CN100428218C - A Method of Realizing General Virtual Environment Roaming Engine - Google Patents

Abstract

The invention belongs to the technical field of computer virtual reality. Its implementation method includes creating general virtual reality application resources, loading scene database files conforming to the scene description specification, customizing the roaming state mechanism, setting the input mapping and interpretation mechanism and accepting an external input in each frame of the simulation cycle, supporting multiple inputs The device is connected to the roaming system, and the viewpoint control, entity manipulation and state setting are performed according to external input commands. During the viewpoint control process, various scene complexity reduction strategies are implemented, collision detection and terrain matching are performed, and standard two-dimensional input devices are supported to select and Procedures for manipulating objects in a 3D scene, setting various roaming control states and environmental effects. The dual-camera model is used for scene observation, collision detection and roaming control, supported by various scene complexity reduction strategies, and the input mapping mechanism is provided, with complete functions and clear interfaces, which can facilitate the expansion of input devices, etc. It is suitable for a user to develop a virtual reality environment system.

Description

A Method of Realizing General Virtual Environment Roaming Engine

technical field

The invention belongs to the technical field of computer virtual reality, in particular to a method for realizing a general virtual environment roaming engine.

Background technique

Due to the important characteristics of urban planning, virtual construction, and scene roaming based on virtual reality technology and the huge social and economic benefits that will be generated, as well as the fundamental changes brought about by related industries, some western developed countries have invested a lot of money since the mid-1980s and personnel to support research in virtual reality. Among them, American Advanced Technology Center, Atlanta Design Company, University of North Carolina, Paradim Corporation and other institutions have achieved many achievements in theory and practice, and have begun to launch preliminary practical and relatively expensive special commercial software toolkits.

At present, two different research directions have formed in the frontier of 3D scene roaming, that is, geometry-based scene rendering using measurable size data and virtual scene representation and rendering based on source image sequences. Geometry-based virtual reality developed earlier. With the rapid development of computer technology, many companies and scientific research institutions have launched a variety of graphics accelerator cards in hardware, 3D modeling tools and 3D graphics environments in software. For example, the performance-price ratio of the graphics accelerator card of American Visual Computing Technology Co., Ltd. has been greatly improved; the modeling software MultiGen II Pro and its database format OpenFlight of Maltejin Company have almost become the industrial standard in the simulation field; VEGA ( Vig), Performer (performer) and other graphics systems have developed into powerful, easy-to-use commercial software. However, virtual reality technology has not been developed for a long time after all, and there are still many problems to be solved in geometry-based virtual reality technology. Relatively speaking, image-based virtual reality technology started late, and its main research direction is to reduce the complexity and labor intensity of scene modeling based on geometric virtual reality technology, and reduce the harsh requirements of virtual reality technology system for hardware performance. However, there are still many theories to be further studied in image-based virtual reality technology in terms of virtual entity manipulation, human-computer interaction with force feedback, and collision detection and response with physical characteristics.

The Scene Walking Studio of the University of Carolina at Berkeley is one of the earliest scientific research institutions in the world engaged in the study of scene roaming and has achieved outstanding results. They began to conduct research on real-time roaming strategies for complex models in 1990. In 1996, real-time roaming of Soda Hall (Soda Hall), a new department building of the Department of Computer Science, was realized on the Power series 320 workstation of SGI (American graphics workstation manufacturer). The Soda Hall model consists of 1,418,807 polygons, occupying 21.5 megabytes of hard disk space. The model uses 406 materials and 58 different textures. Because the research team adopted various technologies such as efficient data storage structure, multi-level detail technology, scene scheduling algorithm, real-time visual area determination algorithm and pre-calculation processing, the real-time simulation rate of Soda Hall, that is, the refresh rate per second is constant. Around 20 frames per second. During years of research work, the Scene Roaming Studio of the University of North Carolina proposed and continuously improved the UNIGRAFIX visual database, and developed some software toolkits, such as data converters for converting AutoCADDXF data format to UNIGRAFIX format, and object multi-level detail technology Automatic generator and simple scene automatic generation tool, etc. But for China, it is actually impossible to reuse a complete roaming engine at the code level.

There are also some domestic research institutions engaged in the research of scene roaming technology, among which the Forbidden City roaming realized by the Industrial Psychology Laboratory of Hangzhou University is the representative. The Forbidden City tour uses bicycles as an interactive tool, allowing the roamer to ride in the virtual Forbidden City without moving. The Forbidden City Tour mainly shows the appearance of the building, that is, the outdoor scene. The indoor part has been greatly simplified, and the modeling method based on texture mapping technology is adopted. The functions of collision detection and collision response are relatively weak. The system does not provide Manipulation capabilities for virtual entities.

Contents of the invention

In order to overcome the above-mentioned shortcomings, the purpose of the present invention is to provide a method for realizing a general-purpose virtual environment roaming engine, which can realize basic but important common functions in virtual environment roaming, and better encapsulate a batch of key technologies, reducing the It is difficult for ordinary users to develop virtual reality application systems, and the reuse of system functions is realized at the code level, which is not only conducive to standardized development, but also can reduce development cycles and costs.

In order to achieve the above object, the implementation method of the present invention includes a personal computer, a graphics accelerator card, a walker, a stereo vision display and tracking device, a scene database and a roaming engine core component, and its implementation method includes: creating a general virtual reality application resource , load the scene database file that conforms to the scene description specification, customize the roaming state mechanism, set the input mapping and interpretation mechanism and accept an external input in each frame of the simulation cycle, support multiple input devices to access the roaming system, according to the external input command Perform viewpoint control, entity manipulation and state setting, implement various scene complexity reduction strategies, perform collision detection and terrain matching in the viewpoint control process, support the use of standard two-dimensional input devices to select and manipulate objects in three-dimensional scenes, and set various Roaming control status and environmental effects, steps to meet the needs of different users.

