaining the distance between each target virtual character within the field of view and the midline position of the field of view; determining the target virtual character with the smallest distance from the midline position of the field of view as the second virtual character; and obtaining the feature information of the second virtual character.

[0010] The technical solution provided by the embodiment of the present application may include the following beneficial effects: The second virtual character selected by the first virtual character must be within its field of view. By determining all the targets within the field of view of the first virtual object, the AI player obtained in this way will not attack targets outside the field of view, thus avoiding anti-human behavior. Moreover, the first virtual character will not simultaneously attack multiple second virtual characters at different positions that are far away. Therefore, it is necessary to determine a second virtual character within the field of view. For a real player, the second virtual character to be selected is generally the target located in the middle of its field of view. Therefore, the second virtual character to be selected by the first virtual character should be the target located in the middle of its field of view, thereby improving the anthropomorphism of the AI player and avoiding anti-human behavior of the trained AI player.

[0011] In a possible implementation, the method further includes: if there are collision bodies between the first virtual character and multiple target points, it is determined that the target virtual character is not within the field of view.

[0012] The technical solution provided by the embodiments of the present application may include the following beneficial effects: For a second virtual character that is not within the field of view of the first virtual character, it can be ignored and not processed, so as to avoid the problem of anti-human behavior caused by the AI player attacking a target outside the field of view, thereby improving the anthropomorphic degree of the AI player.

[0013] In a possible implementation, obtaining the skill landing position of the first virtual character in the game scene image includes: dividing the game scene map where the game scene image is located into multiple regions, where the shapes and sizes of the multiple regions are the same; determining the landing region of the game skill of the first virtual character in the multiple regions; and determining the landing region as the skill landing position.

[0014] The technical solution provided by the embodiments of the present application may include the following beneficial effects: If the landing position of the game skill of the first virtual character in the game scene map is directly used as its skill landing position, since there are infinitely many landing positions in the game scene map, training the neural network model will evolve into a regression model, thus increasing the training difficulty of the neural network model. However, by dividing the game scene map into multiple regions, since the divided regions are limited, using the landing region of the game skill as the skill landing position will evolve the training of the neural network model into a classification model, which can reduce the training difficulty of the neural network model, thereby reducing the training time of the AI player and improving the update speed of the game.

[0015] In a possible implementation, training the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position includes: training the neural network model according to digital features and image features, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located.

[0016] The technical solution provided by the embodiments of the present application may include the following beneficial effects: In addition to the feature information of the second virtual character and the skill landing position, the digital features also include other feature information in the game, such as the feature information of the first virtual character, environmental features, etc., so that the trained AI player can be more in line with real players, thereby improving the anthropomorphic degree of the AI player.

[0017] In a possible implementation, training a neural network model according to digital features and image features includes: one-dimensionally processing the image features to obtain a plurality of one-dimensional arrays; fusing the plurality of one-dimensional arrays and the digital features to obtain a target array; inputting the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

[0018] The technical solution provided by the embodiments of the present application may include the following beneficial effects: By training the neural network model in this way, the anthropomorphic degree of the trained AI player can be improved, and on the premise of achieving the same level of AI players, this technical solution can reduce the training time compared with other reinforcement learning, thereby improving the game update speed.

[0019] In a possible implementation, before one-dimensionally processing the image features, the method further includes: processing the image features using a convolutional neural network.

[0020] The technical solution provided by the embodiments of the present application may include the following beneficial effects: Processing the image features through a convolutional neural network (Convolutional Neural Network, abbreviated as: CNN) can pay more attention to some local features in the image, thereby reducing the computational amount of image processing, reducing the training time of the AI player, and further improving the game update speed.

[0021] In a possible implementation, the digital features further include: the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map.

[0022] The technical solution provided by the embodiments of the present application may include the following beneficial effects: In addition to including the feature information of the second virtual character and the skill landing position, the digital features also include the feature information in other games, such as the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map, etc. In this way, the trained AI player can be more in line with real players, thereby improving the anthropomorphic degree of the AI player.

[0023] In a second aspect, the embodiments of the present application provide a method for processing game images. A graphical user interface is provided through a terminal device. The graphical user interface includes a game scene image, and the game scene image includes a plurality of virtual characters. The method includes: inputting the game scene image into the trained neural network model for processing to obtain the feature information of the second virtual character among the plurality of virtual characters, and the skill landing position of the first virtual character among the plurality of virtual characters, where the second virtual character is within the field of view of the first virtual character and is the closest to the midline position of the field of view.

[0024] The technical solution provided by the embodiment of the present application may include the following beneficial effects: After training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a high degree of anthropomorphism can be obtained. The AI player can play the game in the same way as a real player selects a second virtual object and releases a game skill during the game process, so as to intelligently select a second virtual character and release the game skill at a reasonable skill landing position. In this way, when a novice player battles with the AI player, it is not easy to identify the AI player, which improves the anthropomorphism degree of the AI player and also improves the game experience of the novice player.

[0025] In a possible implementation manner, before inputting the game scene image into the trained neural network model for processing, the method further includes: training the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position to obtain the trained neural network model, where the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image.

[0026] The technical solution provided by the embodiment of the present application may include the following beneficial effects: By obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained by supervised learning can have the effect of a high degree of anthropomorphism, so as to realize the introduction of an AI player in the ABA game and ensure that the anthropomorphism degree of the AI player is relatively high.

[0027] In a third aspect, the embodiment of the present application provides a game image processing device, including: a first acquisition module, configured to acquire the feature information of the second virtual character that is within the field of view of the first virtual character and is closest to the midline position of the field of view in the game scene image; a second acquisition module, configured to acquire the skill landing position of the first virtual character in the game scene image; a training module, configured to train the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position, where the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image.

[0028] The technical solution provided by the embodiment of the present application may include the following beneficial effects: By obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained by supervised learning can have the effect of a high degree of anthropomorphism, so as to realize the introduction of an AI player in the ABA game and ensure that the anthropomorphism degree of the AI player is relatively high.

