CN102662729B - Method for dispatching growth models for simulation of vast forests - Google Patents

Abstract

A method for dispatching growth models for simulation of vast forests. The method includes the following steps of firstly, acquiring simulation data of a field of the vast forests according to simulation target, initializing the structure of the simulation data, storing field information data excluding tree information in a quad-tree manner; storing tree information data by the memory pool technology; secondly, utilizing field blocks as units and traversing all the field blocks during computing tree growth models on the field of the vast forests, and computing tree growth in each field block; and thirdly, visualizing the field of the vast forests and storing computing results of the tree growth models in the field of the vast forest according to the computing results of the tree growth models in the field of the vast forest. The method for dispatching growth model for simulation of vast forests is excellent in instantaneity, high in speed of computing process and smooth in field visualization.

Description

A Scheduling Method for Growth Models Oriented to Large-Scale Forest Simulation

technical field

The invention relates to computer virtual simulation technology, in particular to a scheduling method of a growth model of large-scale forest simulation.

Background technique

Computer virtual simulation technology has changed the traditional forestry management model, enabling people to predict the growth of trees through computers. Dynamically simulating changes in forest scenarios according to specific conditions can enable forestry management decision makers to make more scientific management decisions. However, for the simulation of large-scale forest scenes, the amount of scene data is very huge (including tree information data, terrain data, etc.). For example, to simulate the growth of trees in an area of 5km*5km, the tree information required for the calculation of the plant growth model is about 1.6-2G of data, and the LOD technology is often used for scene rendering, and a fine 3D model is used for places close to the viewpoint For drawing, using a simplified model at a far distance or using Billboard technology to use texture maps, the amount of memory required is roughly about 2G of data. If these data are directly transferred to the computer for growth model calculation and simulation drawing, the rapid simulation of plant growth process will not be realized with the current hardware performance. Therefore, it is necessary to pre-process the original scene data before performing forest scene simulation, and use a reasonable scheduling strategy to schedule and optimize the growth model calculation during the simulation process, so that the simulation of large-scale forest scenes can be faster.

Contents of the invention

In order to overcome the disadvantages of poor real-time performance, long calculation process, and unsmooth scene visualization of the existing forest scene simulation, the present invention provides a large-scale oriented simulation system with good real-time performance, reduced calculation time, and smooth scene visualization. Scheduling methods for growth models for forest simulations.

The technical solution adopted by the present invention to solve its technical problems is:

A method for scheduling a growth model oriented to large-scale forest simulation, the method for scheduling includes the following steps:

1) Acquire the simulation data of a large-scale forest scene according to the simulation target, and initialize the data structure of the simulation data. The scene information data that does not contain tree information is stored in the form of a quadtree; the tree information data is stored using memory pool technology;

2) When calculating the tree growth model in a large-scale forest scene, use the scene block as a unit to traverse all scene blocks, and calculate the tree growth in each scene block. The specific steps of tree growth calculation are as follows:

A. Determine the current tree to be calculated in the scene, and calculate the range of the tree’s circle of influence based on the attributes of the current calculation tree, and then calculate the relevant simulation parameters according to the range of the circle of influence, and search for trees related to the current calculation by traversing the entire scene tree the trees;

B. After obtaining the relevant trees, it is necessary to schedule all the attribute information data of these trees into the memory, and use the scheduling strategy of forest growth simulation to schedule the internal and external memory data during the growth calculation process;

C. Solving the growth model: After all the data required for the current calculation are loaded into the memory, the growth equation is solved and the final result is calculated;

3) Visualize the large-scale forest scene according to the calculation results of the tree growth model in the large-scale forest scene, and save the calculation results of the tree growth model in the large-scale forest scene.

Further, in the step B, the scheduling strategy of the forest growth simulation includes a drawing thread and a scheduling thread;

When the user's viewpoint changes, the drawing thread calculates the current visible area according to the current viewpoint, and at the same time judges the scene data in the memory to confirm whether the data required for redrawing the scene is already in the memory, and if so, proceed directly to the scene Draw and send a viewpoint update message to the data prefetch thread; if the current scene to be drawn is not in the memory, the drawing thread needs to calculate the visible area according to the current viewpoint position, current viewpoint direction and other viewpoint information; after the calculation is completed , the drawing thread hangs up and sends a data request message to the data dispatcher, and the data dispatcher thread continues drawing after loading the scene data from the external memory;

The data scheduling thread is responsible for scheduling data from the external memory. By analyzing the messages passed by the drawing thread, it can determine the specific content of the scheduling data, such as the current scene drawing data, scene prefetching data, or scene calculation data, etc., and delete some from the memory at the same time. scene data;