Viewpoint control is a viewpoint control using a dual camera model for roaming. Create an observation viewpoint camera and a walking viewpoint camera. The observation viewpoint camera is limited by the walking viewpoint camera. The walking camera is used to realize multi-point collision and maintain the direction of travel during roaming. The camera has three rotational degrees of freedom; viewpoint control further includes judging the nature of the collision if a collision is detected during the movement of the viewpoint camera and the walking camera, and if it is a scene control surface, a scene scheduling strategy is activated to schedule and manage indoor and outdoor scenes , if it is a manipulable moving entity, the motion control of the entity in the scene is carried out according to the motion coordinate system and motion parameters defined in the model. If it is confirmed that a collision has occurred, the steps of the collision response are given; the input mapping and interpretation mechanism is used to Isolate the roaming engine from the input device to minimize the impact of the input device on the core of the roaming system, abstract an intermediate layer from the viewpoint control and device functions, map the control of the input device to the intermediate layer and make it a viewpoint Driven control source; during the entire roaming process, it supports multiple scene height and complexity reduction strategies, including scheduling the accuracy replacement of multi-level detail models according to the distance between the observation camera and each scene model, using the three-dimensional space in the scene database Plane and behavior parameters determine the scene control surface, and schedule and divide indoor and outdoor scenes and complex indoor scenes. During the roaming process, after the line-of-sight-based collision detection algorithm detects the model control surface, it analyzes the control behavior identifier defined during modeling , mapped to the scene control object set, according to the object list in the control object set, load visible objects, unload invisible objects, realize scene scheduling management, automatically reduce redundant polygons in the solid model when the model is loaded, and generate the texture of the solid surface; collision The detection and response further include analyzing the cause of camera movement, establishing a forward collision detection line segment between the walking camera and the observing camera, detecting the collision of the observing camera, and detecting the collision of the walking camera. If a collision occurs, the virtual environment scene scheduling process is performed; , then the camera moves forward, and the step of terrain matching is performed; the terrain matching technology further includes walking the camera forward, if the position of the viewpoint changes, the collision detection module is called, and the result of the collision detection is a collision, then ignore the change of the viewpoint position, the walking camera and the observation camera Raise a certain height, and at the same time set the viewing direction of the observation camera according to the direction of the terrain surface. If the result is no collision, the viewpoint reaches a new position; the status setting function includes turning on/off the fog effect, turning on/off the two-dimensional map One or more of wizards, on/off collision detection, selection of transparent processing methods, setting weather conditions, etc.; environmental effects include the steps of the constant frame rate method supported by atomization; the constant frame rate method further includes taking out the first 2 Calculate the average time spent per frame in seconds, calculate the current frame rate, and calculate the difference between the current frame rate and the target frame rate. If the current frame rate is greater than the target frame rate, increase the fog visible distance, otherwise reduce the fog visible distance, and set a new value. atomization parameters and end.

The characteristics of the present invention are: 1. The function is complete and the interface is clear. Using the invention to develop a scene roaming system, the workload is equal to creating a scene database, and the roaming driving part hardly needs to rewrite codes. 2. The dual-camera model is used for scene observation, collision detection and roaming control. The dual-camera model can not only effectively solve the problem of collision detection of low obstacles on the ground, but also better solve the problem of sight direction for terrain matching on sloped roads. 3. Support for various scenario complexity reduction strategies, especially the original scheduling algorithm based on the scenario control plane. The algorithm guarantees the continuous roaming problem of combined indoor and outdoor scenes, and can effectively improve the real-time roaming. At the same time, the algorithm is simple and easy to master and use. 4. Following the virtual reality industry data standards, the roaming engine encapsulates various descriptions and interpretations of the virtual scene, such as the definition of the movement of doors and windows and the behavior characteristics of objects above the ground for roaming. If it is a step, it can perform terrain Match, so that the viewpoint is automatically raised, if it is a green fence, it cannot pass through, and it needs to go around and so on. 5. The mechanism of input mapping is given, and it supports multiple input devices such as mouse, keyboard, joystick, walker, helmet tracker, etc., and can easily expand the access of input devices. 6. Realize the constant frame rate technology based on fogging, control the number of visible geometric patches in the scene by adjusting the fog concentration, and then adjust the roaming frame rate.

Compared with the prior art, the virtual environment roaming engine computer system of the present invention has the beneficial effects that it is a relatively complete roaming engine, which realizes input mapping and viewpoint control, virtual scene scheduling management, multi-level detail model switching, texture , lighting, terrain matching, collision detection and response, entity manipulation, two-dimensional map guide, constant frame rate method and other engine functions, compared with the two typical roaming engines in China, it has obvious advantages, and the overall function is similar to that of the University of Berkeley in the United States. The roaming system is equivalent, although its function is not as powerful in large scene processing, but the scheduling algorithm based on the scene control plane in the present invention is simple to implement, and the function can meet the requirements of medium-sized indoor and outdoor scenes. Compared with the roaming system of Berkeley University, it does not need The preprocessing process of the scene, at the same time, the real-time roaming of the scene can be completed on the basic configuration of the personal computer. In terms of application system development, there is no need to do specific programming work, only need to care about the establishment process of the scene, it is a virtual environment roaming engine with strong versatility, complete functions and advanced technology.

