Adaptive interactive game strategy method and system based on deep learning detection

Abstract

The invention belongs to the technical field of computer software and human-computer interaction, and particularly relates to a self-adaptive interactive game strategy method and system based on deep learning detection. The adaptive interactive game strategy method based on deep learning detection comprises the following steps: S1, regularly obtaining a game picture, identifying scene Chinese in the game picture, and obtaining a game scene name corresponding to the scene Chinese; and S2, acquiring a game strategy corresponding to the game scene name, displaying a strategy window in the front of the current game picture when a calling strategy command is received, displaying the game strategy in the strategy window, and closing the strategy window when a strategy closing command is received. According to the detected game scene, the most suitable strategy is intelligently recommended, the requirements of players are met, the time of manually searching for the strategy by the players is saved, and the smoothness and fun of the game are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer software and human-computer interaction, and specifically relates to an adaptive interactive game strategy method and system based on deep learning detection. Background Art

[0002] With the continuous development of electronic games, players have higher and higher requirements for gaming experience. Traditional game strategies are usually provided in the form of text or video. Finding and using these strategies often requires switching screens or pausing the game, which affects the fluency and experience of the game. Usually, if players want to view the strategy corresponding to the game, they can only search for the corresponding strategy through external devices, use computer browsers, or use other electronic devices such as mobile phones and tablets to open the strategy, and view the strategy while playing the game. Summarize the player steps of the existing solution: identify the current game scene, search for the corresponding strategy from other devices, view the strategy, and continue playing the game.

[0003] It can be seen that the existing technology cannot achieve the goal of directly viewing the corresponding strategy in the game, and players need to search for the strategy on the device by themselves, which greatly destroys the player's immersive gaming experience. Game strategies are generally made by players themselves and shared online, so for game manufacturers, they will not integrate the strategy, so there is no such thing as a game strategy in the game itself, but the game strategy is an important part of the player's game. If the player needs to view the game strategy, he needs to frequently switch between the game screen and the strategy, which is even more cumbersome for players who use a controller to play the game. They also need to switch back and forth between the controller and the keyboard and mouse. This not only makes the player feel tired when playing the game, but also has an impact on some 3A games that require a certain threshold, such as player loss.

[0004] In addition, many games have different requirements for strategies in different stages or scenes, so a more intelligent and convenient way to access these strategies is needed.

[0005] In addition, deep learning is a branch of machine learning. It is based on artificial neural networks, especially deep neural networks, and aims to solve complex learning tasks by imitating the structure and function of the human brain. Artificial neural networks are the basis of deep learning. Artificial neural networks are composed of multiple neurons, which are arranged in a hierarchical structure, usually including input layers, hidden layers, and output layers. Each neuron receives input from the previous layer, and after weighting, summing, and activation function processing, it passes the result to the next layer. Finally, the extracted effective information is summarized and classified, and the probability of this information belonging to each known category is calculated to obtain the content recognized in the picture. Summary of the invention

[0006] The present invention aims to provide an adaptive interactive game strategy method and system based on deep learning detection, in order to solve the technical problem that in the prior art, when players need to check game strategies, they usually need to pause the game and manually search for game strategies, which affects the smoothness and experience of the game.

[0007] In order to solve the above technical problems, the first aspect of the present invention provides an adaptive interactive game strategy method based on deep learning detection, and the adaptive interactive game strategy method based on deep learning detection includes:

[0008] S1, regularly acquiring a game screen, identifying the scene Chinese characters in the game screen, and obtaining the game scene name corresponding to the scene Chinese characters;

[0009] S2, obtaining the game strategy corresponding to the game scene name, and when a strategy call command is received, displaying the strategy window in the front of the current game screen, wherein the game strategy is displayed in the strategy window, and when a strategy close command is received, closing the strategy window.

[0010] Optionally, in the aforementioned adaptive interactive game strategy method based on deep learning detection, in step S1, identifying the scene Chinese in the game screen and obtaining the game scene name corresponding to the scene Chinese includes:

[0011] The trained deep learning model is used to infer the game screen to obtain the name of the game scene.

