CN102800116B - Method for rapidly creating large-scale virtual crowd - Google Patents

Abstract

The invention discloses a method for quickly creating a large-scale virtual crowd, which is used to create a virtual crowd in a scene area input by a user, regards people as particles with distribution and role information, and obtains a small-scale crowd from a real picture Samples of particle information, establish a sample library, and use the samples to make a collection of Wang’s bricks based on the Wang’s brick method, and use Wang’s bricks to cover the scene area with aperiodic splicing, so as to quickly generate diverse virtual crowds in scenes of any scale. The creation method of the present invention is simple and fast. The user only needs to input several samples and the scene area that needs to be covered by the crowd, and the creation work can be completed quickly. Different, with authenticity and naturalness.

Description

A method for rapidly creating large-scale virtual crowds

technical field

The invention relates to the field of virtual reality, in particular to the field of creating virtual crowds, which is used to vividly reproduce various crowds in virtual scenes.

Background technique

People are one of the most important and basic elements in the real world, and virtual characters are equally important to the virtual world. Using group simulation technology, create and simulate groups and their movements in the real world in computer-generated spaces , which can greatly improve the sense of reality and immersion in the virtual world. In recent years, with the continuous development of computer group simulation technology, its application has become more and more extensive. The fields involved include simulation training, computer game animation, special effects for film and television, and auxiliary design for public safety. For example, scenes containing a large number of crowd activities can be realized through computer simulation in film and television screens, thereby improving the efficiency of special effects production and reducing production costs; It plays a very good guiding role in the design, passenger flow management of large-scale events, and the formulation of emergency plans for public places.

For thousands of people, if it is necessary to set and modify simulation-related attributes such as position and orientation of each character in turn, it will be a huge and cumbersome task for users. If these attributes are only set randomly, the results will not conform to the characteristics of real people, and it is difficult to control to achieve the desired effect of users. If simple rules are used to arrange the crowd, the results will be too regular or repetitive. At the same time, complex crowd distribution rules are difficult to describe with mathematical equations.

For large-scale scenes and crowd simulations, how to quickly and efficiently complete the initial distribution and state settings of the crowd, and make it diverse and authentic, is an urgent problem to be solved.

In terms of virtual crowd creation, some models have been proposed, such as Reynolds' flocking model (Reynolds C., Flocks, Herds and Schools: A distributed behavioral model [J]. Computer Graphics, 1987, 21(4): 25-34. ), Helbing's social force model (Helbing D., Molnar P., Social force model for pedestrian dynamics. [J]. Physical Review E, 1995, 51(5): 4282-4286.), can be used for simulation generation with Crowd distribution with a certain complexity, but it is difficult to control the final effect by adjusting local rules, and it is also difficult to simulate to achieve user-specified effects. Ulciny's Crowdbrush system (Ulicny B., Ciechomski O. and Thalmann D., Crowdbrush: Interactive authoring of real-time crowdscenes [C]. In Proceedings of ACM SIGGRAPH/Eurographics Symposium on Computer Animation, 2004, 243-252.) provides The method for users to create and draw crowds in real time uses brushes to edit and control character information. Since the operations are more aimed at a single or several characters, a large amount of user interaction is required. Yersin et al proposed a large-scale scene crowd creation and animation method based on Crowd Patch (Yersin B., Maim J., Pettre J. and Thalmann D., Crowd patches: populating large-scale virtual environments for real-time applications [C].In Proceedings of the 2009symposium on Interactive 3D graphics and games, 2009, 207-214.), the final effect is obtained by pre-storing the movement trajectory of the characters in the Patch. In this method, the design of the Patch is complex and needs to meet the requirements of spatial and temporal continuity. , the crowd distribution and animation types that can be simulated are very limited, and at the same time, the user cannot modify the spliced crowd scene. Ju et al proposed a data-driven crowd simulation method (Ju E., Choi M., Park, M., Lee J., Lee K., Takahashi S., Morphablecrowds[J]. ACM Trans.Graph., 2010 , 29(6): Article 140.), classify and learn from real videos to obtain small-scale crowd distribution samples, in the simulation, add individuals and control movement by detecting whether the distribution of each group in the crowd is in the sample set, because it is necessary to detect each The similarity between the adjacent area of an individual and the entire sample set, the synthesis speed of the population is very slow, and the synthesis results are very regular, which is not suitable for the rapid synthesis of large-scale and diverse populations. Guy et al. proposed a method to parametrically describe the behavior characteristics of characters based on psychological models (Guy S.J., Kim S., Lin M.C., Manocha D., Simulating heterogeneous crowd behaviors using personality trait theory[C].In Proceedings of ACM SIGGRAPH /Eurographics Symposium on Computer Animation, 2011, 43-52.), generate a variety of crowd movement effects, and focus on controlling the parameter settings related to character movement. In addition, there are also macroscopic methods based on the global velocity field, including based on map information, user interactive control navigation, based on continuum theory, based on fluid models, etc. These methods can only be used for global motion control effects, and cannot control local crowd distribution effects.