Description of drawings

Fig. 1 shows the main program flow chart of the present invention;

Fig. 2 shows the input device mapping and viewpoint control diagram of the present invention;

Fig. 3 shows the viewpoint control model diagram in the roaming engine of the present invention;

Fig. 4 shows a schematic diagram of screen area division during mouse input in the present invention;

Fig. 5 shows the working flow chart of the scene scheduling control of the model control plane of the present invention;

Fig. 6 shows the definition diagram of the model control surface of entering and exiting a certain building according to the embodiment of the present invention;

Fig. 7 shows a schematic diagram of terrain matching technology in the viewpoint control process of the present invention;

Fig. 8 shows a two-dimensional map multi-channel model diagram of the present invention;

FIG. 9 shows a diagram of the fog-based constant frame rate algorithm of the present invention.

Table 1 is the main control state table defined in the general roaming engine framework;

Table 2 is the device mapping table used when the keyboard is used for viewpoint control;

Table 3 The device mapping table used when controlling the viewpoint with the mouse;

Table 4 is the keyboard function table of the general roaming engine system.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

The universal virtual environment roaming engine of the present invention is a roaming engine that can operate independently, the software platform that adopts is Virtual Studio C++6.0 and OpenGVS4.3 software, and the operating system is Windwos2000.

Referring to Fig. 1, the present invention first creates general-purpose virtual reality environment resources, these resources are scenes, entities, frame buffers, etc., then loads scene database files conforming to scene description specifications, receives external input and forms input mapping, and then performs input mapping Explanation, input mapping interpretation is divided into state setting, viewpoint control and entity manipulation, state setting includes setting system state, two-dimensional navigation and environmental special effects, viewpoint control viewpoint control includes control of scene scheduling, collision detection, and terrain matching, Whether to exit and whether to release resources is judged by viewpoint control. When customizing the roaming state mechanism, set the input mapping and interpretation mechanism and accept an external input in each frame of the simulation cycle, support multiple input devices to access the roaming system, and perform viewpoint control, entity manipulation and state setting according to external input commands. Implement various scene complexity reduction strategies, collision detection and terrain matching in the process of viewpoint control, support the use of standard two-dimensional input devices to select and manipulate objects in three-dimensional scenes, and set various roaming control states and environmental effects to meet different users demand.

Referring to FIG. 2 , in order to meet various demands of different roaming applications, a state mechanism of customized roaming is set in the roaming engine of the present invention. Most VR applications are OpenGL compatible applications, and the internal implementation of OpenGL is itself a state mechanism. Therefore, setting the state mechanism in the present invention not only maintains the consistency with the graphics system, but also increases the flexibility of the roaming system.

Most of the functions defined in the general roaming framework are defined as optional items, that is, the roamer can turn on or off some functions according to their own needs, such as turning on/off the fog effect, turning on/off the two-dimensional map , Decide whether to perform collision detection, choose a transparent processing method, etc. On the other hand, the rover can also set the initial state of the roaming system, such as the initial position of the observation camera, the speed step of the camera, the step of the corner, the climate conditions simulated by the system (clear, cloudy, overcast), time period (morning, noon, evening) etc. When realizing the viewpoint control of the roaming system, the method of directly reading and processing the device information is mostly adopted. With the development of a large number of I/O devices and new interactive control technologies, more and more interactive devices will be used. Due to the diversity and scalability of the equipped equipment, it will undoubtedly bring complexity to the design and maintenance of the software. Table 1 defines the main control states in the roaming engine framework of the present invention.

Based on the above reasons, the present invention defines a basic device mapping, which abstracts an intermediate layer from viewpoint control and device functions, maps the control of input devices to this intermediate layer and makes it a source of viewpoint-driven control.

It can be seen that the control signal of the input device is mapped to a control instruction, so that the viewpoint control is separated from the direct control signal of the input device and becomes an independent functional module. Using this "isolation" technology not only facilitates the expansion of the roaming system to the input device, but also maintains the relative independence of the roaming engine.

Referring to Figure 3, in the roaming system, the point of view is the "avatar" of the human eye, and its function is the same as that of a video camera or camera in the real world. Viewpoint control generally refers to motion control of the viewing camera. The viewpoint control module uses two different sets of cameras to simulate the movement of human eyes and feet respectively. A camera that simulates the human eye is called an observation camera, and a camera that simulates a human foot is called a walking camera. The movement of the viewing camera is limited by the walking camera. The general roaming system is only equipped with observation cameras. The purpose of the invention is to set up walking cameras for two purposes: one is to realize multi-point collision detection during the roaming process; The degree of freedom of rotation, in addition to rotating left and right along the Y axis, the observation camera can also rotate up and down along the X axis, simulating the movements of people raising and lowering their heads. The basic parameters of the walking camera include: the height from the ground (walker_height), the forward or backward step (step), and the rotation step along the Y axis (θ). The setting of walker_height can improve the overall performance of the roaming system, because the walking camera will participate in the collision detection of the viewpoint control process. When the walker_height is set to a non-zero value such as 0.10, entities below 10cm in the virtual environment will not collide with the viewpoint. Corresponding to the behavior in the real world, all obstacles below 10cm, such as the threshold in a building, can be crossed by humans. In this way, the roaming system will reduce a large number of invalid calculations, thereby improving the overall performance of the system. The basic parameters of the observation camera are basically the same as those of the walking camera: the height from the ground (eye_height), the step size step forward or backward, and the rotation step size θ along the Y axis. In addition, the observation camera also sets another rotation parameter, that is, the angle step α along the X axis, which allows the rover to look up and down to observe the three-dimensional scene.