[0012] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, the deep learning model adopts a convolutional neural network model.

[0013] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, the training process of the deep learning model is:

[0014] Collect several game images in different weather and environmental conditions in each game scene, and mark the scene Chinese characters appearing in the game images as scene identifiers to obtain a data set;

[0015] The deep learning model is trained and tested using the data set. When the model recognition accuracy is not less than a preset value, the model training is considered complete.

[0016] Optionally, in the aforementioned adaptive interactive game strategy method based on deep learning detection, in step S1, after acquiring the game screen and before identifying the scene Chinese in the game screen, it also includes:

[0017] The game screen is preprocessed by at least one of removing screen noise, extracting effective areas and enhancing edges.

[0018] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, when preprocessing the game screen, OpenCV software is used for preprocessing.

[0019] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, step S1 adopts a C++ detection program implemented in C++ language, and uses the ONNX RunTime inference engine to deploy the deep learning model on the C++ side for inference to obtain the name of the game scene.

[0020] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, step S2 uses a JS visualization window program implemented in JS language;

[0021] The game strategy is pre-stored in the back-end server;

[0022] When the strategy window needs to be displayed in front, the strategy window is configured as a window adapted to the terminal screen, and the H5 page loaded with the game strategy is loaded.

[0023] Optionally, in the adaptive interactive game strategy method based on deep learning detection as described above, step S1 adopts a C++ detection program implemented in C++ language, and step S2 adopts a JS visualization window program implemented in JS language, and the game scene name obtained by the C++ detection program is transmitted to the JS visualization window program through an interactive interface.

[0024] In order to solve the aforementioned technical problems, the second aspect of the present invention provides an adaptive interactive game strategy system based on deep learning detection, comprising:

[0025] A game scene detection module is used to periodically obtain the game screen, identify the scene Chinese characters in the game screen, and obtain the game scene name corresponding to the scene Chinese characters;

[0026] The game strategy visualization module is used to obtain the game strategy corresponding to the game scene name. When a strategy call command is received, a strategy window is displayed in front of the current game screen, and the game strategy is displayed in the strategy window. When a strategy close command is received, the strategy window is closed.

[0027] The positive progress of the present invention is that the present invention directly integrates the game strategies of all game scenes in the game, greatly reducing the time cost of players searching for strategies by themselves, and players can directly view the corresponding strategies in the game, which greatly enhances the players' gaming experience and game immersion compared to the existing solutions where players need to switch back and forth between different devices. Players can use the present invention to obtain relevant strategies in real time, reduce stuck time, obtain accurate information, and better understand game mechanisms and strategies. According to the detected game scenes, the most suitable strategies are intelligently recommended to meet the needs of players, saving the time of players manually searching for strategies, and enhancing the fluency and fun of the game. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The disclosure of the present invention will become more apparent with reference to the accompanying drawings. It should be understood that these drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present invention. In the drawings:

[0029] Figure 1 A flow chart of the present invention;

[0030] Figure 2 A flow chart of the training model of the present invention;

[0031] Figure 3 A flow chart of the C++ detection program of the present invention;

[0032] Figure 4 The present invention is a flow chart of the JS visualization window program. DETAILED DESCRIPTION

[0033] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0034] It should be noted that the following embodiments and features therein may be combined with each other if there is no conflict.

[0035] In the description of the present invention, it should be noted that directional words such as the terms "outside", "middle", "inside", "outside", etc., which indicate directions and positional relationships, are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and cannot be understood as limiting the specific protection scope of the present invention.

[0036] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features. Therefore, the definition of "first" and "second" features can explicitly or implicitly include one or more of the features. In the description of the present invention, "several" and "a number" mean two or more, unless otherwise clearly and specifically defined.

[0037] The embodiment of the present invention provides an adaptive interactive game strategy method based on deep learning detection, referring to Figure 1 , the method comprises the following specific steps:

[0038] S1, regularly obtain the game screen, identify the scene Chinese in the game screen, and obtain the game scene name corresponding to the scene Chinese.