Contents of the invention

The purpose of the present invention is to solve the problem of complicated and too regular operation in virtual crowd, and propose a virtual crowd creation method based on real pictures, which is used to quickly create large-scale virtual crowd.

The present invention provides a method for quickly creating a large-scale virtual crowd, which is used to create a virtual crowd in a scene area input by a user, including steps:

(1) Create a virtual crowd sample library based on real pictures;

(2) Synthesize Wang Tile based on the created virtual crowd sample library, and establish a collection of Wang Tile;

(3) Use the Wang Tile in the Wang brick collection to stitch and cover the scene area input by the user.

Further, said step (1) creates a virtual crowd sample library based on real pictures, specifically including steps:

(1.1) Initialize the sample square;

(1.2) Select the square area of interest from the real picture;

(1.3) Select individuals from the selected square area and collect their distribution information and role attribute information, select a role model that meets the role attribute information from the 3D character model library, and perform position and orientation adjustment on the role model according to the collected distribution information Transform, put into the sample square area, as a sample;

(1.4) Set a different color code for each sample, and put the sample into the sample library.

In order to ensure the authenticity and naturalness of the virtual crowd, the present invention uses real pictures to create a sample library. The larger the sample library, the richer the samples can be provided, so the distribution information and roles of individuals in the real photos can be collected as much as possible It is also feasible to construct as many samples as possible based on the attribute information, and it is also feasible to properly establish a sample library according to the size of the virtual crowd to be produced later.

Further, the step of synthesizing Wang's brick Wang Tile in the described step (2) comprises:

(2.1) Select one sample from the sample library in four times, splicing the four samples to form a larger sample block, there is a certain overlapping area between two samples, and the overlapping side length is 0.1 of the sample side length ~0.35, the diamond-shaped area formed by the intersection of the diagonals of the four sample squares is the Wang Tile;

(2.2) In the rhombus area, Voronoi division is performed on the particle set in the overlapping area of adjacent samples, and a dividing line composed of Voronoi edges is selected so that the distance between the separated particles is the farthest. The left sample particles and the right sample particles on the right side of the dividing line ensure the naturalness of particle distribution in the overlapping area;

(2.3) Rotate the clipped diamond-shaped area and the particles inside it to obtain a square WangTile, whose top, bottom, left, and right sides are marked as N, S, W, and E respectively, and set the color code corresponding to the side of the Wang Tile according to the sample color ;

(2.4), traverse all 4 sample combinations, and establish a Wang's brick set, so that the established Wang's brick set is a complete Wang's brick set.

The present invention establishes a complete set of Wang's bricks by traversing the combination of four samples in the sample library, so that the optional color of each side is the entire sample library.

Further, the sample library includes a first sub-sample library and a second sub-sample library, the samples used for stitching the upper and lower edges are put into the first sample library, and the samples used for the left and right edges are put into the second sample library . When using the sample library to synthesize Wang Tile, the optional colors for the upper and lower sides are only selected from the first sub-sample library, and the optional colors for the left and right sides are only selected from the second sub-sample library. Through the above method, the characteristics of the horizontal and vertical distribution of the crowd can be made different.