The view point control model also defines three parameters for collision detection for the observation camera and the walking camera, v_p0, v_p1, v_p2 and w_p0, w_p1, w_p2, which represent the current position of the camera and a point at a certain distance in the forward direction respectively. , a point at the same distance in the backward direction. The viewpoint control model defines camera motion: observation camera motion collection: {FORWARD, BACKWARD, TURN_LEFT, TURN_RIGHT, LOOK_UP, LOOK_DOWN}; walking camera motion collection: {FORWARD, BACKWARD, TURN_LEFT, TURN_RIGHT}

It can be seen that viewpoint control is the motion control of walking camera and observation camera. There are only two ways of camera movement; translation and rotation. The present invention uses the displacement step step in the line of sight direction to define the change of the camera position, and uses the rotation angle steps θ and α of the camera along the coordinate axis to define the change of the line of sight direction. Therefore, the mapping problem of the input device is transformed into the problem of the mathematical transformation of the input device control amount to the camera motion type and step, θ, α.

The mapping of keyboard input is the simplest, using the direct mapping method, taking the observation camera as an example, as shown in Table 2.

Referring to FIG. 4, the schematic diagram of screen area division during mouse input, the viewpoint control mapping of mouse input is different. The present invention adopts the method of screen area division to identify the control function of a mouse input. Note that the current mouse position is (xm, ym). According to the relevant principles of equipment coordinate transformation, xm∈[-1, 1], ym∈[-1, 1], and the coordinates of the lower left corner of the two-dimensional display screen are (- 1, -1), and (1, 1) in the upper right corner. Therefore, divide the screen into ten areas as shown in Figure 5, respectively marked as area 1 to area 10, and set: when using the mouse device to control the viewpoint, the maximum translation step of the camera is MAXSTEP, and the maximum rotation angle step is MAXROTATE, so that speed=MAXSTEP*fabs(ym), rot=MAXROTATE*fabs(xm), the mapping of the mouse device is shown in Table 3. Other input device mappings are basically similar to mouse device mappings.

To describe a moderately complex community roaming virtual environment based on geometric virtual reality technology, especially the indoor environment of buildings, generally requires tens of thousands or even hundreds of thousands of solid surface triangles. Due to the limitations of graphics hardware performance, computer physical memory, CPU frequency and other conditions, a virtual reality system that does not adopt a scene complexity reduction strategy will not be able to guarantee real-time interactive roaming. Using multiple scene complexity reduction strategies can improve the overall performance of the roaming system. The strategy includes several methods such as collision detection-based model control surface scene scheduling method, redundant polygon reduction, texture mapping, and multi-level detail model.

Referring to Figure 5, the scene scheduling management has two meanings. On the one hand, due to the limitation of the physical memory capacity of the computer, sometimes it is impossible to transfer a complex scene database into the memory all at once. At this time, scene scheduling management refers to loading and unloading model data in blocks. On the other hand, due to the limitations of computer hardware, especially the performance of graphics accelerator cards, rendering too many polygons in one frame will also exceed the limit of system processing capabilities. Reducing the number of polygons rendered in one frame is another aspect of scene scheduling management.

Classical indoor scene scheduling control, such as PVS (Potentially Visible Set) algorithm, mostly divides the roaming process and scene scheduling into two stages. For a specific virtual environment, the above algorithm divides the space of the building according to the partition planes parallel to the world coordinate axis, such as the walls, floors, and ceilings of the enclosed rooms, to form Cell (spatial unit). Then mark the visible areas such as doors, windows, etc. on the divided spatial units, and then calculate the visible areas from Cell to Cell and from Cell to Entity one by one, and finally store the calculation results in the scene database. In this way, in the roaming stage, the roaming driver does not need to make complex judgments on the visible area, and directly uses the pre-calculated results to perform scene scheduling control to achieve better real-time effects. However, the precomputation method also has some obvious shortcomings, such as the precomputation of space division and visible area depends on the virtual environment, the precomputation results occupy a large scene database space, and the precomputation workload is relatively large. Especially for hollow buildings, the efficiency of the pre-calculated visible area algorithm is relatively low.

The roaming engine of the present invention proposes a model control surface scene scheduling control method based on collision detection. This method organically combines scene modeling with roaming control, uses subjective evaluation criteria to determine the visible area, and achieves better results.

It can be seen that the model control surfaces are set during the modeling phase and stored in the scene database together with other scene data. The control surface is a space plane, which is determined by the space equation Ax+By+Cz+D=0. Different from the general space plane, the control defines some special properties in the modeling process, the most important of which is the scene scheduling control behavior. The control behavior only defines an identifier in the modeling stage, and the specific control behavior is analyzed and determined by the roaming engine.

During the roaming process, after the line-of-sight-based collision detection algorithm detects the control surface of the model, it parses out the control behavior identifier defined during modeling, maps it to the scene control object set, and loads the visible objects and unloads them according to the object list in the control object set. Invisible objects, realize scene scheduling management. The set of scene control objects is repeatedly practiced by the roaming program and finally determined by subjective criteria. That is, determine which objects are visible objects and which are invisible objects at a certain control plane, and record them in the data file of the roaming drive.