[0039] S2, obtaining the game strategy corresponding to the game scene name. When a call-out strategy command is received, a strategy window is displayed in front of the current game screen. The game strategy is displayed in the strategy window. When a close strategy command is received, the strategy window is closed.

[0040] The timing in the present invention refers to collecting screenshots of the game screen at preset intervals, for example, collecting screenshots of the game screen at 10 frames per second.

[0041] In the present invention, for a single-frame game screen that is obtained, what is identified is the scene Chinese in the game screen. According to the observation of the game screen, when entering different game scenes, the Chinese of the corresponding scene will appear in the middle of the game screen. Therefore, the present invention uses the scene Chinese that appears as a unique identifier of the scene, and believes that the appearance of the corresponding Chinese in the game screen means that the player has entered the corresponding game scene. The present invention uses the Chinese in the game scene as an identifier, which greatly reduces the difficulty of detection, improves the detection speed and accuracy, and directly uses the game scene Chinese to associate with the game strategy, so that the strategy can be quickly obtained.

[0042] Existing solutions require players to search for corresponding strategies from other channels before viewing them, and players need to switch back and forth between different devices, which greatly reduces the real-time nature of the players' gameplay in the game and requires players to constantly switch between being separated from and immersed in the real world and the game world. The present invention can quickly detect and analyze the game screen while the game is in progress, and directly put the corresponding game strategy in the foreground, so that players can view the game strategy of the current game scene at any time.

[0043] The present invention designs a floating window that can be directly called out and closed during the game, and the player can directly use the combination of handle buttons in the game to call out and view the current strategy without switching to other operations such as keyboard and mouse. Compared with the existing solution in which the user needs to switch back and forth between different devices to view the game strategy, the player's interactive experience is greatly enhanced and the player's gaming immersion is greatly enhanced.

[0044] In some embodiments, in step S1, the scene Chinese in the game screen is identified to obtain the game scene name corresponding to the scene Chinese, including: using a trained deep learning model to infer the game screen to obtain the game scene name.

[0045] In deep learning, a model refers to a mathematical structure or algorithm used to represent and learn complex patterns in data. Deep learning models are usually composed of multiple layers, which are organized in a neural network manner. The goal of the model is to learn a mapping function from the input data so that it can make accurate predictions or classifications when given an input. In short, it takes an image as input, and after a series of calculations in the model, it can be obtained that the image is the probability of all the categories learned in the model, and the category with the highest probability is used as the output, and it is determined that the image contains the content of that category.

[0046] The present invention takes into account the strong stability and high robustness of deep learning in recognizing known scenes, and uses deep learning in artificial intelligence as the basic principle of recognition means.

[0047] The present invention utilizes deep learning to simulate the behavior of players observing the current game scene with their own eyes in the existing solution, and automatically identifies the current game scene. The four behavioral steps of the player in the existing solution: identifying the current game scene, searching for the corresponding strategy from other devices, viewing the strategy, and continuing to play the game are changed to viewing the strategy and continuing to play the game. These two behavioral steps greatly reduce the behavioral steps of players searching for game strategies in the existing solution.

[0048] Of course, the present invention may also adopt other methods in the prior art that can recognize scene Chinese from images.

[0049] In some embodiments, the deep learning model adopts a convolutional neural network model.

[0050] Convolutional neural networks can propose features through local perception areas, so that the labeled data can converge effectively. It also has strong generalization capabilities and can accurately identify the learned generally applicable features in images that have never appeared. Therefore, this embodiment selects convolutional neural networks as the model backbone.

[0051] In some embodiments, reference Figure 2 , the training process of the deep learning model is:

[0052] Data collection: Collect several game images in different weather and environmental conditions in each game scene.

[0053] For a game, all game scenes are known and have unique identifiers. Therefore, it is necessary to collect images of all game scenes that appear in the game in various situations. For example, the game scenes will be different in the same scene, different weather, and different environments. As many as possible need to be collected to improve the robustness of the model.

[0054] Data annotation: The Chinese characters in the scenes appearing in the game images are annotated as scene identifiers to obtain a data set.