Further, the step (3) also includes:

(3.1), the user specifies the polygonal scene area that needs to be covered by the virtual crowd, selects the longest side as the boundary that needs to be aligned to cover, and constructs a rectangular area including the polygonal scene area that the user specifies needs to be covered by the crowd;

(3.2), specify the color on the boundary of the constraint that needs to be aligned;

(3.3), select the Wang Tile layout rectangular area that meets the matching conditions from the constraint boundary;

(3.4), according to the original input scene area, delete the particles located outside it, and obtain the virtual crowd that finally covers the scene area.

The color on the specified boundary of the constraint that needs to be aligned is the color on the uppermost entire boundary and the leftmost entire boundary of the specified rectangular area. The selection of Wang Tile that meets the matching conditions from the constraint boundary to lay out the rectangular area means that the layout direction is from top to bottom and from left to right; the Wang Tile that meets the matching conditions means that the upper color of the selected Wang Tile must match The lower side of the upper Wang Tile is the same, the left side must have the same color as the right side of the left Wang Tile, and the colors of the remaining two sides are randomly assigned by the system, but they are different from the colors of the surrounding Wang Tile.

Preferably, the synthesized Wang Tile is recorded into a four-dimensional index table, and the four components of the four-dimensional index table are respectively the colors of the four samples of the synthesized Wang Tile. When splicing the scene area, it is only necessary to give For the colors of the four sides of the Wang Tile, the upper, lower, left, and right sides, the corresponding Wang Tile can be found from the four-dimensional index table, and splicing can be quickly completed.

The invention discloses a method for quickly creating a large-scale virtual crowd, which regards people as particles with distribution and role information, and establishes a sample library by obtaining samples containing particle information of small-scale crowds from real pictures, and based on Wang's The brick method uses samples to make a collection of Wang's bricks, and uses Wang's bricks to splicing aperiodically to cover the scene area, so as to quickly generate diverse virtual crowds in scenes of any scale. The creation method of the present invention is simple and fast, and the user only needs to input a number of samples and scene areas that need to be covered by crowds, and the creation work can be quickly completed.

The present invention makes samples based on people with real and complex characteristics in real pictures, and collects a plurality of different samples to form a sample library. Virtual crowds provide great convenience for the creation of virtual crowds, and at the same time have authenticity and naturalness.

Description of drawings

Fig. 1 is a flow chart of the method for quickly creating large-scale virtual crowds in the present invention;

Fig. 2 is a flow chart of the steps of setting up a sample library in the present invention;

Fig. 3 is the flow chart of the steps of setting up Wang Tile for the present invention;

Fig. 4 is a flow chart of the steps of splicing Wang Tile according to the present invention.

Detailed ways

The technical solution of the present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments, and the following embodiments do not constitute a limitation of the present invention.

In order to quickly create a large-scale virtual crowd, the present invention first establishes a virtual crowd sample library based on real pictures, then uses the sample library to build Wang Tile, and finally splices the Wang Tile to form a large-scale virtual crowd. The specific implementation method is shown in Figure 1, including the following steps:

Step 101, create a virtual crowd sample library based on real pictures.

Step 102: Synthesize Wang Tile based on the created virtual crowd sample library, and establish a collection of Wang Tile.

Among them, the concept of Wang’s brick was proposed by mathematical logician Wang Hao in 1961. It is a group of square tiles of the same size. Each side of the tile is given a color. If the adjacent sides of the two tiles have the same color, it can be Arrange the two tiles together.

The synthesized Wang Tile is marked with N, S, W, and E on its top, bottom, left, and right sides, and the color code corresponding to the side of the Wang Tile is set according to the sample color. The established Wang's brick set is a complete Wang's brick set, that is, the optional color of each side is the entire sample library.

Step 103: Use Wang Tile in the Wang tile set to stitch and cover the scene area input by the user to obtain a crowd particle set in the scene area.

