CN105263050B - Mobile terminal real-time rendering system and method based on cloud platform - Google Patents

Abstract

The invention discloses a mobile terminal real-time rendering system based on a cloud platform and a method thereof. The method includes receiving viewpoint information and interaction information transmitted from a mobile terminal, querying and reading models and scene files, and obtaining three-dimensional scene data; The type of model group of the scene, divide the 3D scene data, and obtain the data of the 3D scene model group; store the data of the 3D scene model group, and automatically adjust the storage location according to different data requirements in different 3D scenes; call the stored 3D scene model group Data, establish and manage the MIC/GPU rendering task of the 3D scene image, finally obtain the 3D scene rendering result data, and compress the obtained 3D scene rendering result data and send it to the mobile terminal; among them, use the dynamic load strategy to MIC/GPU GPU rendering task deployment and management to ensure load balancing of cloud servers. This method consumes the computing power and storage space of the mobile client to a minimum.

Description

Mobile terminal real-time rendering system and method based on cloud platform

technical field

The invention belongs to the field of image rendering, and in particular relates to a real-time rendering system and method for a mobile terminal based on a cloud platform.

Background technique

This year, the market size of China's mobile terminal online games has reached more than 20 billion yuan. 3D modeling and rendering has become the core technology concerned by related industries around the world. Among them, ray tracing-based rendering technology has become a hot spot for major game manufacturers to adopt and research in recent years due to its high-quality rendering effect.

Due to the limitation of scene complexity and computing power of mobile terminals, it is difficult to perform real-time scene ray tracing rendering on mobile terminals. However, in recent years, the development of network speed and cloud computing technology has solved the bottleneck of computing and network for real-time scene rendering of mobile terminals based on cloud environment. At present, foreign rendering engines are in an absolute monopoly position in the game industry, which limits my country's technological development in this field.

The mode of real-time scene rendering of mobile terminals based on the cloud environment is similar to conventional cloud computing, that is, the 3D program is placed on a remote server for rendering, and the user terminal accesses resources through Web software and high-speed Internet access, and commands are issued from the user terminal. The server executes the corresponding rendering task according to the instruction, and the rendering result screen is sent back to the user terminal for display.

The difficulty of real-time scene rendering of mobile terminals based on the cloud environment is far more complicated than that of conventional cloud computing applications, mainly due to the stringent requirements of 3D rendering on hardware performance and command response. However, the real-time scene rendering system for mobile terminals based on the cloud environment may face rendering requests from thousands of users, which will be a huge pressure on the back-end server system—compared with conventional cloud computing applications (such as From commercial programs such as Gmail and Google Docs to scientific computing), the real-time scene rendering of mobile terminals based on the cloud environment meets the needs of the same number of users, and the hardware performance required is at least several times to dozens of times higher than that of cloud computing.

Contents of the invention

In order to solve the shortcomings of the prior art, the present invention provides a real-time rendering system and method for a mobile terminal based on a cloud platform. The system makes full use of the supercomputing and storage capabilities of the cloud to perform real-time ray-tracing rendering of the scene. While achieving a highly realistic rendering effect, the compression algorithm meets the refresh rate of the client screen, and consumes the computing power and storage of the mobile client to a minimum. space.

To achieve the above object, the present invention adopts the following technical solutions:

A real-time rendering system for mobile terminals based on a cloud platform, including:

The central control unit deployed on the cloud server, the 3D scene management system and cluster, the distributed storage system and cluster, the MIC/GPU rendering management service system and cluster, and the load balancing system and cluster respectively connected to the central control unit;

The 3D scene management system and the cluster are used to divide the 3D scene data transmitted from the mobile terminal according to the type of the model group of the 3D scene to obtain the data of the 3D scene model group;

The distributed storage system and cluster are used to store the data of the three-dimensional scene model group, and automatically adjust the storage location according to different data requirements in different three-dimensional scenes;

The MIC/GPU rendering management service system and cluster are used to call the 3D scene model group data stored in the distributed storage system and the cluster, establish and manage the MIC/GPU rendering task of the 3D scene image, and finally obtain the 3D scene rendering result data, and compress the obtained 3D scene rendering result data and send it to the mobile terminal;

The load balancing system and cluster are used to deploy and manage MIC/GPU rendering task processes by using dynamic load strategies to ensure load balancing of cloud servers.

The rendering system also includes a network application service system and a cluster, which are used as a central control unit, a three-dimensional scene management system and a cluster, a distributed storage system and a cluster, a MIC/GPU rendering management service system and a cluster, and a load balancing system and Two clusters provide mutual communication.