[0029] Fourth aspect, an embodiment of the present application provides a processing of game images, including: a processing module and a display module. The display module is used to display a graphical user interface, and the graphical user interface includes a game scene image, and the game scene image includes a plurality of virtual characters; the processing module is used to input the game scene image into a trained neural network model for processing, to obtain the feature information of a second virtual character among the plurality of virtual characters, and the skill landing position of a first virtual character among the plurality of virtual characters, wherein the second virtual character is within the field of view of the first virtual character and is closest to the midline position of the field of view.

[0030] The technical solution provided by the embodiment of the present application may include the following beneficial effects: After training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a high degree of anthropomorphism can be obtained. This AI player can play the game in the same way as a real player selects a second virtual object and releases a game skill during the game process, so as to intelligently select the second virtual character and release the game skill at a reasonable skill landing position. In this way, when a novice player plays against the AI player, it is not easy to identify the AI player, which improves the anthropomorphism degree of the AI player and also improves the game experience of the novice player.

[0031] Fifth aspect, an embodiment of the present application provides a training device, including: a processor, a memory, and a display; the memory is used to store programs and data, and the processor calls the programs stored in the memory to execute the game image processing method of the first aspect.

[0032] The technical solution provided by the embodiment of the present application may include the following beneficial effects: The terminal device can obtain the feature information of the second virtual character and the skill landing position of the first virtual character, and use them as real labels to train the neural network model, so as to obtain a supervised learning model. In this way, the AI player obtained through supervised learning can have the effect of a high degree of anthropomorphism, so as to realize the introduction of an AI player in the ABA game and ensure a high degree of anthropomorphism of the AI player; and after training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a high degree of anthropomorphism can be obtained. This AI player can play the game in the same way as a real player selects a second virtual object and releases a game skill during the game process, so as to intelligently select the second virtual character and release the game skill at a reasonable skill landing position. In this way, when a novice player plays against the AI player, it is not easy to identify the AI player, which improves the anthropomorphism degree of the AI player and also improves the game experience of the novice player.

[0033] Sixth aspect, an embodiment of the present application provides a terminal device, including: a processor, a memory, and a display; the memory is used to store programs and data, and the processor calls the programs stored in the memory to execute the game image processing method of the second aspect.

[0034] The technical solution provided by the embodiment of the present application may include the following beneficial effects: by obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained through supervised learning can have a high degree of anthropomorphism, so as to introduce an AI player in the ABA game and ensure a high degree of anthropomorphism of the AI player; and after training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a high degree of anthropomorphism can be obtained. This AI player can play the game in the same way as a real player selects a second virtual object and releases game skills during the game, so as to intelligently select the second virtual character and release the game skill at a reasonable skill landing position. In this way, when a novice player plays against the AI player, it is not easy to identify the AI player, which improves the anthropomorphism of the AI player and also improves the game experience of the novice player.

[0035] Seventh aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the game image processing method of the first aspect or the second aspect.

[0036] The technical solution provided by the embodiment of the present application may include the following beneficial effects: by obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained through supervised learning can have a high degree of anthropomorphism, so as to introduce an AI player in the ABA game and ensure a high degree of anthropomorphism of the AI player; and after training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a high degree of anthropomorphism can be obtained. This AI player can play the game in the same way as a real player selects a second virtual object and releases game skills during the game, so as to intelligently select the second virtual character and release the game skill at a reasonable skill landing position. In this way, when a novice player plays against the AI player, it is not easy to identify the AI player, which improves the anthropomorphism of the AI player and also improves the game experience of the novice player.

[0037] In an eighth aspect, an embodiment of the present application provides a computer program product, including a computer program which, when executed by a processor, implements the method for processing game images in the first aspect or the second aspect.

[0038] The technical solutions provided by the embodiments of the present application may include the following beneficial effects: By obtaining the feature information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained through supervised learning can have a high degree of anthropomorphism, thus realizing the introduction of an AI player in the ABA game and ensuring a relatively high degree of anthropomorphism of the AI player. And after training the neural network model, only by inputting the game scene image into the trained neural network model, an AI player with a relatively high degree of anthropomorphism can be obtained. This AI player can play the game in the same way as a real player selects a second virtual object and releases game skills during the game process, so as to intelligently select the second virtual character and release game skills at a reasonable skill landing position. In this way, when a novice player battles with the AI player, it is not easy to identify the AI player, which improves the degree of anthropomorphism of the AI player and also improves the game experience of the novice player. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 It is a schematic structural diagram of the game image processing system provided by the embodiment of the present application;

[0041] Figure 2 It is a flowchart of the first embodiment of the method for processing game images provided by the embodiment of the present application;

[0042] Figure 3 It is a flowchart of the second embodiment of the method for processing game images provided by the embodiment of the present application;

[0043] Figure 4 It is a schematic diagram of the supervised learning model provided by the embodiment of the present application;

[0044] Figure 5 It is a schematic diagram of the second virtual character within the field of view of the first virtual character provided by the embodiment of the present application;

[0045] Figure 6Schematic diagram of the second virtual character closest to the midline position within the field of view of the first virtual character provided by an embodiment of the present application;

[0046] Figure 7 Schematic diagram of the skill landing position provided by an embodiment of the present application;

[0047] Figure 8 Schematic diagram of training a neural network model provided by an embodiment of the present application;

[0048] Figure 9 Schematic diagram of the structure of the first embodiment of the game image processing device provided by an embodiment of the present application;

[0049] Figure 10 Schematic diagram of the structure of the second embodiment of the game image processing device provided by an embodiment of the present application;

[0050] Figure 11 Schematic diagram of the structure of a terminal device provided by an embodiment of the present application;

[0051] Figure 12 Schematic diagram of the structure of a training device provided by an embodiment of the present application. Detailed implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application under the inspiration of this embodiment belong to the scope of protection of the present application.