That is, according to the scene area input by the user, the aperiodic fast splicing and coverage of Wang's bricks is used to obtain the crowd particle set in the scene.

In order to ensure the authenticity and naturalness of virtual crowds, the present invention uses real pictures to create a sample library, as shown in Figure 2, creating a virtual crowd sample library based on real pictures includes the following steps:

Step 201: Initialize the sample square, set the side length to 1 unit length, and scale the included crowd location information to 0-1.0 unit length.

Step 202. Select a square area of interest from the real picture.

Step 203: Select individuals from the selected square area and collect their distribution information and role attribute information, select a character model that conforms to the character attribute information from the 3D character model library, and perform position and orientation adjustment on the character model according to the collected distribution information Transform, put into the sample square area, as a sample.

Specifically, each individual in the crowd is regarded as a particle, and each particle contains its initial distribution information and role attribute information. The initial distribution information is the position P(P _x , P _y ) and the orientation O(O _x , O _y ). The character attribute information is mainly the attribute information of the 3D model of the character, including gender, age, and pose (that is, the corresponding action sequence ). After extracting the information, apply the information to qualified character models, and then put these character models applied with the individual distribution information and character attribute information in the real photos into the sample square area to form a sample.

Step 204, setting a different color code for each sample, and putting the samples into the sample library.

Specifically, repeat steps 201 to 204 to form a sample library composed of multiple samples. The larger the sample library, the richer the samples can be provided. Therefore, the distribution information and role attribute information of individuals in real photos can be collected as much as possible to construct It is also feasible to build as many samples as possible according to the size of the virtual crowd to be produced later.

After the sample library is created, each sample has a color code, and then Wang’s bricks can be synthesized based on the created sample library, and the distribution of particles in the overlapping area of adjacent samples inside the Wang Tile can be processed, as shown in Figure 3 , including the following steps:

Step 301, select one sample from the sample library in four times, splice the four samples to form a larger sample block, there is a certain overlapping area between two samples, and the diagonal lines of the four sample squares intersect each other to form a The diamond-shaped area of is the Wang Tile, and the side length of the overlapping part is Overlap_size∈(0.1, 0.35).

Step 302, in the diamond-shaped area, perform Voronoi division on the particle set in the overlapping area of adjacent samples, select a dividing line composed of Voronoi edges, so that the distance between the separated particles is the farthest, and respectively retain the particles located on the left side of the dividing line The left sample particle and the right sample particle located on the right side of the dividing line ensure the naturalness of the particle distribution in the overlapping area.

Step 303: Rotate the cropped diamond-shaped area and the particles inside it to obtain a square Wang Tile, whose upper, lower, left, and right sides are marked as N, S, W, and E respectively, and set the color corresponding to the side of the Wang Tile according to the sample color coding.

Step 304, traversing all 4 sample combinations to establish a Wang's brick set, so that the established Wang's brick set is a complete Wang's brick set, that is, the optional color of each edge is the entire sample library.

In order to obtain a complete set of Wang tiles, it is necessary to traverse each combination of 4 samples in the sample library, and record the Wang Tile synthesized into a four-dimensional index table, the four components of which are 4 The color of the four samples, that is, the color of the N/S/W/E four sides of the Wang Tile synthesized by these four samples. In the next step, you only need to give the colors of the four sides N/S/W/E, and you can find the corresponding Wang Tile in the four-dimensional index table.

The complete collection of Wang tiles lays the foundation for large-scale creation of virtual crowds. In the virtual scene, the user specifies a polygonal area that needs to be covered by the crowd, selects the longest side as the boundary that Wang Tile splicing needs to align, and constructs a polygon that covers the polygonal area. rectangular area. Then calculate the Wang Tile number H_NUM required for each row and the Wang Tile number C_NUM required for each column in the rectangular area according to the length and width of the rectangular border and the border length of Wang Tile. In the rectangular coverage area, from Wang Tile Randomly select (or partially specify) H_NUM×C_NUM tiles from the brick collection to arrange and splice according to the matching rules. After all Wang tiles are put together, there are HNUM+1 lines in the horizontal side (N/S side), and each line is C NUM , represented by a two-dimensional matrix B _N [H_NUM+1][C_NUM]; the vertical side (W/E side) has a total of C NUM+1 columns, each column HNUM, with a two-dimensional matrix B _W [C_NUM +1][H_NUM] for that.

