Graphic Generation technology - Dr. Xuesong Tian
Acceleration of graphics generation in virtual reality technology
Abstract: Virtual Reality technology, English Virtual Reality, referred to as VR. It is an ideal man-machine communication technology. The three essential characteristics of this technology are immersion, interactivity and imagination. The graphics generation technology is one of the core technologies in virtual reality. This paper introduces and summarizes the real-time graphics generation technology and its development in detail. The main content of this paper is hardware architecture of graphics generation and various effective techniques for acceleration in realistic graphics generation of virtual reality.
Keywords: virtual reality technology, graphics generation, graphics acceleration
1 the concept of virtual reality technology
The most essential characteristics of virtual reality technology can be summarized by three I, namely, Immersion, Interaction and Imagination. Immersion is the degree to which the user feels real in the virtual world. Interactivity refers to the degree to which users can manipulate objects in the virtual world and the real degree to which users can get feedback from the virtual world. Virtual reality technology as a new man-machine communication technology, different from the traditional mode of keyboard and mouse, using digital helmet, digital gloves and other complex sensor equipment, three-dimensional interaction technology, voice recognition, voice input technology has become an important means of human-machine interaction. In recent years, virtual reality technology has achieved rapid development, showing some new characteristics and development trends, mainly: virtual reality modeling technology, real-time THREE-DIMENSIONAL graphics generation and haptic interaction technology, large-scale network distributed virtual reality research.
2. 3D graphics generation technology
Immersive features require computers to create three-dimensional virtual environments that look, sound and feel real. Undoubtedly, the application of feasible technology that makes the virtual environment as immersive is the necessary condition for the success of virtual reality. This includes the development and application of advanced interactive, video and other technologies. For the virtual environment of virtual reality, the real-time dynamic graphic visual effect is the primary condition to produce the feeling of reality, without which there is no "reality" at all. So the graphics fast generation technology becomes the key technology of virtual reality.
2.1 Types of geometric modeling
In VR the geometrical model of objects in the virtual environment with traditional CAD modeling and animation are essential difference between the latter and modelling is given priority to in order to improve the fidelity must increase the complexity of geometric shape and VR 3 d graphics generation needs to both real-time and realistic reason is adopted to reduce the complexity of the modelling and enhance texture display object geometric modeling method which includes rules and irregular objects Geometric modeling and irregular fuzzy object geometric modeling.
2.2 Geometric description of objects
Modeling of object shape and geometric properties is a huge and rich research field. Some approaches pursue modeling accuracy in order to accurately describe the geometry of objects in the real world; Others pursue simplicity of expression. In terms of portraying
Speed. Most VR systems sacrifice detail and precision to achieve simplicity.
The simplest object is a one-dimensional point, followed by a two-dimensional vector. Many CAD systems generate and exchange data in a two-dimensional view. 2D information is of little use to VR systems, except when displayed on a two-dimensional surface in a virtual world. Some programs can reconstruct a three-dimensional model of an object from a given two or more two-dimensional views. The choice of modeling method is closely related to the depiction method chosen. Some descriptors can handle multiple types of models, but most can only handle one, especially for VR. The complexity of the model is inversely proportional to the drawing speed. The more complex the modeling, the more details, and the slower the frame speed .
3 graphics generation acceleration
3D graphics generation technology develops rapidly, and people have rich experience, mainly can form corresponding graphics in real time according to the requirements. In view of this situation, people need to ensure that the graphics refresh rate of this technology is 30 frames /s in the research process. On the basis of ensuring the integrity and quality of the graphics, the main research purpose of this technology is to make the refresh speed faster.
3.1 Judgment of visibility
Due to the directionality of sight, the limitation of visual Angle and mutual occlusion of objects, what the human eye sees is often only a part of the scene, and the generation of graphics is the transformation from object space to image space. In order to make full use of the limited resources of rendering hardware, it is necessary to make full use of the relativity of object space. The correlation of image space and time space can speed up the judgment of visibility and reduce the depth of drawing.
For a very complex scene, the detailed description of objects contains a large number of polygons, which requires an effective data structure to compose the scene to speed up the judgment of visibility . The development of hierarchical representation of objects in the late 1970s greatly accelerated the determination of visibility. The main methods of hierarchical representation are bounding box technology and octree technology. The main characteristic of these two methods is that the scene is organized into a tree, and the spatial coherence is fully utilized to speed up the scene traversal, thus greatly reducing the spatial complexity of the drawing process. In real-time roaming, the algorithm uses the predicted results to manage the scheduling of memory and external memory, and computs the visibility set from the superset, thus effectively accelerating the calculation of visibility. However, the prediction calculation of this algorithm is complicated, and it is limited to the axial building model.