Figure 6 is a definition diagram of the model control surface for entering and exiting a building. When the roaming viewpoint is close to the model control surface in front of the main entrance of a building, the collision detection algorithm gives the control behavior specified by the control surface at this place, such as entering the building, After the analysis of the behavior analysis module, the corresponding scene scheduling control object set is found out. Naturally, the building outline model should be closed, and the building indoor model should be transferred, and the visible and invisible objects on the six floors should be displayed accordingly Dealing with display suppression.

The control surface of the model is described by a six-tuple (A, B, C, D, arg ₁ , arg ₂ ), where A, B, C, and D uniquely determine a plane Ax+By+Cz+D=0 in the three-dimensional space , arg ₁ and arg ₂ are the control behavior parameters given by the control plane, which are passed to the control behavior analysis module by the detection algorithm. Multiple model control surfaces are set in the scene model of a building. Figure 6 shows the scene scheduling control diagram for entering and exiting a building. Six model control surfaces set the scene scheduling behavior for entering and exiting a building. In addition, control surfaces are also set up at the stairs where each floor is switched and near the floor of the floor near the outer window. , most objects are invisible, and the floor switching control surface can control this kind of scene scheduling management, greatly reducing the number of polygons drawn in one frame.

A basic principle of 3D scene modeling is to use the least number of polygons to achieve the same realistic visual effect. There is often redundancy in the surface data describing the entity model, and data redundancy also occurs when the separately modeled entities are integrated. Eliminating these redundant surface polygons can greatly reduce the overall scene complexity. For example, the multi-level detail window model of the prior art is composed of 156 triangles, and a window edge on the top of the window is described by 12 triangles (6 rectangular planes) in the usual modeling technology. The coincidence of the vertical planes on the side is not only redundant data, but also causes Z value contention, so it should be deleted in advance. In this way, the entire window is deleted once, and the window with the same visual effect can be composed of 104 triangles. Considering the data redundancy caused by model integration, the uppermost facet will coincide with the wall, and can also be deleted, and so on. The final window model only needs 96 triangles to fully describe its geometric topology information. Compared with the model represented by 156 triangles, its visual effect has no loss, but it improves the system performance by nearly 40%.

Texture mapping technology is widely used in computer graphics applications to express realistic details of solid surfaces, and it is also an effective method to control scene complexity and accelerate graphics rendering speed. It maps a two-dimensional texture image defined by a rectangular array to a three-dimensional solid surface, or modifies the light intensity distribution on a solid surface through a process.

The general method of generating texture is to predefine the texture pattern on a plane area, that is, the texture space, and then establish a mapping between the points on the surface of the object and the points in the texture space. After the visible points on the surface of the object are determined, the texture pattern can be attached to the surface of the object by multiplying the value of the corresponding point in the texture space by the brightness value. A similar method can be used to create a bumpy appearance, or bump texture, to the surface of an object. But this time the texture value acts on the normal vector, not the color brightness. Whether it is to generate color textures or convex-concave textures, it is generally only required to be roughly close to the real pattern, and it is not necessary to do precise simulations, so as to greatly increase the realism of the solid surface without significantly increasing the amount of calculation.

Texture mapping technology can greatly reduce scene complexity, but this method also has certain limitations. When the rover gets close to the solid model, the textures represent a lack of realism to the solid details. If the multi-level detail model is used in combination with texture mapping technology, it can effectively reduce the complexity of real-time rendering of the scene while ensuring the realism of entity details.

The multi-level detail model refers to a group of models obtained by using description methods with different details for the same scene or entities in the scene, which can be selected for use in drawing. Since virtual reality usually uses polygonal meshes to describe the geometric entities in the scene, the multi-level level of detail model uses meshes of different degrees of complexity to achieve real-time rendering of virtual scenes at different fine levels. Typically, the LOD model uses more polygons to create the most detailed description of the entity, and uses fewer polygons to create the outline description of the entity. When the roaming viewpoint is close to the entity, the detailed model is selected, and when the viewpoint is far away, the contour model is selected. The general roaming system adopts the three-level multi-level detail model technology. Taking the window described above as an example, the high-level multi-level detail model is based on the design data, and all use the geometric modeling method to give the entity description. The low-level multi-level detail model replaces the window part with texture, but the window sill still uses a geometric model, and the low-level multi-level detail model uses the entire window as a texture plane.

The roaming engine is responsible for realizing the control behavior analysis and scene scheduling management of the scene scheduling control method based on the model control plane. Because the control object set is stored in the roaming system, which gives the object set to be loaded when the control surface of a certain model is opened and the object set to be unloaded when the control surface is closed, the implementation of behavior analysis and scene scheduling becomes very simple.

When the viewpoint collides with the model control surface and returns the control parameters, the behavior analysis only needs to simply locate the control object set corresponding to the model control surface. Then, for a certain model control surface, the objects in the object set are unloaded and loaded one by one according to the following method.

As shown in Figure 7, in the process of viewpoint control, the roaming system must perform collision detection between the viewpoint and various entities in the virtual environment, and then give a reasonable collision response. The collision detection algorithm is included in the camera motion control module. Once a collision occurs between the viewpoint and the virtual entity, the forward or backward command given by the input device will be ignored, that is, the camera will no longer move forward or backward. A simple response to collisions given in the invention. Terrain matching usually refers to the movement of dynamic entities in the virtual environment, such as ground vehicles, whose posture fluctuates with the terrain and deflects left and right. In the roaming system, the rover controls the "incarnation" of the human eye, that is, the viewpoint, so terrain matching means that the height and direction of the viewpoint change with terrain changes.