[0053] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above accompanying drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0054] In the prior art provided in the background art, there are at least the following technical problems:

[0055] Technicians in the field of online games have proposed the idea of introducing AI players into ABA games, so that novice players can become familiar with the game mechanisms in ABA games through the confrontation between AI players and novice players. However, technicians in this field have not further pointed out how to implement the technical solution of introducing AI players into ABA games. Instead, the technology of introducing AI players into symmetric confrontation competitive games is relatively mature. Introducing AI players into symmetric confrontation competitive games is usually achieved by training AI players using expert rule systems such as finite state machines or behavior trees. If the technology of introducing AI players in symmetric confrontation competitive games is used to introduce AI players into online games in ABA mode, due to the limitations of the expert rule system, the trained AI players cannot show diversity in ABA games, resulting in real players being very easy to identify AI players, reducing the game experience of real players, and thus resulting in a lower degree of anthropomorphism of the trained AI players. Therefore, there is currently a problem in the existing technology of how to introduce AI players into ABA games and ensure a relatively high degree of anthropomorphism of the introduced AI players.

[0056] In response to the above problems, this application proposes a method for processing game images. By obtaining video data of real players during the game process, parsing the video data, and determining the feature data of the target player closest to the midline of the field of view within the field of view of the real player, this feature data is used to represent the target player closest to the midline of the field of view within the field of view of the real player, and the feature data of this target player is used as a real label; then the game scene map is divided into multiple regions with the same size and shape, and the landing area of the real player's game skills in the game scene map is determined as its skill landing position, and this skill landing position is used as another real label. Then, a large amount of video data of real players, the feature data of the target player, and the skill landing position are used to train a neural network model, so that a supervised learning model can be obtained, and thus a highly anthropomorphic AI player can be trained. Moreover, the efficiency of training AI players through the method of supervised learning is relatively high, and it also meets the current requirement of the relatively fast game update speed. First, the terms involved in this application are explained below.

[0057] Graphical User Interface (GUI for short): Also known as the graphical user interface, it refers to the computer operation user interface displayed in a graphical way.

[0058] Neural network model: A neural network (Neural Networks, abbreviated as NN) is a complex network system formed by a large number of simple processing units (called neurons) interconnected extensively. It reflects many basic characteristics of the human brain function and is a highly complex non-linear dynamic learning system. Neural networks have the capabilities of large-scale parallelism, distributed storage and processing, self-organization, self-adaptation, and self-learning, and are particularly suitable for processing imprecise and fuzzy information processing problems that require considering many factors and conditions simultaneously. The neural network model is described based on the mathematical model of neurons. The neural network model is represented by network topology, node characteristics, and learning rules.

[0059] Convolutional Neural Networks (abbreviated as CNN): It is a class of feedforward neural networks (Feedforward Neural Networks) that contain convolutional calculations and have a deep structure, and is one of the representative algorithms of deep learning. Convolutional neural networks have the ability of representation learning and can perform translation-invariant classification on input information according to their hierarchical structure.

[0060] The core idea of the game image processing method provided in this application is to train an AI player through supervised learning. First, a large amount of game video data of real players is obtained. By parsing and processing these video data, the tasks, scene states, and events at each moment can be obtained, and thus the feature data of the target player closest to the midline of the field of view within the field of view of the real player can be obtained. The feature data of this target player is used as a real label. Then, the game scene map is divided into multiple regions with the same size and shape, and the landing region of the real player's game skills in the multiple regions is determined. This landing region is determined as the skill landing position, and this skill landing position is used as another real label. Then, the neural network model is trained using the video data, the target player, and the skill landing position, so as to obtain a supervised learning model, and then an AI player with a high degree of anthropomorphism can be obtained, making the game behavior of this AI player in the game conform to the game behavior of real players. Therefore, the technical solution of this application can improve the anthropomorphism degree of the AI player, increase the training speed of the AI player, and enhance the update speed of the game.

[0061] In one embodiment, the game image processing method in one of the embodiments of this application can run on a local terminal device or a server. When the game image processing method runs on the server, the game image processing method can be implemented and executed based on a cloud interaction system, where the cloud interaction system includes a server and a client device.

[0062] In an alternative embodiment, various cloud applications can run under the cloud interaction system, such as cloud games. Taking cloud games as an example, cloud games refer to a game mode based on cloud computing. In the operation mode of cloud games, the running entity of the game program and the presenting entity of the game screen are separated. The storage and operation of the game image processing method are completed on the cloud game server. The role of the client device is to receive and send data and present the game screen. For example, the client device can be a display device with data transmission function near the user side, such as a mobile terminal, a television, a computer, a palm computer, etc.; however, the cloud game server in the cloud performs information processing. When playing a game, the player operates the client device to send an operation instruction to the cloud game server. The cloud game server runs the game according to the operation instruction, encodes and compresses data such as the game screen, and returns it to the client device through the network. Finally, the client device decodes and outputs the game screen.

[0063] In an alternative embodiment, taking a game as an example, the local terminal device stores a game program and is used to present the game screen. The local terminal device is used to interact with the player through a graphical user interface, that is, conventionally, the game program is downloaded and installed on an electronic device and run. The way for the local terminal device to provide the graphical user interface to the player can include various methods. For example, it can be rendered and displayed on the display screen of the terminal, or provided to the player through holographic projection. For example, the local terminal device can include a display screen and a processor. The display screen is used to present the graphical user interface, and the graphical user interface includes the game screen. The processor is used to run the game, generate the graphical user interface, and control the display of the graphical user interface on the display screen.

[0064] In one embodiment, the game image processing method can be applied in an application scenario. Figure 1 It is a schematic structural diagram of the game image processing system provided by the embodiment of the present application, as Figure 1 shown. In this scenario, the game image processing system can include a data acquisition device 101, a database 102, a training device 103, an execution device 104, a data storage system 105, and a user device 106. Among them, the execution device 103 includes a computing module 107 and an I / O interface 108, and the computing module 107 includes a target model / rule 109.

[0065] The data acquisition device 101 can be used to obtain the feature information of the second virtual character that is within the field of view of the first virtual character in the game scene image and is closest to the midline position of the field of view, and obtain the skill landing position of the first virtual character in the game scene image, and store the game scene image, the feature information of the second virtual character, and the skill landing position of the first virtual character in the game scene image into the database 102. Among them, the feature information of the second virtual character and the skill landing position of the first virtual character can be used as the label data of the game scene image.