The 3D scene management system and cluster include:

The 3D model and scene file loading unit is connected to the mobile terminal through the network application service system and the cluster, and is used to receive the viewpoint information and interaction information transmitted by the mobile terminal, query and read the model and scene file, and obtain the 3D scene data;

A three-dimensional model and scene file segmentation unit, which is used to divide the three-dimensional model and scene file into several model blocks according to the type of the model group of the three-dimensional scene, and obtain the data of the three-dimensional scene model group;

The 3D model block storage allocation unit is used to automatically allocate the data of the 3D scene model group to the distributed storage system and the cluster for storage.

The MIC/GPU rendering management service system and cluster include:

A three-dimensional scene model group data retrieval unit, which is used to retrieve the three-dimensional scene model group data stored in the distributed storage system and the cluster;

A rendering task creation module, which is used to create a task for rendering a three-dimensional scene;

A rendering task data loading module, which is used to load the retrieved 3D scene model group data into the established task of rendering a 3D scene;

A rendering task scheduling module, which is used to call rendering tasks, perform parallel ray tracing rendering on the loaded 3D scene model group data, and obtain 3D scene rendering result data;

Rendering result data management, which is used to compress the 3D scene rendering result data, and transmit the compressed 3D scene rendering result data to the mobile terminal for display.

A method for realizing a real-time rendering system of a mobile terminal based on a cloud platform, comprising:

Step (1): Receive the viewpoint information and interaction information sent by the mobile terminal, query and read the model and scene files, and obtain 3D scene data;

Step (2): According to the type of the model group of the 3D scene, divide the 3D scene data to obtain the data of the 3D scene model group;

Step (3): store the data of the 3D scene model group, and automatically adjust the storage location according to different data requirements in different 3D scenes;

Step (4): Call the stored 3D scene model group data, establish and manage the MIC/GPU rendering task of the 3D scene image, and finally obtain the 3D scene rendering result data, and compress the obtained 3D scene rendering result data before delivering To the mobile terminal; Among them, use the dynamic load strategy to deploy and manage the MIC/GPU rendering tasks to ensure the load balance of the cloud server.

In the step (1), the mobile terminal communicates with the cloud server.

The process of dividing the three-dimensional scene data in the step (2) is:

Step (2.1): divide the 3D scene data into several model groups according to preset units, and number these model groups;

Step (2.2): For each model group divided into, according to the density of different regional models in the model group, the group model is divided into several closed cube model blocks, the number of models in each model block is equal, and the model group Each model block is then numbered.

The process of obtaining the three-dimensional scene rendering result data in the step (4) includes:

Step (4.1): call the 3D scene model group data stored in the distributed storage system and the cluster;

Step (4.2): establish the task of rendering the 3D scene;

Step (4.3): Load the retrieved 3D scene model group data into the established task of rendering the 3D scene;

Step (4.4): Call the rendering task, perform parallel ray tracing rendering on the loaded 3D scene model group data, obtain the rendered image of the 3D scene, and then obtain the rendering result data;

Step (4.5): compressing the rendered image of the 3D scene, and transmitting the compressed image of the rendered 3D scene to the mobile terminal for display.

The process of compressing the rendered image of the 3D scene in the step (4.5) is:

Step (4.5.1): Using the HEVC coding framework, according to the flatness of the rendered image of the 3D scene, adaptively segment the rendered image of the 3D scene into several coding units, and determine the best segmentation scheme;

Step (4.5.2): Identify the segmentation scheme, select the optimal prediction mode, and use differential coding to perform preliminary prediction; wherein, the prediction mode includes intra-frame prediction and inter-frame prediction;

Step (4.5.3): According to the prediction mode, the residual scanning sequence of the rendered image of the 3D scene is determined, and the entropy coding is optimized.

In step (4.5.2), the prediction mode also includes spatial prediction, through the combination of time domain and space domain, the optimal prediction mode is selected pixel by pixel.

The beneficial effects of the present invention are:

(1) The mobile terminal real-time rendering system based on the cloud platform of the present invention deploys a three-dimensional scene management system and clusters, a distributed storage system cluster, and a MIC/GPU rendering management service system and clusters in the cloud server, and the system makes full use of these clusters The computing and storage capabilities of the server load data from the storage node into the memory of the currently allocated local computing unit in advance, reducing memory access delays, improving rendering efficiency, and finally realizing real-time ray tracing rendering of the scene;

(2) There are also load balancing systems and clusters deployed in the cloud servers. The load balancing systems and clusters use dynamic load strategies to deploy and manage MIC/GPU rendering tasks to ensure the load balancing of the cloud servers, thereby maximizing the use of cloud computing capabilities. rendering.