In the collision detection algorithm, you can see when the terrain matching module is called. Any external input that changes the position of the viewpoint will cause the call of the collision detection module. Once the result of the collision detection shows that a collision has occurred, the collision response will be to ignore the change of the viewpoint and the viewpoint will stay in place. If the result of the collision detection is that no collision occurs, the viewpoint will arrive at a new position according to the specified step size and direction, but this viewpoint position and viewing direction are not the final position of the viewpoint’s response to external input. The final position must be Determined after terrain matching.

After the camera advances the step distance in the viewing direction, its position is recorded as WPOS(xw, yw, zw). The camera position after terrain matching must be above the Yv axis of the viewpoint coordinate system. Establish the vertical straight line direction through the point WPOS, that is, the direction of the Yv axis, and use the method of intersecting the line of sight and the space polygon set to find the plane intersecting with the Yv axis and the coordinates of the intersection point. Among the intersection points, the Y component does not exceed a given value and the largest intersecting plane is determined as the terrain surface. Position the walking camera above the surface's Y height. The height value of the observation camera is set to the walking camera plus a fixed value eye_height-walker_height. The terrain matching problem of line of sight is relatively simpler. The direction of the terrain surface polygon is taken, so that the direction of the observation camera is consistent with the direction of the terrain surface, but the line of sight of the walking camera remains unchanged.

Two-dimensional maps are widely used navigation tools in navigation systems. Compared with the three-dimensional scene view, the advantage of the two-dimensional map is that it can provide a wider field of view, which is convenient for the roamer to grasp the current position and the surrounding environment as a whole.

In general, to develop a 2D map display module, it is necessary to extract 2D basic features in the modeling space, such as road and building size data, use the OPENGL graphics system to draw lines, and then fill the polygonal area with the corresponding color, and control a Viewpoint pointer, which moves in line with the viewpoint in the 3D scene.

As shown in FIG. 8 , the present invention uses another simpler two-dimensional map development method, which generates a two-dimensional map by utilizing the orthographic projection of a three-dimensional scene. According to the principle of orthographic projection in computer graphics, this method "compresses" the 3D scene model onto a plane, and then uses camera resources to realize map display, zooming, and synchronous movement of 2D and 3D viewpoints. The multi-channel programming technology is used to implement the two-dimensional map guide, that is, the channel model shown in Figure 8 is established. The main channel (left channel) in the figure is also the main channel, which is used to display the three-dimensional scene of the roaming system, and channel2D (right channel) channel) to display a two-dimensional map. The main channel occupies the entire screen, and the two-dimensional map channel only occupies one-eighth of the upper right corner of the screen. In channel2D, not only the camera is set, but also the lighting and observation body are preset, but the lens of the camera is always facing the model plane representing the two-dimensional map. Of course, it is impossible for any system to transfer all the models representing the three-dimensional world into channel2D in order to quickly and conveniently realize two-dimensional map display, which will lead to a double increase in system load, which is not worth the candle. Only the surface data (that is, all the solid parts higher than the ground) are simply repaired to form a three-dimensional representation model of the two-dimensional map. After establishing the multi-channel mechanism and the two-dimensional map representation model, the problem of synchronizing the viewpoint indication of the two-dimensional map and the three-dimensional scene is transformed into the problem of the camera in channel2D tracking the walking camera in the main channel. During the tracking process, the lens direction of the camera in channel2D remains constant. The zoom function of the 2D map is realized by setting different focal lengths for the camera.

Virtual reality applications not only pursue a high interactive simulation rate, that is, the frame rate, but also hope that the frame rate remains consistent and constant, at least within a certain range expected by the user. Especially for combat simulation type virtual reality applications, jumping frame rates will lead to False simulation results and leave participants feeling overwhelmed. The frame rate is the reciprocal of the time it takes to draw a frame of the scene, so to obtain a constant frame rate, you must control the number of polygons on the surface of the scene drawn in one frame.

最大算法。在对象三元组上建立两个评价函数Cost(O，L，R)和Benefit(O，L，R)，Cost(O，L，R)用于计算对象选用L级LOD时R算法的绘制时间，Benefit(O，L，R)用于评价L级的多级层次细节及R算法条件下该对象对整个场景的视觉方面的贡献。这样，恒定帧频的问题转化为算法中使用一个线程专门负责上述两个函数的计算，在每一帧绘制之前确定选用的多级层次细节级别及绘制算法。In 1996, Thomas Van Kehaus and Thursteller of the University of North Carolina proposed an effective but complicated method of constant frame rate in the SodaHall roaming system. They set up a triplet (O, L, R) for each object in the scene database, where O represents an object in the scene, L represents a LOD model level of the object, and R represents the drawing applied to the object $\underset{\mathrm{the\; s}}{\mathrm{\Σ}}\mathrm{cost}(o,L,R)\≤T\arg \mathrm{et\; Frame\; Time}$ and

Maximum Algorithm. Establish two evaluation functions Cost (O, L, R) and Benefit (O, L, R) on the object triplet, Cost (O, L, R) is used to calculate the drawing of the R algorithm when the object selects L-level LOD Time, Benefit(O, L, R) is used to evaluate the L-level multi-level details and the contribution of the object to the visual aspects of the entire scene under the condition of the R algorithm. In this way, the problem of constant frame rate is transformed into the use of a thread in the algorithm to be responsible for the calculation of the above two functions, and the selected multi-level detail level and rendering algorithm are determined before each frame is drawn.

The constant frame rate method of the University of North Carolina is an efficient prefetch-based optimization algorithm that can be applied to most virtual reality applications, but the calculation in this algorithm mainly relies on subjective criteria, and the entire algorithm is relatively complex.