[0066] The data acquisition device 101 can determine the feature information of the second virtual character and the skill landing position of the first virtual character by acquiring the game scene images of real players. A large number of game scene images of real players, the feature information of the second virtual character, and the skill landing position of the first virtual character are stored in the database 102. The game scene image can be a game video and / or a game picture.

[0067] The training device 103 generates the target model / rule 109 based on the game scene image, the feature information of the second virtual character, and the skill landing position of the first virtual character in the database 102. Among them, the feature information of the second virtual character can include: character class, skill cooldown status, character status (current HP, whether hanging on a rocket chair, etc.), movement speed, position, orientation, talent, healing speed, etc.; the target model / rule 109 can be a supervised learning model, etc.

[0068] The training device 103 can execute the game image processing method in the embodiments of the present application, so as to train and obtain the target model / rule 109 for obtaining AI players. The target model / rule 109 obtained by the training device 103 can be applied to different systems or devices.

[0069] The execution device 104 is configured with an I / O interface 108 and can perform data interaction with the user device 106. The user can input a game scene image to the I / O interface 108 through the user device 106; the computing module 107 in the execution device 104 processes the game scene image input to the target model / rule 109, so as to obtain the feature information of the second virtual character and the skill landing position of the first virtual character; the I / O interface 108 returns the feature information of the second virtual character and the skill landing position of the first virtual character to the user device 106, and the user device 106 provides it to the user. The user can be a game developer or a game planner.

[0070] The execution device 104 can call the data, code, etc. in the data storage system 105, and can also store data, instructions, etc. in the data storage system 105.

[0071] In the above scenario, in one case, the user can manually input a game scene image to the I / O interface 108 through the user device 106. For example, the user can operate in the interface provided by the I / O interface 108. In another case, the user device 106 can automatically input a game scene image to the I / O interface 108 and obtain the feature information of the second virtual character and the skill landing position of the first virtual character returned by the I / O interface 108. It should be noted that if the user device 106 automatically inputs data to the I / O interface 108 and obtains the result returned by the I / O interface 108, the user device 106 needs to obtain the authorization of the user, and the user can set the corresponding permissions in the user device 106.

[0072] In the above scenario, the user device 106 can also be used as a data acquisition end to store a large number of game scene images of real players, the feature information of the second virtual character, and the skill landing position of the first virtual character collected into the database 102.

[0073] It should be noted that Figure 1 the structure of the game image processing system shown in Figure 1 is only a schematic diagram. The positional relationship between the devices, components, modules, etc. shown in the figure does not constitute any limitation. For example, in

[0074] in this case, the data storage system 105 is an external memory relative to the execution device 104. In other cases, the data storage system 105 can also be placed in the execution device 104; the database 102 is an external memory relative to the training device 103. In other cases, the database 102 can also be placed in the training device 103.

[0074] Combined with the above scenario, the technical solution of the game image processing method provided by the present application will be described in detail through several specific embodiments below.

[0075] Figure 2 FIG. is a flowchart of the first embodiment of the game image processing method provided by the embodiment of the present application. As Figure 2 shown, this method can be executed by the training device in Figure 1 and this method includes the following steps:

[0076] S201: Obtain the feature information of the second virtual character that is within the field of view of the first virtual character in the game scene image and is closest to the midline position of the field of view.

[0077] In this step, it can be achieved through Figure 1The training device in [device name] obtains a large number of game scene images of real players, thereby obtaining the feature information of the second virtual character. The game scene images include multiple virtual characters, and these virtual characters can be virtual characters controlled by real players. When training an AI player, the two most crucial parts are the feature information of the target object (i.e., the second virtual character) to be selected by the AI player, and the skill landing position of the AI player's game skills. Therefore, it is necessary to obtain the feature information of the second virtual character in the game scene image and use the obtained feature information of the second virtual character as a true label. The feature information of the second virtual character is used to represent the second virtual character.

[0078] In the above solution, for real players, during the game process, it is impossible to choose a target object outside their field of vision to release game skills. Therefore, the target object to be selected must be within their field of vision, that is, the second virtual character must be within the field of vision of the first virtual character. In this way, the trained AI player will not attack target objects outside the field of vision, and thus will not cause the AI player to exhibit anti-human behavior. Also, for real players, when the virtual character they control releases a certain game skill, it will not attack multiple target objects at different positions that are far away at the same time. Therefore, it is necessary to determine one second virtual character, or determine multiple second virtual characters within the skill attack range of the first virtual character. At the same time, for real players, the target object they want to select is generally located in the middle of their field of vision. Therefore, the second virtual character needs to meet two conditions: being within the field of vision of the first virtual character and being the closest to the midline position of the field of vision. In this way, the trained AI player will not only not exhibit anti-human behavior, but also be more in line with the game behavior of real players. Therefore, the trained AI player has a high degree of anthropomorphism.

[0079] In the above solution, the game scene image can be a game video image or a game picture image. Taking the game video image as an example, real players record videos during the game process, and the game data of real players will be saved in the form of videos. The essence of a video is frame snapshots plus event triggers. By simply parsing and processing the video, the status of characters, scenes, and events that occur at each moment can be obtained.

[0080] S202: Obtain the skill landing position of the first virtual character in the game scene image.

[0081] In this step, when training an AI player, the other most crucial part is the skill landing position of the AI player's game skills. For real players, after determining the second virtual character they have selected, they need to control the first virtual character to release game skills for attack. Therefore, it can be achieved through Figure 1The training device in [device name] obtains the skill landing position of the first virtual character in the game scene image, and uses the obtained skill landing position as another true label.

[0082] S203: Train the neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position.

[0083] In this step, the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image. After the training device trains the neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position, a supervised learning model can be obtained, thereby obtaining a trained AI player. Figure 4 This is a schematic diagram of the supervised learning model provided by the embodiments of the present application. As Figure 4 shown, the training of the supervised learning model (Supervised Learning Model, abbreviated as: SL Model) requires (x (i) , y (i) ), where x (i) is the sample feature, y (i) is Figure 4 the y in target , that is, the true label, and y pred is the result predicted by the supervised learning model. The optimizer continuously reduces the difference between y target and y pred , and finally makes the y pred output by the supervised learning model approximate to y target . For the first virtual character in the game, in a set of training data (x (i) , y (i) ), where, is a feature in a certain dimension. This feature is mainly divided into two major parts: image features and digital features. The image features and digital features will be introduced later.