[0055] For deep learning, data labeling is of utmost importance, which directly affects the stability and recognition accuracy of the model. Therefore, the unique identifiers in the game scenes can be selected to represent the game scenes, which can greatly reduce the impact of useless information data in the game scenes, and can also greatly enhance the recognition accuracy of the model. According to the observation of the game screen, when entering different game scenes, the Chinese characters of the corresponding scenes will appear in the middle of the game screen. Therefore, the Chinese characters of the scenes that appear are used as the unique identifiers of the scenes. It is believed that the corresponding Chinese characters appearing in the screen represent that the player has entered the corresponding game scene.

[0056] Model training and testing: Use the data set to train and test the deep learning model. When the model recognition accuracy is not less than the preset value, the model training is considered complete.

[0057] After marking all the Chinese characters that appear in all scenes, you can train the model. Train a model that can recognize game scenes. Before training, the data set can be divided into training set and test set. The training set is used for model training, and the test set is used for model testing.

[0058] The preset value in this embodiment is preferably 99%, that is, when the model recognition accuracy is ≥99%, the model training is considered to be completed, otherwise it is necessary to modify the model parameters such as weights or biases, or re-collect data.

[0059] The present invention can train different models according to different games to quickly become compatible with the games. It only needs to train the game scene pictures that appear in the game. The deep learning model backbone is highly robust and has high recognition accuracy. The amount of data for each game scene is only about ten pictures, which greatly reduces the time to adapt to new games.

[0060] In some embodiments, in step S1, after acquiring the game screen and before recognizing the scene Chinese in the game screen, a pre-processing process is also included:

[0061] The game screen is preprocessed by at least one of removing screen noise, extracting effective areas and enhancing edges, so as to reduce interference caused by complex scenes in the game screen.

[0062] When preprocessing the game screen, it is preferred to use OpenCV software for preprocessing.

[0063] In some embodiments, step S1 uses a C++ detection program implemented in C++ language, and uses the ONNX RunTime inference engine to deploy the deep learning model on the C++ side for inference to obtain the name of the game scene.

[0064] Since the interfaces provided by Windows system for C++ language are relatively rich and complete, this embodiment uses C++ language to capture the game screen, and uses C++ language to detect and output the content of the game screen.

[0065] In some embodiments, step S2 uses a JS visualization window program implemented in JS language; the game strategy is pre-stored in the back-end server; when the strategy window needs to be displayed in front, the strategy window is configured as a window adapted to the terminal screen, and the H5 page loaded with the game strategy is loaded.

[0066] Since it involves many backend operations, JS has powerful and efficient application support for this, so the JS language is chosen to develop a visual window for game strategies. Players can use the handle to operate and view the game strategies in the strategy window.

[0067] A large number of game strategies corresponding to each game scenario are pre-stored in the back-end server, and users can customize or update the game strategies according to different needs.

[0068] In some embodiments, step S1 adopts a C++ detection program implemented in C++ language, and step S2 adopts a JS visualization window program implemented in JS language. The game scene name obtained by the C++ detection program is transmitted to the JS visualization window program through an interactive interface.

[0069] This embodiment uses two machine languages ​​to divide the two sections. The functions of the two sections are different. By combining the two computer languages ​​​​C++ and JS and using different computer languages ​​​​for different functions, the software execution efficiency is higher and the software development speed is faster.

[0070] An embodiment of the present invention provides an adaptive interactive game strategy system based on deep learning detection, the system comprising:

[0071] The game scene detection module is used to periodically obtain the game screen, identify the scene Chinese in the game screen, and obtain the game scene name corresponding to the scene Chinese.

[0072] The game strategy visualization module is used to obtain the game strategy corresponding to the game scene name. When a strategy call command is received, the strategy window is displayed in the front of the current game screen. The game strategy is displayed in the strategy window. When a strategy close command is received, the strategy window is closed.

[0073] Embodiment 1:

[0074] Reference Figure 3 In this embodiment, a C++ detection program implemented in C++ language is used, and the C++ detection program performs the following steps:

[0075] The current game screen is obtained, and the game screen is preprocessed by using OpenCV software, including at least one of removing screen noise, extracting effective areas, and enhancing edges. The preprocessed game screen is inferred by using a deep learning model to obtain the name of the current game scene. The game scene name is directly transmitted to the JS visualization window program through an interactive interface and interacted with the JS visualization window program.