(3) After the present invention completes the rendering, the cloud compresses the video stream in real time, and with the help of a large number of computing cores of the MIC/GPU, the video stream is quickly and efficiently compressed to minimize its occupied volume, and the decompression is optimized. Adapting to the relatively weak computing power of the client, the client can decompress the transmitted video stream in a short period of time.

Description of drawings

FIG. 1 is a flow chart of the real-time rendering method for a mobile terminal based on a cloud platform in the present invention.

Fig. 2 is a schematic diagram of the principle of the real-time rendering method of the mobile terminal based on the cloud platform of the present invention.

Figure 3 is a flow chart of rendering technology implementation.

Fig. 4 is a flow chart of compression and transmission.

Figure 5 is a flow chart of user display and control.

Detailed ways

Below in conjunction with accompanying drawing and embodiment the present invention will be further described:

The mobile terminal real-time rendering system based on the cloud platform of the present invention can receive the user's interactive operation in real time, and update and render the game scene according to the interactive operation and the established program.

The working principle of the mobile terminal real-time rendering system based on the cloud platform of the present invention is shown in FIG. 1 .

The system makes full use of the supercomputing and storage capabilities of the cloud to perform real-time ray tracing rendering of the scene. While achieving a highly realistic rendering effect, the system meets the refresh rate of the client screen through efficient compression algorithms and communication strategies, and minimizes the consumption of mobile clients. terminal computing power and storage space.

In Figure 1, MIC/GPU high-performance computing clusters are deployed in the cloud. In order to make full use of its performance, the present invention uses a grid-based distributed ray tracing rendering technology to perform MIC/GPU rendering tasks by using a dynamic load strategy on the CPU Deployment and management, so as to maximize the use of cloud computing capabilities for rendering.

At the same time, scene division is introduced to accurately calculate the scene content that may appear in the current situation, and the data is loaded from the storage node to the memory of the currently allocated local computing unit in advance through the data prefetch method, which reduces memory access delay and improves rendering efficiency. Complex scenes are accurately divided, and based on the division, the extremely high-quality model is stored in the current optimal storage node through the distributed storage technology through the size of the model and the number of visits, and the storage location is determined according to different data requirements in different scenarios automatic adjustment.

After the rendering is completed, the cloud compresses the video stream in real time. Through a parallel and lossless video stream compression technology, with the help of a large number of computing cores of MIC/GPU, the video stream is quickly and efficiently compressed to minimize its occupied volume and solve the problem. The compression is optimized to adapt to the relatively weak computing power of the client, so that the client can decompress the transmitted video stream in a short period of time.

After the compression is completed, the video stream and some corresponding control streams will be transmitted through the high-speed network link. During the transmission process, the order of sending the streams will be determined according to the priority to ensure that the client does not need to wait for a long time to receive some current operations. data.

After receiving the required data, the client makes full use of its computing power to decompress the video stream, and uses the API supported by its own platform to synchronize the picture and sound and output in full screen. While outputting, it performs specified interaction with some specified peripherals according to some operation instructions provided by the control flow, and at the same time stores some video stream and audio stream data that may be reused in future situations. In some embodiments, overutilization of the limited computing power of the client is further reduced by storing reassembled and pre-rendered video or audio elements,

The specific structural framework of the mobile terminal real-time rendering system based on the cloud platform of the present invention is introduced in detail below:

As shown in Figure 2, the mobile terminal real-time rendering system based on the cloud platform of the present invention includes:

The central control unit deployed on the cloud server, and the 3D scene management system and cluster, distributed storage system and cluster, MIC/GPU rendering management service system and cluster, and load balancing system and cluster respectively connected to the central control unit;

3D scene management system and cluster, which are used to divide the 3D scene data transmitted from the mobile terminal according to the type of the model group of the 3D scene, and obtain the data of the 3D scene model group;

Distributed storage system and cluster, which are used to store the data of the 3D scene model group, and automatically adjust the storage location according to different data requirements in different 3D scenes;

MIC/GPU rendering management service system and cluster, which are used to call 3D scene model group data stored in distributed storage systems and clusters, establish and manage MIC/GPU rendering tasks for 3D scene images, and finally obtain 3D scene rendering result data , and send the obtained 3D scene rendering result data to the mobile terminal;

The load balancing system and cluster are used to deploy and manage MIC/GPU rendering tasks using dynamic load strategies to ensure load balancing of cloud servers.