[0084] In the above steps, when introducing the trained AI player into the game, the feature information of the second virtual character to be selected by the AI player and the skill landing position can be determined according to the game scene image. Therefore, in the actual game process, the AI player can reasonably determine the specific second virtual character and the skill landing position of its game skill. Therefore, the trained AI player has a high degree of anthropomorphism.

[0085] In the above steps, when training the neural network model, a large amount of game image data of real players can be used, so that the trained AI player will be more in line with the game behaviors of real players.

[0086] The method for processing game images provided in this embodiment can obtain the feature information of the second virtual character and the skill landing position of the first virtual character, and use them as real labels to train a neural network model, so as to obtain a supervised learning model. In this way, the AI player obtained through supervised learning can have a high degree of anthropomorphism, thereby realizing the introduction of an AI player in the ABA game and ensuring a high degree of anthropomorphism of the AI player.

[0087] In one embodiment, obtaining the feature information of the second virtual character within the field of view of the first virtual character in the game scene image includes: determining the target virtual character in the game scene image; obtaining multiple target points on the target virtual character; performing ray detection on the first virtual character and the multiple target points to determine whether there is a collision body between the first virtual character and the target points; if there is no collision body between the first virtual character and at least one target point, determining that the target virtual character is within the field of view; obtaining the distance between each target virtual character within the field of view and the midline position of the field of view; determining the target virtual character with the smallest distance from the midline position of the field of view as the second virtual character; and obtaining the feature information of the second virtual character.

[0088] In this solution, in addition to a first virtual character in the game scene image, there are also multiple target virtual characters. The target object to be selected by the first virtual character needs to be within its field of view. Therefore, it is necessary to first determine the multiple target virtual characters within its field of view, and then determine the target virtual character with the smallest distance from the midline position of the field of view among the multiple target virtual characters within the field of view, and determine this target virtual character as the second virtual character.

[0089] In the above solution, when determining the target virtual character within the field of view of the first virtual character, multiple target points can be selected on the target virtual character first, and then ray detection is performed with the first virtual character as the starting point and the multiple target points on the target virtual character as the ending points to determine whether each ray detects a collision body. If any one ray does not detect a collision body, it means that the target virtual character is within the field of view of the first virtual character, as Figure 5 shown. Among them, the collision body can be any object with an occlusion effect in the game scene image, such as buildings, trees, stones, boxes, etc. For example, when a certain ray does not detect a wall occlusion between the first virtual character and the target virtual character, while other rays detect a wall occlusion between the first virtual character and the target virtual character, it means that the first virtual character can at least see the part where a certain target point on the target virtual character is located. Therefore, the target virtual character is within the field of view of the first virtual character.

[0090] Figure 5 This is a schematic diagram showing that the second virtual character provided in the embodiment of the present application is within the field of view of the first virtual character. In Figure 5 , taking the game scene displayed on the current game screen as the field of view of the first virtual character, for a computer, it is necessary to use the ray detection method to determine which virtual characters in the game scene image are within the field of view of the first virtual character. Therefore, 5 target points can be set on the target virtual character, and then, taking the first virtual character as the starting point and the 5 target points as the ending points, ray detections are respectively performed. When there is a collider between the starting point and the ending point, False will be returned to the computer, and when there is no collider between the starting point and the ending point, True will be returned. When the computer detects that any one ray returns True, it means that the target virtual character is within the field of view of the first virtual character. If all rays return False, it means that the target virtual character is not within the field of view of the first virtual character.

[0091] In the above solution, for a real player, when the virtual character controlled by him releases a certain game skill, he will not attack multiple target objects at different positions that are far away at the same time. Therefore, it is necessary to determine a second virtual character, or determine multiple second virtual characters within the skill attack range of the first virtual character. At the same time, for a real player, the target object he wants to select is generally located in the middle of his field of view. Therefore, the second virtual character needs to meet two conditions at the same time: being within the field of view of the first virtual character and being the closest to the midline position of the field of view. In this way, the trained AI player will not only not show anti-human behaviors, but also be more in line with the game behaviors of real players. Therefore, the trained AI player has a high degree of anthropomorphism, as Figure 6 shown.

[0092] Figure 6 This is a schematic diagram of the second virtual character closest to the midline position of the field of view of the first virtual character provided in the embodiment of the present application. In Figure 6 , taking the game scene displayed on the current game screen as the field of view of the first virtual character as an example, the midline position of the field of view is the black solid line shown in Figure 6 . Generally speaking, the target object selected by a player is generally in the middle of the player's field of view. It is extremely unlikely that the target object is in the front, but the player is walking away from the target object with his back turned. When a player finds that there are multiple target objects within his field of view, he will give priority to attacking the target object in the middle of his field of view. Therefore, it is necessary to determine the target object closest to the midline position of the field of view. As Figure 6As shown, the distance A between the target virtual character A and the black solid line is greater than the distance B between the target virtual character B and the black solid line. Therefore, the target virtual character B is the second virtual character that is finally determined to be within the field of view of the first virtual character and is the closest to the midline position of the field of view.

[0093] In the above solution, if there are collision bodies between the first virtual character and multiple target points, it is determined that the target virtual character is not within the field of view. That is, if all the rays detect collision bodies, it means that there is an occluding object between the target virtual character and the first virtual character, and the first virtual character cannot see any part of the target virtual character at all. Therefore, the target virtual character is not within the field of view of the first virtual character. For the second virtual character that is not within the field of view of the first virtual character, it can be ignored and not processed. This can avoid the problem of anti-human behavior caused by the AI player attacking targets outside the field of view, thereby improving the anthropomorphic degree of the AI player.

[0094] In one embodiment, obtaining the skill landing position of the first virtual character in the game scene image includes: dividing the game scene map where the game scene image is located into multiple regions, where the shapes and sizes of the multiple regions are the same; determining the landing region of the game skill of the first virtual character in the multiple regions; and determining the landing region as the skill landing position.