3.2 Level of detail model
Providing different levels of detail for objects is a very effective way to control the complexity of a scene. As we know in life, as an object gets further and further away, people no longer recognize many of the details of the object, such as the smooth surface. Therefore, when drawing a distant object, it is completely unnecessary to describe a very complex model with details of the object, which is bound to waste graphics processing resources and processing time . In 1987, it was pointed out that when the object only covers a small area of the screen, it can be used to draw the model with a thick description of the object, so as to achieve fast and effective rendering of complex scenes. At the same time, the algorithm gives the approximate set of seed slices according to the user-defined error, and modifies the approximate set of seed slices according to the continuously merged seed slices until the set is empty. Once you have a detail level model of the object, how do you choose the current level description of the object? According to the visual characteristics of people, mainly based on the following criteria: (1) according to the distance between the object and the point of view and the size of the space occupied by the object in the projection plane to choose, this is the most commonly used method; ② According to the Angle between the object and the direction of sight to choose, because the image of the retina center can see more clearly, and the peripheral vision is easier to perceive the movement of the object; (3) The description of the fine degree of the object is determined by whether there is relative motion between the person and the object. Because of the sudden motion, how to achieve smooth transition of different fine models remains to be solved. The current effective method is to further refine the local area by using the improved fractal spatial displacement interpolation method, so as to solve the problem of the establishment and real-time display of the detail level model of the surface model.
3.3 Prediction calculation
Predictive computing is a way to reduce system latency. According to the speed and acceleration of various movements, such as the rotation speed of human head, the input of head tracking system and other input devices can be estimated by prediction and extrapolation before drawing the next frame, so as to reduce the delay caused by input devices. The human body test shows that the maximum rotation rate of the human head is 100 degrees/second, so it can be predicted that the change of the direction of the human eye will not exceed 3~5 degrees in the next frame . Although prediction method based on real-time test data can be effectively solve the problem of delay, but the system for the correct response to a user action depends on the correct forecast, on the basis of the practice, people often use of all kinds of simulated data extrapolation method income changes, such as position and viewpoints of the observer the location and direction of gaze direction, various kinds of sports to reduce the delay. For example, the system guesses the user's viewpoint position and line of sight direction according to the prediction, but when judging that the user can meet according to the visual body, the system increases the normal visual body Angle by 5 degrees according to the test result of the rotation rate of the human head. After the judgment of the visibility of the viewing body is increased, the system can draw a graph according to the more accurate position and direction of the viewing point input by the peripheral. This method not only overcomes the delay caused by the input device, but also solves the error caused by the prediction. In distributed network virtual reality system, the system can reduce the load of network transmission and the delay caused by network transmission according to the principle of prediction and estimation.
3.4 Offline computing
Offline computing is one of the common methods to reduce the load of real-time computing. Since VR system is a multi-task simulation system, it is necessary to pre-calculate and store some pre-calculated results in the corresponding structure as much as possible, including the calculation of global illumination model and dynamic model, etc. The hardware capability of feature processing is far from enough to meet the needs of VR, so it is almost impossible to generate realistic features in real time under current conditions. For most VR applications, the simple local lighting model is still adopted, and the low-quality feature display is realized with the support of hardware components. However, from the perspective of future development, the use of global lighting model and display of realistic graphics with complex scenes must be the goal that VR should pursue. In the case of global lighting models, ray-tracking technology is almost impossible to be applied in VR due to the large amount of computation and the difficulty of prediction due to the inherent viewpoint-related results generated.  In contrast, radiometric technology is the most promising tool for realistic graphics generation because its calculation results are independent of viewpoint and a large number of calculations can be used for prediction.
Current situation of graph generation technology
The limitation of hardware foundation, a high quality virtual reality system, must require higher specifications of external equipment, equipment accuracy and comfort must be improved. Graphics generation technology is not mature, in the virtual reality scene graphics change must be consistent with human action changes, but at present our graphics generation technology is still relatively backward, the conversion rate is not high enough, graphics conversion is not ideal. The amount of 3D data is large, and virtual reality technology needs many models, so it occupies a large space and transmits more data. Current broadband speeds are extremely difficult to keep up with and can load computers.
The graphics generation problem solves the problem of object modeling very well. The acceleration of graphics generation accelerates the speed of modeling, improves the modeling efficiency, and makes the environment more real. As one of the core technologies of virtual reality technology, it fits the reality of virtual reality technology, and it can be applied to many other fields. Just like other emerging science and technology, virtual reality technology is the product of many related disciplines. We must clearly recognize that although the technical potential of this field is huge and the application prospect is very broad, there are still many unsolved theoretical problems and technical guarantees that have not been overcome. Nevertheless, its unique advantages provide a new breakthrough for the development of various fields. At the same time, I believe that with the rapid development of network technology, virtual reality technology will get more rapid development, will be widely used in human daily life, to bring a new look to people's life.
 Yu Hongchuan, WU Fuchao, RUan Zongcai, et al. Research on Several Key Technologies of VR Environment image generation [J]. Computer Research and Development, 1999(11):1349-1357.
 ZUO Zhiqiang, HU Shiyun, LIANG Jianqi, et al. Research on New Scene Reproduction System Based on Computer Graphics and VR Technology [J]. Computer and Information Technology, 2019, 27(02):24-26.
 Cao Shuo. Research and design of graphic rendering and visual communication in VR system [J]. Consumer Guide, 2019, 000(033):45.
 Chen Yunhai. Research on VR Panoramic Image Processing Technology [J]. Guangdong Communications Technology, 2018, V. 38; No.308(08):42-47.
 Rao Cheng, Li Xun-xiang, Chen Dingfang. Application of detail Level technology in VR Visual System [J]. Journal of Hubei University of Technology, 2005, 20(003):52-54.