Further, the rendering system of the present invention also includes a network application service system and a cluster, which are used for the central control unit, the three-dimensional scene management system and the cluster, the distributed storage system cluster, the MIC/GPU rendering management service system and the cluster and the load The balance system and the cluster provide mutual communication between the two.

Further, the 3D scene management system and cluster include:

The 3D model and scene file loading unit is connected to the mobile terminal through the network application service system and the cluster, and is used to receive the viewpoint information and interaction information transmitted by the mobile terminal, query and read the model and scene file, and obtain the 3D scene data;

A three-dimensional model and scene file segmentation unit, which is used to divide the three-dimensional model and scene file into several model blocks according to the type of the model group of the three-dimensional scene, and obtain the data of the three-dimensional scene model group;

The 3D model block storage allocation unit is used to automatically allocate the data of the 3D scene model group to the distributed storage system and the cluster for storage.

Furthermore, the MIC/GPU rendering management service system and cluster include:

A three-dimensional scene model group data retrieval unit, which is used to retrieve the three-dimensional scene model group data stored in the distributed storage system and the cluster;

A rendering task creation module, which is used to create a task for rendering a three-dimensional scene;

A rendering task data loading module, which is used to load the retrieved 3D scene model group data into the established task of rendering a 3D scene;

A rendering task scheduling module, which is used to call rendering tasks, perform parallel ray tracing rendering on the loaded 3D scene model group data, and obtain 3D scene rendering result data;

Rendering result data management, which is used to compress the 3D scene rendering result data, and transmit the compressed 3D scene rendering result data to the mobile terminal for display.

As shown in Figure 1, the implementation method of the mobile terminal real-time rendering system based on the cloud platform of the present invention includes:

Step (1): Receive the viewpoint information and interaction information sent by the mobile terminal, query and read the model and scene files, and obtain 3D scene data;

Step (2): According to the type of the model group of the 3D scene, divide the 3D scene data to obtain the data of the 3D scene model group;

Step (3): store the data of the 3D scene model group, and automatically adjust the storage location according to different data requirements in different 3D scenes;

Step (4): Call the stored 3D scene model group data, establish and manage the MIC/GPU rendering task of the 3D scene image, finally obtain the 3D scene rendering result data, and send the obtained 3D scene rendering result data to the mobile terminal ; Among them, the dynamic load strategy is used to deploy and manage the MIC/GPU rendering tasks to ensure the load balance of the cloud server.

In step (1), the mobile terminal communicates with the cloud server.

Take the game 3D scene as an example:

Distributed storage systems and clusters are deployed to cloud servers, and the parameters and performance of the distributed storage system are first debugged:

Set up two accounts in the distributed storage system: one is called the management account, and the other is the rendering account. The specific permissions are shown in Table 1 below. The management account is responsible for managing all stored 3D models, and the rendering account is responsible for calling the models to the computing unit for rendering. Configuring files in the distributed storage system environment enables the system to implement faster call operations under the current cluster.

Table 1 Permissions of accounts in the distributed storage system

The process of dividing the three-dimensional scene data in step (2) is:

Step (2.1): Divide the 3D scene data of the game into N model groups according to preset units, and number these model groups from 0 to N-1; wherein, the preset units are game levels;

Step (2.2): For each model group divided into, according to the density of different regional models in the model group, the group model is divided into M airtight cube model blocks, the number of models in each model block is equal, and the model group Each model nugget in is numbered from 0 to M-1.

Wherein, both M and N are positive integers greater than or equal to 1.

To store model groups in the cloud, it is necessary to ensure that the storage locations of the model blocks of any model group are adjacent or similar.

Further, the process of obtaining the 3D scene rendering result data in step (4) includes:

Step (4.1): call the 3D scene model group data stored in the distributed storage system and the cluster;

Step (4.2): establish the task of rendering the 3D scene;

Step (4.3): Load the retrieved 3D scene model group data into the established task of rendering the 3D scene;

Step (4.4): Call the rendering task, perform parallel ray tracing rendering on the loaded 3D scene model group data, obtain the rendered image of the 3D scene, and then obtain the rendering result data;

Step (4.5): compressing the rendered image of the 3D scene, and transmitting the compressed image of the rendered 3D scene to the mobile terminal for display.