Fog can not only improve the realism of the scene, but also improve the overall performance of the virtual reality system. The essence of atomization is to fuse the original color of the scene with the color of the fog according to a certain scale factor, and the value of the scale factor is related to the atomization model. Most graphics acceleration engines compatible with OPENGL provide three fog models:

Where f is the scale factor, density is the density of the fog, z is the distance from the viewpoint to the center point of the scene patch in the observation coordinate system,

f＝e ^{-(density z)} (GL_EXP)

$f=e^{-{(\mathrm{density}\&Center\; Dot;z)}^2}$ (GL_EXP2)

$f=\frac{\mathrm{end}-z}{\mathrm{end}-\mathrm{start}}$ (GL_LINEAR)

start and end are the start value and end value of the atomization depth.

Under the RGBA color model, the color C of each pixel on the observation plane after atomization is calculated by the following formula:

C＝fC ₁ +(1-f)C _f

C ₁ is the RGBA color value of the scene patch, C _f is the color of the fog, and f is the scale factor discussed above.

The above fog model gives an important property of fog, that is, the farther the distance between the viewpoint and the entity is, the greater the degree of fogging the scene is, and the less visible the entity is in the scene. The spline transformation atomization model has a more significant atomization effect on the scene. When the distance between the viewpoint and the entity exceeds a given value, the atomized color of the entity is replaced by the color of the fog, and no real-time drawing is required.

Based on the above analysis, the present invention designs a method to adjust the average display time of each frame of the scene in a small range by adjusting the parameters of the atomization model, and then achieve the purpose of constant interactive simulation rate.

Referring to FIG. 9 , the present invention adopts a simple but very effective constant frame rate technology, that is, a constant frame rate method based on fog visible distance adjustment. This method can be considered as an after-the-fact adjustment method. The so-called post-adjustment refers to dynamically setting the fog effect according to the difference between the frame rate in the previous period and the target frame rate, so that the number of polygons on the solid surface processed in each frame increases or decreases, so as to achieve the goal of adjusting the frame rate. The specific algorithm can be represented by the flowchart shown in FIG. 9 . It can be seen that the algorithm of the present invention to achieve the atomization effect is to use the current frame rate to gradually approach the target frame rate, so that the frame rate of the roaming system "swings" around the target frame rate, so as to achieve a constant frame rate at one The purpose of the fixed interval. This algorithm is suitable for virtual reality applications with large scenes.

Table 4 shows the keyboard function table of the universal roaming engine system according to the embodiment of the present invention. According to the steps above, a universal roaming drive computer system using the virtual reality environment of the present invention can be developed.

state control Hotkey definition default state Atomization effect enable/disable CTRL+Z Enabled state 2D map guide enable/disable m Disabled Outdoor scene visible/invisible P visible status Collision detection enable/disable for camera forward movement CTRL+G Enable collision detection Collision detection enable/disable for camera backward movement CTRL+R Enable collision detection Texture mapping on/off CTRL+X open state Wireframe display on/off CTRL+P Disabled Frame count statistics display/not display TAB+'-' Disable display status Debugging information output enable/disable CTRL+T Prohibited status Automatic manipulation of DOF entities allowed/disabled o Allowed state Transparent algorithm switching selection F8, F9 AF status 2D map camera zoom selection F1-F5 Maximum focal length state Frame buffer mode single/dual mode switching CTRL+E Single frame buffer mode Display/hide mouse cursor CTRL+B Display state Permission and prohibition of looking up/down of the camera's head CTRL+H Allowed state

Table 1

key definition Sports type step θ α UPARROW KEY FORWARD 0.7 n/a n/a DOWNARROW KEY BACKWARD 0.5 n/a n/a left ARROW KEY TURN_left n/a 0.3 n/a rightARROW KEY TURN_right n/a 0.3 n/a U KEY LOOK_UP n/a n/a 0.15 J KEY LOOK_DOWN n/a n/a 0.15

Table 2

The area where the mouse is located Type of camera movement step θ α Area 1 FORWARD speed n/a n/a Area 2 BACKWARD speed n/a n/a Area 3 FORWARD, TURN_right speed rot n/a Zone 4 BACKWARD, TURN_right speed rot n/a Zone 5 TURN_right n/a rot n/a Zone 6 TURN_right n/a rot n/a Area 7 FORWARD, TURN_left speed rot n/a Zone 8 BACKWARD, TURN_left speed rot n/a Area 9 TURN_left n/a rot n/a Area 10 TURN_left n/a rot n/a