When the data scheduling thread receives the scene data request message sent by the drawing thread, it performs direct scheduling. At this time, the scheduling thread directly schedules the data required by the drawing thread from the external memory according to the message. At this time, the data scheduling priority is the highest;

When the data scheduling thread receives the new viewpoint coordinate information sent by the drawing thread, it means that the scene data drawn by the current scene is already in the memory, and the drawing thread has entered the drawing operation; at this time, the main work of the data scheduling thread is based on the new viewpoint The data to determine the data that may be used for the next drawing, and schedule these data into the memory; after receiving the information of the viewpoint sent by the drawing thread, the data dispatching thread updates the viewpoint view area according to the viewpoint information, and calculates the data that needs to be prefetched area and judge the LOD level corresponding to the prefetch area; the number of the scene block in the prefetch area is directly calculated according to the number of the visible area; after calculating all the scene blocks in the prefetch area, since the data prefetch time must be shorter than the Time, so it is necessary to classify the scheduling priority of these scene blocks, and the judgment is mainly based on the direction of the viewpoint at the previous moment;

The prefetch area is divided into three priorities according to the angle between the scene block and the direction of view movement, and the priority Ki of scene block i is as in formula (1):

Among them, K _i represents the priority of scene block i; θ _i represents the angle between the scene block and the movement direction of the viewpoint; R represents the set of scene blocks in the prefetch area;

When scheduling the prefetched data into the memory, it is necessary to delete scene data of the corresponding size from the memory, and select the LRU strategy to select deleted scene blocks.

Further, in the step 1), the scene linked list is used as an index, and the tree information is stored in the MemoryBlock. Each scene block in the scene linked list may have a plurality of MemoryData to store the tree information in the block, and at the same time, the trees in the MemoryData Sort to speed up the search for related trees later.

The beneficial effects of the present invention are mainly manifested in: (1) by analyzing the information characteristics of the whole forest classification simulation of the overall dynamic evolution of the forest, the forest classification simulation of the interaction between plants and the single tree level simulation of the dynamic growth of individual trees, the scene The external memory data structure is designed and organized, and the terrain, ground features (non-tree features), and texture data are segmented, and an interrelated index and external memory storage structure are established for the three, making the data reading speed faster during simulation calculations ;

(2) According to the relatively time-consuming steps in the solution process of the coarse-grained scene dynamic evolution model, the medium-grained plant interaction model, and the fine-grained tree individual dynamic growth model, the storage of scene data in memory is designed and organized to reduce The time of the calculation process of the forest scene model;

(3) Aiming at the problem that the scene data cannot be loaded into the memory at one time when solving the large-scale forest scene growth model and visualizing the growth results, the internal and external memory data scheduling method is optimized during the growth model solving and visualization process, including coarse-grained, Scheduling optimization during calculation and simulation of medium-grained and fine-grained scene growth models makes the final scene visualization smoother.

Description of drawings

Figure 1 is a schematic diagram of the scene data memory organization structure.

FIG. 2 is a schematic diagram of a storage structure of scene data in memory.

Fig. 3 is a schematic diagram of scene multi-thread scheduling.

Fig. 4 is a schematic diagram of LOD hierarchical model selection.

FIG. 5 is a schematic diagram of scene rendering data prefetching.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings.

Referring to Figures 1 to 5, a scheduling method for growth models oriented to large-scale forest simulation, which obtains simulation data of large-scale forest scenes according to the simulation objectives and stores them in computer memory, and then processes the large-scale forest scene data in the simulation process according to the large-scale forest scene data. According to different scheduling methods, the data structures (storage structures in the internal and external memory) are initialized respectively, as shown in FIG. 1 .

In the computer memory, for the scene information data (not including tree information), we store it in the form of a quadtree according to the quadtree segmentation of the terrain.

For the tree information data, considering that the data in the memory needs to be deleted and stored more frequently during the calculation process, the memory pool technology is used for storage, as shown in Figure 2.

In the memory, the scene linked list is used as the index, and the tree information is stored in MemoryBlock. Each scene block in the scene linked list may have multiple MemoryData to store the tree information in the block. At the same time, the trees in MemoryData are sorted to speed up the search for related trees later.

2. Calculate the tree growth model for large-scale forest scenes, and obtain the results of interaction calculations for each tree in the scene.