[0095] In this solution, the game scene displayed in the game scene image is a game scene at a certain position on the game scene map. Therefore, the game scene map shows all the regions in the game. When the first virtual character releases a game skill, the skill landing position can be any position on the game scene map, and there can be an infinite number of such positions. If the landing position of the game skill of the first virtual character in the game scene map is directly used as its skill landing position, training the neural network model will evolve into a regression model, which will increase the training difficulty of the neural network model. However, if the game scene map is divided into multiple regions, since the divided regions are limited, using the landing region of the game skill as the skill landing position will evolve the training of the neural network model into a classification model, which can reduce the training difficulty of the neural network model, thereby reducing the training time for the AI player and improving the update speed of the game.

[0096] In the above solution, as Figure 7 shown, Figure 7 is a schematic diagram of the skill landing position provided by the embodiment of the present application. In Figure 7 , the game scene map is divided into 11 * 10 = 110 blocks. If the landing point of the game skill of the first virtual character is in Figure 7If it is within the black area in [the figure], then this black area is the skill landing position of the first virtual character.

[0097] In one embodiment, training the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position includes: training the neural network model according to the digital features and the image features, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located.

[0098] In this solution, a deep neural network can be used as the model for supervised learning. The deep neural network mainly includes two parts: digital features and image features. The game scene image and the game scene map are used to display the scene layout during the game process, so they can be used as image features. The feature information of the second virtual character is used to determine the second virtual character, and the skill landing position is used to determine the landing position of the game skill of the first virtual character. For a computer, it can only determine the specific second virtual character and the skill landing position of the first virtual character through numbers. Therefore, the feature information of the second virtual character and the skill landing position are used as digital features.

[0099] In the above solution, by using a deep neural network as the model for supervised learning, the training process of the neural network model can be simplified, the training speed can be improved, thereby improving the training speed of the AI player, and further improving the update speed of the game.

[0100] In one embodiment, training the neural network model according to the digital features and the image features includes: performing one-dimensional processing on the image features to obtain multiple one-dimensional arrays; fusing the multiple one-dimensional arrays and the digital features to obtain a target array; and inputting the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

[0101] In this solution, the one-dimensional processing of the image features can be implemented by using the flatten layer in the neural network. Since the digital features are already one-dimensional, there is no need to process the digital features. The fusion processing of the obtained multiple one-dimensional arrays and the digital features can be implemented by using the concat layer in the neural network. Inputting the target array into the neural network model for classification processing can be implemented by using the fully connect (fc) layer in the neural network model, as Figure 8 shown.

[0102] Figure 8 is a schematic diagram of training the neural network model provided by the embodiment of the present application. In Figure 8Among them, the image features include the game scene image and the game scene map. The game scene image is also the player's field of view, and the game scene map is also the mini-map in the game. After the image features are expanded through the flatten layer of the neural network to obtain multiple one-dimensional arrays, the digital features and the multiple one-dimensional arrays are fused and connected through the concat layer of the neural network, and then classified by the fully connected layer in the neural network. Finally, the feature information of the second virtual character and the skill landing position are output through the two fully connected layers after classification.

[0103] In the above solution, training the neural network model in this way can improve the anthropomorphic degree of the trained AI player, and on the premise of achieving the same level of AI players, this technical solution can reduce the training time compared with other reinforcement learning, thereby improving the game update speed.

[0104] In one embodiment, before the image features and the digital features are respectively one-dimensionalized, the method further includes: processing the image features by using a convolutional neural network.

[0105] In this solution, as Figure 8 shown, before the image features and the digital features are respectively one-dimensionalized, the image features can be processed by using a convolutional neural network first. Processing the image features through the CNN can pay more attention to some local features in the image, thereby reducing the computational amount of image processing, reducing the training time of the AI player, and further improving the game update speed.

[0106] In one embodiment, the digital features further include: the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map.

[0107] In this solution, in addition to including the feature information of the second virtual character and the skill landing position, the digital features also include other feature information in the game, such as the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map, etc. In this way, the trained AI player can be more in line with real players, thereby improving the anthropomorphic degree of the AI player.

[0108] For example, digital features include in-game information during the game process: environmental features, such as the status of cipher machines in the game (whether the cipher machine has been deciphered, etc.), the status of rocket chairs (whether there is a player hanging on the rocket chair, whether the rocket chair is damaged, etc.), the situation of boards or windows, etc.; the feature information of the second virtual character, such as character class, skill cooldown status, character status (current health, whether hanging on the rocket chair, etc.), movement speed, position, orientation, talent, healing speed, etc.; the feature information of the first virtual character, such as character class, skill cooldown status (including the cooldown status of auxiliary skills and normal attack skills), movement speed, position, orientation, talent, etc. Digital features also include out-of-game information during the game process: the feature information of the second virtual character, such as character rank information, account information, historical teaming information, historical game battle information, etc.; map features, such as map type, the location of the cellar in the map, the spawn points of the first virtual character and the second virtual character, the location of cipher machines, the location of rocket chairs, etc.

[0109] The game image processing method provided in this embodiment obtains the video data of real players during the game process, analyzes the video data, determines the feature information of the target player closest to the midline of the field of view within the field of view of the real player, and uses the feature information of the target player as a real label; then divides the game scene map into multiple regions with the same size and shape, and determines the landing area of the real player's game skills in the game scene map as the skill landing position, and uses this skill landing position as another real label. Then, a large amount of video data of real players, the feature information of the target player, and the skill landing position are used to train the neural network model, so that a supervised learning model can be obtained, and thus a highly anthropomorphic AI player can be trained. Moreover, the efficiency of training the AI player by the supervised learning method is relatively high, and it also meets the current requirement of the relatively fast game update speed.

[0110] Figure 3 It is a flowchart of the second embodiment of the game image processing method provided in this application. As Figure 3 shown, this method can be executed by the execution device in Figure 1 and includes the following steps:

[0111] S301: Input the game scene image into the trained neural network model for processing to obtain the feature information of the second virtual character among multiple virtual characters, and the skill landing position of the first virtual character among multiple virtual characters.