The random aperiodic splicing steps are shown in Figure 4, including steps:

Step 401, the user specifies the polygonal scene area that needs to be covered by virtual crowds, selects the longest side as the boundary that needs to be aligned for coverage, and constructs a rectangular area including the polygonal scene area that needs to be covered by virtual crowds specified by the user;

Step 402, specify the color on the constraint boundary that needs to be aligned, usually specify the color on the uppermost and leftmost entire boundary of the rectangular area, that is, the uppermost row B _N [H_NUM+1] in the two-dimensional matrix B _N and the color value of the leftmost column B _W [C_NUM+1] in the two-dimensional matrix B _W. Boundary constraints can be specified by the user or automatically by the system.

Step 403: Select the Wang Tile laying area that meets the matching conditions from the constraint boundary, and the laying direction is from N to S, and from W to E. The Wang Tile that meets the matching conditions is: the color of the N side of the selected WangTile must be the same as the S side of the Wang Tile above it, the W side color must be the same as the E side of the Wang Tile to the left, and the colors of the remaining two sides can be randomly specified by the system , but try to be different from the color of the side of the Wang Tile that has been laid out around. When the colors of the four sides of the Wang Tile have been determined, the Wang Tile is obtained from the index table INDEX and placed in the layout area. After laying out, get the crowd particles CrowdParticles in the rectangular area.

Step 404: According to the original input scene area, delete the particles located outside it, if there are obstacles in the scene area, delete the particles located within the obstacle polygon, and finally obtain the crowd particles CrowdParticles covering the scene.

It should be noted that since the tiles in the Wang brick set are stored in the four-dimensional index table according to the colors of their four sides as components of the four-dimensional index table, the colors of the four sides (N/S/W/E) can be used to quickly search for other tiles. Position in the index table to find the Wang Tile you want to use to stitch and cover the scene area.

What needs to be explained here is that since the N side of the sample is only spliced with the S side, and the W side is only spliced with the E side, in order to be able to flexibly generate crowds, an optional sub-sample library 1 can also be set for the N/S side, and Set up another subsample bank 2 for the W/E sides. When the sample library further includes a sub-sample library, the combination method of selecting samples when synthesizing Wang Tile is changed, that is, when synthesizing Wang Tile, the optional colors of N/S sides are only selected from sub-sample library 1, W /E Side optional colors are selected from subsample library 2 only.

Users can also independently select desired samples from the sample library according to the characteristics of the samples and put them into sub-sample library 1 and sub-sample library 2 (for example: the population density in sub-sample library 1 is relatively large, and the population density in sub-sample library 2 is relatively high). smaller, etc.). Through the above method, the characteristics of the horizontal and vertical distribution of the crowd can be made different.

The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and changes according to the present invention. deformation, but these corresponding changes and deformations should belong to the scope of protection of the appended claims of the present invention.

Claims (8)

