CN-122019373-A - Game testing method and system based on virtual reality

Abstract

The invention discloses a game testing method and system based on virtual reality. The method comprises the steps of determining optimal layout of a plurality of virtual cameras through a self-adaptive view angle prediction algorithm, generating a panoramic test view, constructing a multi-dimensional operation data structure, recording test operation at high frequency, realizing compression storage, collecting performance data, generating visual performance distribution thermodynamic diagrams through spatial interpolation, realizing natural gesture interaction and multi-user collaborative annotation based on hand feature recognition, and deploying a plurality of AI models to automatically detect visual, behavioral and time sequence anomalies. The system comprises six modules, namely VR environment construction, operation record analysis, performance visualization, cooperative interaction, intelligent detection and data storage, and the six modules work cooperatively through a standardized interface. The invention converts the traditional planar game test into an immersive VR test environment, solves the technical problems of limited visual field, inaccurate operation record, non-visual performance analysis and low cooperation efficiency, and provides an innovative technical means for game quality assurance.

Inventors

• WANG YING
• CAO YIFAN

Assignees

• 上海日兮禾戈信息科技有限公司

Dates

Publication Date

20260512

Application Date

20260120

Claims (6)

1. 1. The game testing method based on virtual reality is characterized by comprising the following steps of: S1, generating a multi-view panoramic test picture, namely acquiring game rendering data through a graphic API interception technology, comprehensively evaluating three dimensions of motion complexity, content richness and historical problem density by adopting a self-adaptive view prediction algorithm to determine the optimal layout of 3 to 7 virtual cameras, and fusing the multi-view picture into a 270-degree panoramic view through an image stitching technology; S2, recording operation track data, namely constructing four-dimensional operation tensors comprising space, channels and time dimensions, collecting input data of a mouse, a keyboard, a game handle and a VR controller at a sampling rate not lower than 120Hz, compressing and storing the input data through a tensor decomposition technology, and converting the operation data into three-dimensional tracks for visualization in the VR space; S3, calculating performance thermodynamic distribution, namely periodically collecting GPU rendering index, CPU service condition and memory resource consumption data, constructing a continuous three-dimensional performance field through a spatial interpolation algorithm, mapping the performance data into color-coded thermodynamic diagrams, and superposing and displaying the thermodynamic diagrams on a game scene; S4, realizing gesture annotation interaction, namely extracting joint angles, fingertip distances, palm orientations and motion characteristics based on a hand skeleton model, recognizing six predefined gestures through a deep learning network, and maintaining data synchronization of multi-user collaborative annotation by adopting a distributed consistency algorithm; S5, intelligent problem detection is carried out, namely three detection models of visual abnormality detection, behavior abnormality detection and time sequence analysis are deployed, rendering errors, game logic abnormality and time related problems are respectively identified, and a unified detection report is generated through weighted fusion.
2. 2. The method according to claim 1, wherein the adaptive view prediction algorithm: The motion complexity evaluates the dynamic change degree of the scene by analyzing the pixel motion velocity distribution between adjacent frames; the content richness quantifies the visual information richness by calculating the information entropy of the image area; The history problem density converts discrete history problem positions into continuous probability distribution through a kernel density estimation method; and determining the virtual camera layout by adopting a heuristic optimization algorithm under the condition of meeting the constraint of view coverage maximization and overlapping degree.
3. 3. The method of claim 1, wherein the four-dimensional operation tensor comprises: Space dimension corresponding to screen resolution; 16 operation channels for respectively recording the mouse coordinates and keys, key of the main function of the keyboard, joystick rocker and trigger, VR controller speed and operation intensity comprehensive index; Recording an operation sequence in a time dimension; And adopting a tensor compression method based on singular value decomposition, and realizing data storage with high compression rate by reserving a main characteristic mode.
4. 4. The method of claim 1, wherein the performance thermodynamic distribution calculation comprises: defining a comprehensive performance index, and combining a frame rate, memory use and rendering load by weighted powers; determining influence weight according to the spatial distance and the performance similarity of the sampling points by adopting a radial basis function interpolation method; the influence radius is adaptively adjusted according to the density of the scene objects, and a smaller influence radius is used in the object dense area; interpolation parameters are determined by solving a system of linear equations containing regularization terms.
5. 5. The method of claim 1, wherein in the gesture annotation interaction: Extracting multidimensional features including joint bending angle, fingertip relative distance, palm space orientation and hand movement speed from 21 hand key points; processing time sequence characteristics by using a recurrent neural network comprising a long-short-term memory unit to realize gesture classification; Maintaining the causal relation of operation through a vector clock mechanism, and ensuring consistency in a distributed environment; When the annotation position conflicts, spatial adjustment is performed by shifting a fixed distance in the direction of the conflict.
6. 6. A virtual reality-based game testing system, comprising: the VR environment construction module is used for acquiring game rendering data, executing a view angle optimization algorithm to generate a plurality of virtual camera views, generating a panoramic view through an image fusion technology and outputting the panoramic view to the VR display device; the operation record analysis module is used for collecting input equipment data at high frequency, constructing a multidimensional operation data structure, reducing storage overhead through a data compression algorithm and generating a visualized operation track; the performance visualization module is used for collecting multi-level performance indexes, constructing continuous performance distribution through spatial interpolation and generating a color coding performance thermodynamic diagram; The collaborative interaction module is used for identifying gesture input, managing multi-user labeling operation and maintaining data consistency through a distributed algorithm; the intelligent detection module is used for running a plurality of special detection models, identifying different types of game problems and generating a comprehensive detection report; and the data storage module is used for managing the persistent storage of the test data, providing multidimensional index service and supporting a data export function.