As shown in Figure 3, before the parallel ray tracing rendering in step (4.4), the rendering system first uses the GPU and the visibility applicable to the ray tracing algorithm to calculate the visibility of the viewpoint; then, uses the MIC system according to the visibility The calculation results are further accurately divided for the 3D scene.

In the process of parallel ray tracing rendering, the rendering system of the present invention uses GPU to perform real-time and efficient ray tracing rendering on the segmented scene.

Among them, the specific process of calculating the visibility of the viewpoint is as follows:

Set the threshold, obtain the material information of the object, and calibrate it as a node when the reflectance of the material is greater than the set threshold;

Create a queue and store the initial position in the queue;

Divide the memory area to save the visible range required for this rendering;

Calculate the visible range of all nodes in the queue, delete the calculated nodes, save all objects in the visible range, if a node appears in the visible range, store the node in the queue; otherwise divide the memory area again to save the rendering Visibility required.

Furthermore, the process of compressing the rendered image of the 3D scene in step (4.5) is:

Step (4.5.1): Using the HEVC coding framework, according to the flatness of the rendered image of the 3D scene, adaptively segment the rendered image of the 3D scene into several coding units, and determine the best segmentation scheme;

Step (4.5.2): Identify the segmentation scheme, select the optimal prediction mode, and use differential coding to perform preliminary prediction; wherein, the prediction mode includes intra-frame prediction and inter-frame prediction;

Step (4.5.3): According to the prediction mode, the residual scanning sequence of the rendered image of the 3D scene is determined, and the entropy coding is optimized.

In step (4.5.2), the prediction mode also includes spatial prediction, through the combination of time domain and space domain, the optimal prediction mode is selected pixel by pixel.

Hierarchical prediction, template matching and adaptive prediction methods are also used for intra-frame prediction to reduce spatial redundancy and improve coding efficiency. In addition, coding efficiency is improved by inter prediction.

In step (4.5), the rendered image of the compressed 3D scene is transmitted to the mobile terminal for display, and a fast and low-delay user terminal display method is adopted, as shown in Figure 4, and the specific process is as follows:

Automatically detect the hardware information of the mobile terminal device and send it to the cloud, and the cloud intelligent analysis technology will customize and intelligently select the most suitable compression algorithm for different users according to their device computing power and link speed;

Video stream decompression is based on the HEVC framework itself, because the image is optimally segmented during the compression process;

The decompression process is decompressed according to the calculation results in the cloud. With good parallelism, the conversion of the relevant rendering instructions will be performed immediately after the decompression of the segmented primitives, instead of the overall conversion after the decompression of the entire image. One frame performs multiple state updates in the form of multiple batches;

Update the screen display according to the content in the video memory.

Introduce behavior prediction, predict user input, and communicate relevant information with the cloud. The cloud can use the information to update the view in advance. Once the accurate information of the user's operation is received, the wrongly predicted changes will be canceled. Combined with the predicted information, the real Changes are updated, and new predictions are made based on the user's new operating information.

The cloud recognizes and learns according to the operating habits of different users, and continuously improves the prediction accuracy during the interaction process.

In the process of link communication, it interacts with the operating system to increase the priority of the current application's network use, so as to minimize user delays.

Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it does not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.

Claims (7)