1. A method for quickly creating a large-scale virtual crowd, for creating a virtual crowd in a scene region input by a user, characterized in that it comprises steps: (1) Create a virtual crowd sample library based on real pictures, including steps: (1.1) Initialize the sample square; (1.2) Select the square area of interest from the real picture; (1.3) Select individuals from the selected square area and collect their distribution information and role attribute information, select a role model that meets the role attribute information from the 3D character model library, and perform position and orientation adjustment on the role model according to the collected distribution information Transform, put it into the sample square area, as a sample: specifically, each individual in the crowd is regarded as a particle, and each particle contains its initial distribution information and role attribute information; the initial distribution information is the position and orientation, and the role Attribute information includes gender, age, and posture; after extracting this information, apply this information to qualified character models, and then put these character models applied with individual distribution information and character attribute information in real photos into the sample square area , forming a sample; (1.4) Set a different color code for each sample, and put the sample into the sample library; (2) Based on the created virtual crowd sample library, Wang Tile is synthesized to set up a collection of Wang bricks, and the steps of said synthetic Wang Tile include: (2.1) Select one sample from the sample library in four times, splicing the four samples to form a larger sample block, there is a certain overlapping area between two samples, and the overlapping side length is 0.1 of the sample side length ~0.35, the diamond-shaped area formed by the intersection of the diagonals of the four sample squares is the Wang Tile; (2.2) In the rhombus area, Voronoi division is performed on the particle set in the overlapping area of adjacent samples, and a dividing line composed of Voronoi edges is selected so that the distance between the separated particles is the farthest. The left sample particles and the right sample particles on the right side of the dividing line ensure the naturalness of particle distribution in the overlapping area; (2.3) Rotate the cropped diamond-shaped area and the particles inside it to obtain a square Wang Tile, whose upper, lower, left, and right sides are marked as N, S, W, and E respectively, and set the color of the corresponding Wang Tile edge according to the sample color coding; (2.4), traverse all 4 sample combinations, set up the Wang's brick set, so that the Wang's brick set set up is a complete Wang's brick set; (3) Use the Wang Tile in the Wang brick collection to stitch and cover the scene area input by the user. 2. The method for creating a large-scale virtual crowd as claimed in claim 1, characterized in that: the sample library includes a first sub-sample library and a second sub-sample library, and the samples for splicing the upper and lower edges are put into the first sub-sample library. A sub-sample bank, the samples used for splicing the left and right sides are put into the second sub-sample bank. 3. the method for creating large-scale virtual crowd as claimed in claim 2, is characterized in that: when adopting described sample storehouse to synthesize Wang Tile, the optional color of upper and lower sides only selects from the first sub-sample storehouse, left and right sides can be selected. The selected color is selected only from the second subsample library. 4. the method for creating large-scale virtual crowd as claimed in claim 1, is characterized in that: described step (3) also comprises: (3.1), the user specifies the polygonal scene area that needs to be covered by the virtual crowd, selects the longest side as the boundary that needs to be aligned to cover, and constructs a rectangular area including the polygonal scene area that the user specifies needs to be covered by the virtual crowd; (3.2), specify the color on the boundary of the constraint that needs to be aligned; (3.3), select the Wang Tile layout rectangular area that meets the matching conditions from the constraint boundary; (3.4), according to the original input scene area, delete the particles located outside it, and obtain the virtual crowd that finally covers the scene area. 5. The method for creating a large-scale virtual crowd as claimed in claim 4, characterized in that: the specified color on the boundary of the constraint that needs to be aligned is the uppermost entire line and the leftmost entire line of the specified rectangular area on the color. 6. The method for creating large-scale virtual crowds as claimed in claim 4, characterized in that: the selection of Wang Tile that satisfies matching conditions from the constraint boundary to lay out the rectangular area refers to that the laying direction is from top to bottom, from left to right Right; the Wang Tile that meets the matching conditions means that the top color of the selected Wang Tile must be the same as the bottom side of the Wang Tile above it, the left side color must be the same as the right side of the Wang Tile to the left, and the colors of the remaining two sides are randomized by the system Specified, but not the same color as the edges of the surrounding tiled Wang Tile. 7. The method for creating a large-scale virtual crowd as claimed in claim 1, characterized in that, the method also includes the step of: recording the synthetic Wang Tile into a four-dimensional index table, the four-dimensional index table The components are the colors of the 4 samples for synthesizing the Wang Tile. 8. The method for creating a large-scale virtual crowd as claimed in claim 7, wherein the selection of the Wang Tile satisfying the matching condition refers to the colors of the four sides of the Wang Tile, up, down, left, and right, from the described Wang Tile Find the corresponding Wang Tile in the four-dimensional index table.