[0112] In this step, a graphical user interface can be provided by a terminal device. The graphical user interface includes a game scene image, which can be a game video image or a game picture image. The game scene image includes multiple virtual characters. The first virtual character among these virtual characters is an AI player, and the second virtual character can be an AI player or a virtual character controlled by a real player. The second virtual character is within the field of view of the first virtual character and is closest to the midline position of the field of view. The above terminal device can be Figure 1 the execution device in

[0113] In the above step, when introducing the trained AI player into the game, the execution device can determine the characteristic information of the second virtual character to be selected by the AI player and the skill landing position according to the game scene image. Therefore, in the actual game process, the AI player can determine the specific second virtual character through the characteristic information of the second virtual character and reasonably determine the skill landing position of its game skills. Therefore, the trained AI player has a high degree of anthropomorphism.

[0114] In one embodiment, before inputting the game scene image into the trained neural network model for processing, the method further includes: training the neural network model according to the game scene image, the characteristic information of the second virtual character, and the skill landing position to obtain the trained neural network model, where the characteristic information of the second virtual character and the skill landing position are used as the label data of the game scene image.

[0115] In this solution, by obtaining the characteristic information of the second virtual character and the skill landing position of the first virtual character and using them as real labels to train the neural network model, a supervised learning model can be obtained. In this way, the AI player obtained through supervised learning can have the effect of a high degree of anthropomorphism, so as to introduce the AI player into the ABA game and ensure a high degree of anthropomorphism of the AI player.

[0116] In the above solution, the method of training the neural network model can be trained according to the method as Figure 2 shown.

[0117] The game image processing method provided in this embodiment introduces the trained AI player into the game. Through the game scene image, the AI player can reasonably select the second virtual character and the skill landing position of the AI player's game skills, so that the AI player has a high degree of anthropomorphism. Moreover, the efficiency of training the AI player through the supervised learning method is relatively high, which also meets the requirement of the current relatively fast game update speed.

[0118] Generally speaking, the technical solution provided by this application uses a deep neural network as a supervised learning model, and trains the neural network model based on two real labels, namely the feature information of the second virtual character and the skill landing position, and the game scene image. This avoids the problems of the need for a large amount of computing power for trial and error in reinforcement learning and the relatively weak generalization performance. It is a technical implementation method that can not only make the trained AI player have a high degree of anthropomorphism, but also improve the training speed of the AI player and the game update speed.

[0119] Figure 9 FIG. 4 is a schematic structural diagram of the first embodiment of the game image processing device provided by the embodiment of this application. The game image processing device 900 includes:

[0120] A first acquisition module 901, configured to acquire the feature information of the second virtual character within the field of view of the first virtual character in the game scene image and closest to the midline position of the field of view;

[0121] A second acquisition module 902 is further configured to acquire the skill landing position of the first virtual character in the game scene image;

[0122] A training module 903, configured to train the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position, wherein the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image.

[0123] Optionally, the first acquisition module 901 is further configured to determine the target virtual character in the game scene image; acquire multiple target points on the target virtual character; perform a ray detection on the first virtual character and the multiple target points to determine whether there is a collision body between the first virtual character and the target points; if there is no collision body between the first virtual character and at least one target point, determine that the target virtual character is within the field of view; acquire the distance between each target virtual character within the field of view and the midline position of the field of view; determine the target virtual character with the smallest distance from the midline position of the field of view as the second virtual character; and acquire the feature information of the second virtual character.

[0124] Optionally, the second acquisition module 902 is further configured to divide the game scene map where the game scene image is located into multiple regions, where the shapes and sizes of the multiple regions are the same; determine the landing region of the game skill of the first virtual character in the multiple regions; and determine the landing region as the skill landing position.

[0125] Optionally, the training module 903 is further configured to train the neural network model according to digital features and image features, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located.

[0126] Optionally, the training module 903 is further configured to perform one-dimensional processing on the image features to obtain a plurality of one-dimensional arrays; fuse the plurality of one-dimensional arrays and the digital features to obtain a target array; and input the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

[0127] Optionally, the digital features further include: the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map.

[0128] The game image processing device provided in this embodiment is used to execute the technical solution of the game image processing method in the foregoing method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here.

[0129] Figure 10 FIG. 12 is a schematic structural diagram of Embodiment 2 of the game image processing device provided in this application embodiment. The game image processing device 1000 includes: a processing module 1001 and a display module 1002. The display module 1002 is used to display a graphical user interface, and the graphical user interface includes a game scene image, and the game scene image includes a plurality of virtual characters.

[0130] The processing module 1001 is configured to input the game scene image into the trained neural network model for processing to obtain the feature information of the second virtual character among the plurality of virtual characters, and the skill landing position of the first virtual character among the plurality of virtual characters, where the second virtual character is located within the field of view of the first virtual character and is closest to the midline position of the field of view.

[0131] Optionally, the processing module 1001 is further configured to, before inputting the game scene image into the trained neural network model for processing, train the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position to obtain a trained neural network model, where the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image.

[0132] The game image processing device provided in this embodiment is used to execute the technical solution of the game image processing method in the foregoing method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here.

[0133] Figure 11 FIG. 25 is a schematic structural diagram of a terminal device provided in an embodiment of this application. As Figure 11 shown, the terminal device 1100 includes:

[0134] a processor 1111, a memory 1112, and a display 1113;

[0135] The memory 1112 is used to store programs and data, and the processor 1111 calls the programs stored in the memory to execute the technical solution of the method for processing game images provided by the foregoing method embodiments.

[0136] In the above terminal device, the memory 1112 and the processor 1111 are directly or indirectly electrically connected to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines, such as being connected through a bus. The memory 1112 stores computer-executable instructions for implementing the method for processing game images, including at least one software function module stored in the memory in the form of software or firmware. The processor 1111 executes various functional applications and data processing by running the software programs and modules stored in the memory 1112.