[0076] Reference Figure 4 In this embodiment, a JS visualization window program implemented in JS language is used, and the JS visualization window program performs the following steps:

[0077] Interact with the C++ detection program to obtain the name of the current game scene, obtain the game strategy corresponding to the current game scene name in the backend server, and wait for player commands.

[0078] After the player uses the button group on the handle controller to call out the strategy command, the strategy call command is received. At this time, the strategy window loaded with the game strategy can be directly displayed in the front of the current game screen, or a prompt "Do you want to call out the strategy window" can be popped up. After the player confirms, the strategy window is displayed.

[0079] After the player uses the button group on the handle controller to call out the command to close the strategy, the strategy window can be closed directly when the command to close the strategy is received, or a prompt "Do you want to close the strategy window" can be popped up. After the player confirms, the strategy window can be closed.

[0080] The present invention is described in detail above in conjunction with the embodiments of the drawings, and those skilled in the art can make various variations of the present invention according to the above description. Therefore, certain details in the embodiments should not constitute limitations on the present invention, and the present invention shall be protected by the scope defined by the attached claims.

Claims

1. An adaptive interactive game strategy method based on deep learning detection, characterized in that: The adaptive interactive game strategy method based on deep learning detection includes: S1, regularly acquiring a game screen, identifying the scene Chinese characters in the game screen, and obtaining the game scene name corresponding to the scene Chinese characters; S2, obtaining the game strategy corresponding to the game scene name, and when a strategy call command is received, displaying the strategy window in the front of the current game screen, wherein the game strategy is displayed in the strategy window, and when a strategy close command is received, closing the strategy window.

2. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 1, characterized in that: In step S1, the scene Chinese characters in the game screen are identified to obtain the game scene name corresponding to the scene Chinese characters, including: The trained deep learning model is used to infer the game screen to obtain the name of the game scene.

3. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 2, characterized in that: The deep learning model adopts a convolutional neural network model.

4. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 2, characterized in that: The training process of the deep learning model is: Collect several game images in different weather and environmental conditions in each game scene, and mark the scene Chinese characters appearing in the game images as scene identifiers to obtain a data set; The deep learning model is trained and tested using the data set. When the model recognition accuracy is not less than a preset value, the model training is considered complete.

5. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 1, characterized in that: In step S1, after acquiring the game screen and before identifying the scene Chinese in the game screen, it also includes: The game screen is preprocessed by at least one of removing screen noise, extracting effective areas and enhancing edges.

6. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 5, characterized in that: When preprocessing the game screen, OpenCV software is used for preprocessing.

7. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 1, characterized in that: Step S1 uses a C++ detection program implemented in C++ language, and uses the ONNX RunTime inference engine to deploy the deep learning model on the C++ side for inference to obtain the name of the game scene.

8. The adaptive interactive game strategy method based on deep learning detection as claimed in claim 1, characterized in that: Step S2 uses a JS visualization window program implemented in JS language; The game strategy is pre-stored in the back-end server; When the strategy window needs to be displayed in front, the strategy window is configured as a window adapted to the terminal screen, and the H5 page loaded with the game strategy is loaded.

9. The adaptive interactive game strategy method based on deep learning detection according to any one of claims 1 to 8, characterized in that: Step S1 adopts a C++ detection program implemented in C++ language, and step S2 adopts a JS visualization window program implemented in JS language. The game scene name obtained by the C++ detection program is transmitted to the JS visualization window program through an interactive interface.

10. An adaptive interactive game strategy system based on deep learning detection, characterized in that: include: A game scene detection module is used to periodically obtain the game screen, identify the scene Chinese characters in the game screen, and obtain the game scene name corresponding to the scene Chinese characters; The game strategy visualization module is used to obtain the game strategy corresponding to the game scene name. When a strategy call command is received, a strategy window is displayed in front of the current game screen, and the game strategy is displayed in the strategy window. When a strategy close command is received, the strategy window is closed.

Images