1. A mobile terminal real-time rendering system based on cloud platform, characterized in that, comprising: The central control unit deployed on the cloud server, the 3D scene management system and cluster, the distributed storage system and cluster, the MIC/GPU rendering management service system and cluster, and the load balancing system and cluster respectively connected to the central control unit; The 3D scene management system and the cluster are used to divide the 3D scene data transmitted from the mobile terminal according to the type of the model group of the 3D scene to obtain the data of the 3D scene model group; The distributed storage system and cluster are used to store the data of the three-dimensional scene model group, and automatically adjust the storage location according to different data requirements in different three-dimensional scenes; The MIC/GPU rendering management service system and cluster are used to call the 3D scene model group data stored in the distributed storage system and the cluster, establish and manage the MIC/GPU rendering task of the 3D scene image, and finally obtain the 3D scene rendering result data, and compress the obtained 3D scene rendering result data and send it to the mobile terminal; The load balancing system and cluster are used to deploy and manage the MIC/GPU rendering task process by using a dynamic load strategy to ensure load balancing of cloud servers; The rendering system also includes a network application service system and a cluster, which are used as a central control unit, a three-dimensional scene management system and a cluster, a distributed storage system and a cluster, a MIC/GPU rendering management service system and a cluster, and a load balancing system and Provide mutual communication between clusters; The 3D scene management system and cluster include: The 3D model and scene file loading unit is connected to the mobile terminal through the network application service system and the cluster, and is used to receive the viewpoint information and interaction information transmitted by the mobile terminal, query and read the model and scene file, and obtain the 3D scene data; A three-dimensional model and scene file segmentation unit, which is used to divide the three-dimensional model and scene file into several model blocks according to the type of the model group of the three-dimensional scene, and obtain the data of the three-dimensional scene model group; A three-dimensional model block storage allocation unit, which is used to automatically allocate the data of the three-dimensional scene model group to the distributed storage system and the cluster for storage; The MIC/GPU rendering management service system and cluster include: A three-dimensional scene model group data retrieval unit, which is used to retrieve the three-dimensional scene model group data stored in the distributed storage system and the cluster; A rendering task creation module, which is used to create a task for rendering a three-dimensional scene; A rendering task data loading module, which is used to load the retrieved 3D scene model group data into the established task of rendering a 3D scene; A rendering task scheduling module, which is used to call rendering tasks, perform parallel ray tracing rendering on the loaded 3D scene model group data, and obtain 3D scene rendering result data; Rendering result data management, which is used to compress the 3D scene rendering result data, and transmit the compressed 3D scene rendering result data to the mobile terminal for display. 2. an implementation method of a mobile terminal real-time rendering system based on cloud platform as claimed in claim 1, characterized in that, comprising: Step (1): Receive the viewpoint information and interaction information sent by the mobile terminal, query and read the model and scene files, and obtain 3D scene data; Step (2): According to the type of the model group of the 3D scene, divide the 3D scene data to obtain the data of the 3D scene model group; Step (3): Store the data of the 3D scene model group, and automatically adjust the storage location according to different data requirements in different 3D scenes; Step (4): Call the stored 3D scene model group data, establish and manage the MIC/GPU rendering task of the 3D scene image, finally obtain the 3D scene rendering result data, and compress the obtained 3D scene rendering result data before sending To the mobile terminal; Among them, use the dynamic load strategy to deploy and manage the MIC/GPU rendering tasks to ensure the load balance of the cloud server. 3 . The method for implementing a real-time rendering system for mobile terminals based on a cloud platform according to claim 2 , wherein, in the step (1), the mobile terminal communicates with the cloud server. 4 . 4. A method for realizing a real-time rendering system for mobile terminals based on a cloud platform as claimed in claim 2, wherein the process of dividing the three-dimensional scene data in the step (2) is: Step (2.1): divide the 3D scene data into several model groups according to preset units, and number these model groups; Step (2.2): For each model group divided, divide the model group into several closed cube model blocks according to the density of different regional models in the model group, and the number of models in each model block is equal. Divide the model group Each model block is then numbered. 5. The method for realizing a real-time rendering system of a mobile terminal based on a cloud platform as claimed in claim 2, wherein the process of obtaining 3D scene rendering result data in the step (4) includes: Step (4.1): Retrieve the 3D scene model group data stored in the distributed storage system and cluster; Step (4.2): Establish the task of rendering the 3D scene; Step (4.3): Load the retrieved 3D scene model group data into the established task of rendering the 3D scene; Step (4.4): Call the rendering task, perform parallel ray tracing rendering on the loaded 3D scene model group data, obtain the rendered image of the 3D scene, and then obtain the rendering result data; Step (4.5): Compress the parallelized ray tracing rendering image obtained based on step (4.4), and transmit the compressed image to the mobile terminal for display. 6. A method for realizing a real-time rendering system for a mobile terminal based on a cloud platform as claimed in claim 5, wherein the process of compressing the rendered image of the 3D scene in the step (4.5) is as follows: Step (4.5.1): Using the HEVC coding framework, according to the flatness of the rendered image of the 3D scene, adaptively segment the rendered image of the 3D scene into several coding units, and determine the best segmentation scheme; Step (4.5.2): Identify the segmentation scheme, select the optimal prediction mode, and use differential coding for preliminary prediction; where the prediction mode includes intra-frame prediction and inter-frame prediction; Step (4.5.3): According to the prediction mode, the residual scanning order of the rendered image of the 3D scene is determined, and the entropy coding is optimized. 7. The implementation method of a real-time rendering system for mobile terminals based on cloud platform as claimed in claim 6, characterized in that in step (4.5.2), the prediction mode also includes spatial prediction, through time domain and space domain Combined, the optimal prediction mode is selected pixel by pixel.