[0137] Figure 12 The following is a schematic structural diagram of a training device provided by an embodiment of the present application, as Figure 12 shown, the training device 1200 includes:

[0138] a processor 1211, a memory 1212, and a display 1213;

[0139] The memory 1212 is used to store programs and data, and the processor 1211 calls the programs stored in the memory to execute the technical solution of the method for processing game images provided by the foregoing method embodiments.

[0140] In the above training device, the memory 1212 and the processor 1211 are directly or indirectly electrically connected to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines, such as being connected through a bus. The memory 1212 stores computer-executable instructions for implementing the method for processing game images, including at least one software function module stored in the memory in the form of software or firmware. The processor 1211 executes various functional applications and data processing by running the software programs and modules stored in the memory 1212.

[0141] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electric Erasable Programmable Read-Only Memory (EEPROM), etc. Among them, the memory is used to store programs, and after receiving the execution instruction, the processor executes the program. Further, the software programs and modules in the above-mentioned memory may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide a running environment for other software components.

[0142] The processor can be an integrated circuit chip with the ability to process signals. The above-mentioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0143] The embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium includes a program, and the program is used to implement the technical solution of the game image processing method provided in the method embodiments when executed by the processor.

[0144] The present application also provides a computer program product, including: a computer program, and the computer program is used to implement the technical solution of the game image processing method provided in the foregoing method embodiments when executed by the processor.

[0145] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: ROM, RAM, magnetic disk, or optical disk and other various media that can store program codes.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for processing game images, characterized in that, The method includes: Obtaining the feature information of a second virtual character that is within the field of view of a first virtual character in a game scene image and is closest to the midline position of the field of view; Obtaining the skill landing position of the first virtual character in the game scene image; Training a neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position, where the feature information of the second virtual character and the skill landing position serve as the labeled data for the game scene image; The training of the neural network model based on the game scene image, the feature information of the second virtual character, and the skill landing position includes: Performing one-dimensional processing on the image features to obtain multiple one-dimensional arrays; Fusing the multiple one-dimensional arrays and digital features to obtain a target array, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located; Inputting the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

2. The method according to claim 1, characterized in that, The obtaining of the feature information of the second virtual character that is within the field of view of the first virtual character in the game scene image and is closest to the midline position of the field of view includes: Determining the target virtual character in the game scene image; Obtaining multiple target points on the target virtual character; Performing ray detection on the first virtual character and the multiple target points to determine whether there is a collision body between the first virtual character and the target points; If there is no collision body between the first virtual character and at least one of the target points, determining that the target virtual character is within the field of view; Obtaining the distance between each target virtual character within the field of view and the midline position of the field of view; Determining the target virtual character with the smallest distance from the midline position of the field of view as the second virtual character; Obtaining the feature information of the second virtual character.

3. The method according to claim 1, wherein The obtaining of the skill landing position of the first virtual character in the game scene image includes: Dividing the game scene map where the game scene image is located into multiple regions, where the shapes and sizes of the multiple regions are the same; Determining the landing region of the game skill of the first virtual character in the multiple regions; Determining the landing region as the skill landing position.

4. The method according to claim 1, wherein The digital features further include: the feature information of the first virtual character, the environmental features of the game scene image, and the map features of the game scene map.

5. A method for processing game images, characterized in that, Providing a graphical user interface through a terminal device, where the graphical user interface includes a game scene image, and the game scene image includes multiple virtual characters, and the method includes: Input the game scene image into the trained neural network model for processing to obtain the feature information of the second virtual character among the multiple virtual characters and the skill landing position of the first virtual character among the multiple virtual characters, where the second virtual character is within the field of view of the first virtual character and is closest to the midline position of the field of view; Before inputting the game scene image into the trained neural network model for processing, the method further includes: Perform one-dimensional processing on the image features to obtain multiple one-dimensional arrays; Fuse the multiple one-dimensional arrays and the digital features to obtain a target array, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located; Input the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

6. A processing device for game images, characterized in that, It includes: A first acquisition module for acquiring the feature information of the second virtual character that is within the field of view of the first virtual character and is closest to the midline position of the field of view in the game scene image; A second acquisition module for acquiring the skill landing position of the first virtual character in the game scene image; A training module for training the neural network model according to the game scene image, the feature information of the second virtual character, and the skill landing position, where the feature information of the second virtual character and the skill landing position are used as the label data of the game scene image; The training module is further used for training the neural network model according to the digital features and the image features, where the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located; The training module is further used for performing one-dimensional processing on the image features to obtain multiple one-dimensional arrays; Fuse the multiple one-dimensional arrays and the digital features to obtain a target array; Input the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

7. A processing device for game images, characterized in that It includes: A processing module and a display module, the display module is used for displaying a graphical user interface, and the graphical user interface includes a game scene image, and the game scene image includes multiple virtual characters; The processing module is used for inputting the game scene image into the trained neural network model for processing to obtain the feature information of the second virtual character among the multiple virtual characters and the skill landing position of the first virtual character among the multiple virtual characters, where the second virtual character is within the field of view of the first virtual character and is closest to the midline position of the field of view; The processing module is further configured to train the neural network model according to the digital features and the image features before inputting the game scene image into the trained neural network model for processing, wherein the digital features include the feature information of the second virtual character and the skill landing position, and the image features include the game scene image and the game scene map where the game scene image is located; The processing module is further configured to perform one-dimensional processing on the image features to obtain a plurality of one-dimensional arrays; Fuse the plurality of one-dimensional arrays and the digital features to obtain a target array; Input the target array into the neural network model for classification processing to obtain the feature information of the second virtual character and the skill landing position.

8. A training device, characterized in that, It includes: A processor, a memory, and a display; The memory is used to store programs and data, and the processor calls the programs stored in the memory to execute the game image processing method according to any one of claims 1 to 4.

9. A terminal device, characterized in that, It includes: A processor, a memory, and a display; The memory is used to store programs and data, and the processor calls the programs stored in the memory to execute the game image processing method according to claim 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the game image processing method according to any one of claims 1 to 5.

11. A computer program product, characterized in that, It includes a computer program, and when the computer program is executed by the processor, it is used to implement the game image processing method according to any one of claims 1 to 5.

Images