table 3

Hotkey definition Functional description Hotkey definition Functional description ↑Up cursor key Move the viewpoint forward f Reduced fog density ↓Cursor key down Move the viewpoint backwards F Fog concentration increased ←Left cursor key Turn view left o Setting of automatic opening/closing door and window function →Right cursor key turn view right m 2D map on/off u look up P Outdoor scene on/off J lower your head T Debug information output on/off L Helicopter viewpoint descent r Reduce the red component in the material h Helicopter viewpoint up R Increase the red component in the material PageUp Set the helicopter viewpoint g Reduce the green component in the material PageDown Reset Helicopter Viewpoint G Increase the green component in the material Home to the front of a building b Reduce the blue component in the material End before a specified point B Increase the blue component in the material F1 2D map camera zoom C cloud on/off F2 2D map camera zoom CTRL+G Forward collision detection on/off F3 2D map camera zoom CTRL+R Backward collision detection on/off F4 2D map camera zoom CTRL+B Mouse pointer display on/off F5 2D map camera zoom CTRL+X Texture drawing on/off F7 Transparent Algorithm One Choice CTRL+I Statistics On/Off F8 Transparent Algorithm 2 Choice CTRL+C Camera Control On/Off F9 Atomization concentration setting CTRL+H View camera on/off 1 Place the camera on the first floor of a building CTRL+P Wireframe drawing on/off 2 Place the camera on the second floor of a building CTRL+L magnification of virtual entities 3 Place the camera on the third floor of a building CTRL+S Shrinkage of virtual entities 4 Place the camera on the fourth floor of a building CTRL+M Forward rotation of the virtual entity 5 Place the camera on the fifth floor of a building CTRL+N Negative rotation of virtual entities a Increase the ambient value of sunlight CTRL+Z Atomization on/off A Reduce the ambient value of sunlight CTRL+T Texture replacement function key d Increase the diffuse value of sunlight '-' Detailed statistics display on/off D Reduce the diffuse value of sunlight s Sun rise S the sun sets

Table 4

Claims (10)

1, a kind of method that realizes universal virtual environment roaming engine comprises personal computer, graphics acceleration card, walker, stereoscopic vision demonstration and tracking equipment, scene database and roaming engine core component, and this method may further comprise the steps:

(a) create general virtual reality applications resource;

(b) load the contextual data library file that meets the scene description standard;

(c) customization roaming state mechanism;

It is characterized in that also comprising:

(d) input mapping and explanation facility and all accept once outside the input at each frame of simulation cycles is set, supports multiple input equipment access roaming system;

(e) carry out viewpoint control, manipulations of physical and state setting according to outside input command, implementing the several scenes complexity in the viewpoint control procedure subdues strategy, carries out collision detection and terrain match, support use standard 2-d input device to select and handle the object in the three-dimensional scenic, multiple roaming state of a control and environment special efficacy are set, satisfy requirements of different users.

2, the method for realization universal virtual environment roaming engine according to claim 1, it is characterized in that: the viewpoint control that described viewpoint control is to use the double camera model to roam, create and observe view camera and walking view camera, observe view camera and be subject to the walking view camera, the walking camera is in order to realize the multiple spot collision in the roam procedure and to keep direct of travel that observing camera has three rotational freedoms.

3, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that described viewpoint control further comprises:

(a) in the motion of view camera and walking camera as detect collision, then judge the character of collision;

(b) if the scene chain of command then starts the scene scheduling strategy, carry out the management and running of indoor and outdoor scene;

(c) if steerable movement entity then carries out the motion control of entity in the scene by moving coordinate system that defines in the model and kinematic parameter;

(d) if confirm collision has taken place, then provide collision response.

4, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that step (d) further comprises:

(a) input mapping and explanation facility are in order to roaming engine and input equipment are isolated, to reduce the influence of input equipment to the roaming system kernel to greatest extent;

(b) from viewpoint control and functions of the equipments, take out a middle layer;

(c) control with input equipment is mapped to this middle layer and makes it become the Controlling Source that viewpoint drives.

5, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that: in whole roam procedure, support several scenes height and complexity to subdue strategy simultaneously, may further comprise the steps:

(a) replace by the precision of the multi-level detail model of distance scheduling of observing view camera and each model of place;

(b) determine the scene chain of command with three dimensions plane in the scene database and behavior parameter, carrying out the scheduling of indoor and outdoor scene and complex indoor scene divides, in roam procedure, after detecting the model chain of command based on the collision detection algorithm of sight line, the control behavior identification (RNC-ID) analytic that defines during with modeling comes out, and is mapped to the scene controlling object and concentrates, according to the concentrated list object of controlling object, load viewable objects, unload invisible object, realize the scene management and running;

Automatically subdue the redundant polygon in the solid model when (c) model loads;

(d) texture of generation solid object surface.

6, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that: described collision detection further comprises with response:

(a) reason of analysis camera motion is set up the walking camera and is observed camera collision detection line segment forward;

(b) observe the camera collision detection;

(c) walking camera collision detection if bump, is then carried out the scheduling of virtual environment scene and is handled;

(d) if do not bump, then camera advances, and carries out terrain match.

7, according to the method for the realization universal virtual environment roaming engine of claim 1, it is characterized in that: described terrain match technology further may further comprise the steps:

(a) the walking camera advances;

(b), call the collision detection module if viewpoint position changes;

(c) the collision detection result then ignores viewpoint position and changes for bumping, and walking camera and observation camera rising certain altitude are provided with the line of vision of observing camera by the topographical surface direction simultaneously;

(d) if the result is not for bumping, then viewpoint arrives a reposition.

8, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that: state is provided with function and comprises ON/OFF atomizing effect, ON/OFF two-dimensional map guide, ON/OFF collision detection, selects the transparent processing mode, one or more of weather conditions etc. are set.

9, the method for realization universal virtual environment roaming engine according to claim 1 is characterized in that: affiliated environment special efficacy comprises with the atomizing step of the constant frame frequency method that is technical support.

10, the method for realization universal virtual environment roaming engine according to claim 9 is characterized in that: described constant frame frequency method further comprises:

(a) take out that every frame on average consumes the time in preceding 2 seconds;

(b) calculate current frame frequency;

(c) calculate the poor of current frame frequency and target frame frequency,

(d) if current frame frequency greater than the target frame frequency then increase the atomizing visibility, otherwise reduces the visibility that atomizes;

(e) new atomization parameter and end are set.

Images