Description

Game testing method and system based on virtual reality Technical Field The invention belongs to the technical field of game testing, and particularly relates to a game testing method and system based on virtual reality. Background With the vigorous development of the game industry, the complexity of game software is continuously improved, and higher requirements are put on game testing. The game test is a key link for ensuring the game quality and improving the user experience, and directly influences the market performance and the user satisfaction of the game product. The traditional game test method mainly relies on a tester to perform manual test through standard display equipment, and has the following technical pain points: 1. limited field of view coverage The tester is limited by the fixed viewing angle of the display, and can only observe partial pictures, and can cover the whole content only by frequently switching the viewing angle. Problems such as rendering errors of screen edges and view angle blind areas, mapping anomalies and the like are easily missed. Especially in large open world games, visual-type bugs often appear in non-home view areas. 2. Difficulty in operation reproduction The existing input recording mode is insufficient in time precision and lacks operation context, force and speed information. After the tester finds the bug, the developer often needs to try for many times to successfully reproduce the problem, and the problem repairing efficiency is seriously affected. The problems of complex operating sequences and specific timing triggers are particularly difficult to reproduce. 3. Performance bottleneck positioning is not intuitive Traditional performance monitoring outputs abstract numerical indexes (FPS, CPU occupancy rate, memory usage amount and the like), and visual connection with specific scene positions is difficult to establish. Determining specific areas and objects with reduced performance requires repeated testing and investigation, which is time-consuming and laborious. 4. Multiple people cooperate with low efficiency Team collaboration relies on screenshot, screen recording and text descriptions, with information distortion and understanding bias. The remote testers cannot observe the same game state at the same time, the problem confirmation period is obviously prolonged when the time zone is crossed for cooperation, and the communication cost is high. 5. The test experience is difficult to inherit. Disclosure of Invention In order to solve the technical problems in the prior art, the invention provides a game testing method and system based on virtual reality, which aim to realize omnibearing coverage, accurate operation record, visual performance analysis, high-efficiency team cooperation and intelligent auxiliary detection of game testing through the space display capability, interaction advantage and data visualization characteristics of the virtual reality technology, thereby remarkably improving the efficiency and quality of game testing. In order to achieve the aim of the invention, the invention adopts the following technical scheme: On one hand, the invention provides a game testing method based on virtual reality, which realizes comprehensive improvement of testing capability by converting a traditional planar game test into a three-dimensional immersive VR testing environment. The method comprises the following steps: step S1, generating a multi-view panoramic test picture In this step, scene data, including depth information, geometry information, and rendering states, is acquired through a rendering pipeline that intercepts the game. Based on the scene characteristics analyzed by the self-adaptive visual angle prediction algorithm, three dimensions of motion complexity, content richness and historical problem distribution are comprehensively considered, and an optimal virtual camera layout scheme is dynamically determined. The motion complexity is evaluated by analyzing the degree of change between adjacent frames, and the important attention is paid to the areas with dense dynamic objects and frequent scene switching. The content richness adopts an information theory method to quantify the information quantity of different areas, so that the important game elements are fully covered. The historical problem distribution identifies high risk areas that are prone to problems based on the accumulated test data. After the virtual camera position is determined, game pictures of a plurality of view angles are generated by modifying rendering parameters. And precisely aligning the multi-view images by adopting a characteristic matching and image registration technology, and realizing natural transition in an overlapping area by using a multi-band image fusion method to finally generate a continuous and seamless panoramic view. The generated panoramic picture is transmitted to the head display equipment in real time through the VR interface, and the displ