When calculating the tree growth model in a large-scale forest scene, use the scene block as a unit to traverse all scene blocks, and calculate the tree growth in each scene block. The specific steps of tree growth calculation are as follows:

A. Determine the current tree to be calculated in the scene, and calculate the range of the tree’s circle of influence based on the attributes of the current calculation tree, and then calculate the relevant simulation parameters according to the range of the circle of influence, and search for trees related to the current calculation by traversing the entire scene tree trees.

B. After obtaining the relevant trees, it is necessary to schedule all the attribute information data of these trees into the memory, so that they can be quickly used later in the growth calculation of the current tree. However, due to the limited memory size of the computer, it may not be possible to schedule all the tree data into the memory at one time, so it may be necessary to schedule internal and external memory data during the calculation process. In order to make full use of computer resources, the scheduling strategy of forest growth simulation is used to schedule the internal and external memory data during the growth calculation process.

Threads are divided into drawing threads and data dispatching threads according to different functions. When the viewpoint changes, the rendering and data loading of the scene can work simultaneously to improve the drawing efficiency of the scene. During the scene drawing process, the CPU is mainly responsible for the calculation of the visible area, the calculation of the scene biomass, and the judgment of the LOD level of the scene model, and the drawing of the scene is performed by the graphics processor. Therefore, when the CPU hands over the processed scene data to the graphics processor for drawing, the CPU will be in an idle state at this time, and this idle time can be used to schedule the scene data. Figure 3 shows the multi-thread work model.

(1) Drawing thread

The main function of the drawing thread is to calculate the visible area of the current scene according to the viewpoint information, judge whether the drawing data of the visible area is in the memory, and send a prefetch message to the data scheduling thread.

When the user's viewpoint changes, the drawing thread calculates the current visible area according to the current viewpoint, and at the same time judges the scene data in the memory to confirm whether the data required for redrawing the scene is already in the memory, and if so, proceed directly to the scene Draw and send view update messages to the data prefetch thread. If the scene currently to be drawn is not in the memory, the drawing thread needs to calculate the visible area according to the viewpoint information such as the position of the current viewpoint and the direction of the current viewpoint. After the calculation is completed, the drawing thread hangs up and sends a data request message to the data dispatcher, and the data dispatcher thread loads the scene data from the external memory and continues drawing. As shown in Figure 4, LODO in the figure is a fine 3D model of plants, LOD1 is a simplified 3D model of plants, and LOD2 uses Billboard technology to directly use texture maps.

(2) Scheduling threads

The work of the data scheduling thread is mainly responsible for scheduling data from the external memory, by analyzing the messages passed by the drawing thread, to determine the content of the specific data that needs to be scheduled, such as the current scene drawing data, scene prefetching data, or scene calculation data, etc. Part of the scene data is deleted from the internal memory.

When the data scheduling thread receives the scene data request message sent by the drawing thread, it means that the scene data to be drawn is not completely in the memory, and direct scheduling is required. At this time, the scheduling thread directly schedules the data required by the drawing thread from the external memory according to the message, and the data scheduling priority at this time is the highest.

When the data scheduling thread receives the new viewpoint coordinate information sent by the drawing thread, it means that the scene data for the current scene drawing is already in the memory, and the drawing thread has entered the drawing operation. At this time, the main work of the data scheduling thread is to judge the data that may be used for the next drawing according to the data of the new viewpoint, and schedule these data into the memory. As shown in Figure 5, scene block A is the location of the current viewpoint, the white area is the currently visible area, that is, the area that needs to be drawn currently, and areas B, C, and D are prefetch areas. After receiving the viewpoint information (coordinates and direction) sent by the drawing thread, the data scheduling thread updates the viewpoint view area according to the viewpoint information, calculates the area to be prefetched and judges the LOD level corresponding to the prefetched area. The number of the scene block in the prefetch area can be directly calculated according to the number of the visible area. After calculating all the scene blocks in the prefetch area, due to the time limit of each data prefetch (the data prefetch time must be less than the scene drawing time), it is necessary to classify the scheduling priority of these scene blocks, and the basis for judgment It is mainly the direction in which the point of view is heading at the previous moment.

According to the size of the included angle between the scene block and the viewpoint movement direction, the prefetch area is divided into three priorities. As shown in Figure 4, the arrow indicates the movement direction of the viewpoint at the previous moment, the prefetch area B is marked as high priority (its priority is still the data priority of the current scene block in the region), and the prefetch area C is marked as medium priority. Prefetch area D is marked as low priority. The priority Ki of scene block i is as formula (1):

Among them, K _i represents the priority of scene block i; θ _i represents the angle between the scene block and the movement direction of the viewpoint; R represents the set of scene blocks in the prefetching area.

When scheduling the prefetched data into the memory, it is necessary to delete the corresponding size of scene data from the memory. This paper chooses the LRU (least recently used) strategy to select the deleted scene blocks.

C. Growth model solution. After all the data required for pre-calculation are loaded into the memory, the growth equation is solved and the final result is calculated.

3. According to the calculation results of the tree growth model in the large-scale forest scene, visualize the large-scale forest scene, and save the calculation results of the tree growth model in the large-scale forest scene.

Claims (3)

1. a dispatching method for the growth model emulated towards extensive forest, is characterized in that: described dispatching method comprises the following steps:

1) obtain the emulated data of extensive scale Forest Scene according to simulation objectives, carry out initialization to the data structure of emulated data, the scene information data acquisition not comprising tree information stores by the mode of quaternary tree; Tree information data acquisition memory pool technique stores;

2) when carrying out Trees growing models calculating to extensive scale Forest Scene, in units of world subdivision, travel through all world subdivisions, and calculate the arboreal growth in each world subdivision, the concrete steps that arboreal growth calculates are as follows:

A. in scene, determine the current trees that will calculate, and according to the current influence circle scope calculating the property calculation trees of trees, then, according to influence circle scope, calculate relevant simulation parameter, and calculate the relevant trees of trees by the whole scene trees searching of traversal to current;

B. after obtaining relevant trees, needing the data of attribute information of these trees to be all dispatched in internal memory, adopting the scheduling strategy of Forest Growth emulation to dispatch growing deposit data inside and outside in computation process;

C. growth model solves: the data required for current calculating solve growth equation after all calling in internal memory, calculates last result;

3) according to Trees growing models result of calculation in extensive scale Forest Scene, carry out the visual of extensive scale Forest Scene, and preserve the result of calculation of Trees growing models in extensive scale Forest Scene.

2. the dispatching method of the growth model emulated towards extensive forest as claimed in claim 1, it is characterized in that: in described step B, the scheduling strategy of described Forest Growth emulation comprises utilization drafting thread and the internal outer deposit data of data dispatch thread is dispatched;

When user's viewpoint changes, draw thread and calculate current visibility region according to current view point, contextual data in internal memory is judged simultaneously, confirm that scene repaints required data whether in internal memory, if it is directly carry out the drafting of scene and send viewpoint updating message to data prefetching thread; If the scene that current needs are drawn is not in internal memory, drafting thread then needs the position according to current view point, the direction view information of current view point calculates in visibility region; After calculating completes, draw thread suspension and send data request information to data dispatch, proceeding to draw after calling in contextual data by data dispatch thread from external memory;

Data dispatch thread is responsible for data dispatching from external memory, by analyzing the message of drawing thread and passing over, judge the content specifically needing data dispatching, as current scene drawing data, scene prefetch data or scene calculate data, deletion contextual data from internal memory simultaneously;

When data dispatch thread receive draw contextual data request message that thread sends time, directly dispatch, now scheduling thread according to message directly from external memory scheduling draw the data required for thread, data dispatch priority is now the highest;

When data dispatch thread receive draw new viewpoint coordinate information that thread sends time, illustrates that contextual data that current scene draws is all at internal memory, drafting thread has entered drafting and has operated; At this moment, the groundwork of data dispatch thread is then judge to draw the data that may use according to the data of new viewpoint next time, and by these data dispatchs in internal memory; After receiving the information of drawing the viewpoint that thread sends, data dispatch thread is according to the view information more new viewpoint ken, and the region that calculating needs are looked ahead also judges the LOD level corresponding to prefetching areas; The numbering of prefetching areas scenario block directly calculates according to the numbering of visibility region; After the scenario block calculating all prefetching areas, because the data pre-fetching time must be less than the scene drawing time, so need to classify to the dispatching priority of these scenario block, the working direction according to mainly eve viewpoint of judgement;

Prefetching areas is divided into three priority by the corner dimension according to scenario block and viewpoint direction of motion, and scenario block i priority Ki is such as formula (1):

Wherein, K _irepresent the priority of scenario block i; θ _irepresent the angle of scenario block and viewpoint direction of motion; R represents the set of prefetching areas scenario block;

When being dispatched in internal memory by prefetch data, needing the contextual data of deleting corresponding size from internal memory, selecting LRU strategy to select the scenario block of deleting.

3. the dispatching method of the growth model emulated towards extensive forest as claimed in claim 1 or 2, it is characterized in that: described step 1) in, with scene chained list for index, tree information is stored in MemoryBlock, each scenario block in scene chained list may have the tree information that multiple MemoryData comes in storage block, trees in MemoryData are sorted, with the search of trees relevant after accelerating simultaneously.